No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Comparative Analysis of Levenberg-Marquardt and Bayesian Regularization Backpropagation Algorithms in Photovoltaic Power Estimation Using Artificial Neural Network
Abstract
This paper presents a comparative analysis of Levenberg-Marquardt (LM) and Bayesian Regularization (BR) backpropagation algorithms in development of different Artificial Neural Networks (ANNs) to estimate the output power of a Photovoltaic (PV) module. The proposed ANNs undergo training, validation and testing phases on 10000+ combinations of data including the real-time measurements of irradiance level (W/m2) and PV output power (W) as well as the calculations of the Sun’s position in the sky and the PV module surface temperature (°C). The overall performance of the LM and the BR algorithms are analyzed during the development phases of the ANNs, and also the results of implementation of each ANN in different time intervals with different input types are compared. The comparative study presents the trade-offs of utilizing LM and BR algorithms in order to develop the best ANN architecture for PV output power estimation.
... It is concluded, that the Bayesian-Regularization algorithm produces more accurate results at the expense of higher training time, than the Levenberg-Marquardt algorithm. Based on this Jazayeri et al. (2016) compare and compete the Levenberg-Marquardt algorithm and the Bayesian-Regularization algorithm for power estimation of photovoltaic modules. The better accuracy at the expense of higher training time by using the Bayesian-Regularization algorithm can also be approved for the described use case. ...
... The default set and frequently recommended Levenberg-Marquardt algorithm is used to determine the most suitable input variant (Jazayeri, Jazayeri & Uysal, 2016;Payal, Rai & Reddy, 2013). The training is partly random and therefore does not deliver uniform results. ...
... Analogous to Jazayeri et al. (2016), the resulting functions are evaluated on the basis of the performance metrics achieved. A relatively small deviation can be seen in the data. ...
The increasing electrification of powertrains leads to increased demands for the test technology to ensure the required functions. For conventional test rigs in particular, it is necessary to have knowledge of the test technology's capabilities that can be applied in practical testing. Modelling enables early knowledge of the test rigs dynamic capabilities and the feasibility of planned testing scenarios. This paper describes the modelling of complex subsystems by experimental modelling with artificial neural networks taking transmission efficiency as an example. For data generation, the experimental design and execution is described. The generated data is pre-processed with suitable methods and optimized for the neural networks. Modelling is executed with different variants of the inputs as well as different algorithms. The variants compare and compete with each other. The most suitable variant is validated using statistical methods and other adequate techniques. The result represents reality well and enables the performance investigation of the test systems in a realistic manner.
... The combinatorial optimisation approach is also used with two different ANN training algorithms: the Levenberg-Marquardt algorithm (LM) (Marquardt 1963) and the Bayesian regularisation algorithm (BR) (MacKay 1992). The Levenberg-Mardquart (LM) algorithm was designed for a fast convergence of back-propagation processes and to deal with moderate-sized problems (Kayri 2016;Jazayeri et al. 2016). A detailed description of the mathematical basis of the LM back-propagation algorithm is available in Jazayeri et al. (2016); Wilamowski and Yu (2010); Marquardt (1963). ...
... The Levenberg-Mardquart (LM) algorithm was designed for a fast convergence of back-propagation processes and to deal with moderate-sized problems (Kayri 2016;Jazayeri et al. 2016). A detailed description of the mathematical basis of the LM back-propagation algorithm is available in Jazayeri et al. (2016); Wilamowski and Yu (2010); Marquardt (1963). The Bayesian Regularisation (BR) proposed by MacKay (1992) is another well-known algorithm used in the training of ANN. ...
... The verification subset is not required in the BR algorithm [ Fig. 3b]. (Jazayeri et al. 2016;MacKay 1992;Buntine and Weigend 1991). As in the case of the LM algorithm, the data division is carried out by means of a interleaving procedure as it is shown in Fig. 4(b). ...
Ensemble modelling is a numerical technique used to combine the results of a number of different individual models in order to obtain more robust, better-fitting predictions. The main drawback of ensemble modeling is the identification of the individual models that can be efficiently combined. The present study proposes a strategy based on the Random-Restart Hill-Climbing algorithm to efficiently build ANN-based hydrological ensemble models. The proposed technique is applied in a case study, using three different criteria for identifying the model combinations, different number of individual models to build the ensemble, and two different ANN training algorithms. The results show that model combinations based on the Pearson coefficient produce the best ensembles, outperforming the best individual model in 100% of the cases, and reaching NSE values up to 0.91 in the validation period. Furthermore, the Levenberg-Marquardt training algorithm showed a much lower computational cost than the Bayesian regularisation algorithm, with no significant differences in terms of accuracy.
... The block diagram consistof PV cell, DC-DC boost converter, PWM generator and with ANN controller is shown in Fig.1 [6]. It consist of Artificial Neural Network (ANN) having input G (radiance),T(temperature), V pv (voltage across PV module), I pv (current from PV module), PWM generator which will generate duty cycle D. ...
... Three algorithms Levenberg Marquardt, Bayesian Regularization and Scaled Conjugate Gradient is used. [6] Comparative analysis has been done for these three methods and discussed in SIMULATION section. To extract the maximum power from PV cell Maximum Power Point Tracking (MPPT) algorithm has been implemented by using these three ANN techniques. ...
... BR algorithm does not require cross-validation. [6] Overtrain the BR algorithm is difficult since evidence procedures enable a Bayesian objective criterion for terminating training. Because the BRANN calculates and trains on a number of effective network characteristics or weights, essentially turning off those that are not significant, they are also challenging to overfit. ...
This paper presents the comparative analysis of Artificial Neural Network (ANN) based algorithms in maximum power point tracking (MPPT) for solar photovoltaic system. The algorithms deployed in this paper are Bayesian Regularization (BR), Levenberg- Marquardt (LM) and Scaled Conjugate Gradient algorithm (SCG).The MPPT model for solar photovoltaic system was designed in MATLAB/Simulink environment and ANN toolbox was used to for analysis. For training 70% data was used and rest 30% was used for validation and testing purpose, which was 15% each. The proposed model was trained seven times for each algorithm and best result was taken. The performance of BR algorithm was better in terms of mean square error which was less than LM algorithm. But with LM algorithm the learning rate, thus time required for training is less so it can be preferred over Time constrained system. SCG algorithm trained the system perfectly with low performance hence it is not suitable for MPPT module. Solar module of 200W with 2 modules in series and 1 module in parallel were taken. The output generated from the trained MPPT solar energy system was 400 W.
... It is concluded, that the Bayesian-Regularization algorithm produces more accurate results at the expense of higher training time, than the Levenberg-Marquardt algorithm. Based on this Jazayeri et al. (2016) compare and compete the Levenberg-Marquardt algorithm and the Bayesian-Regularization algorithm for power estimation of photovoltaic modules. The better accuracy at the expense of higher training time by using the Bayesian-Regularization algorithm can also be approved for the described use case. ...
... The default set and frequently recommended Levenberg-Marquardt algorithm is used to determine the most suitable input variant (Jazayeri, Jazayeri & Uysal, 2016;Payal, Rai & Reddy, 2013). The training is partly random and therefore does not deliver uniform results. ...
... Analogous to Jazayeri et al. (2016), the resulting functions are evaluated on the basis of the performance metrics achieved. A relatively small deviation can be seen in the data. ...
The increasing electrification of powertrains leads to increased demands for the test technology to ensure the required functions. For conventional test rigs in particular, it is necessary to have knowledge of the test technology's capabilities that can be applied in practical testing. Modelling enables early knowledge of the test rigs dynamic capabilities and the feasibility of planned testing scenarios. This paper describes the modelling of complex subsystems by experimental modelling with artificial neural networks taking transmission efficiency as an example. For data generation, the experimental design and execution is described. The generated data is pre-processed with suitable methods and optimized for the neural networks. Modelling is executed with different variants of the inputs as well as different algorithms. The variants compare and compete with each other. The most suitable variant is validated using statistical methods and other adequate techniques. The result represents reality well and enables the performance investigation of the test systems in a realistic manner.
... Payal [15] compares the Levenberg-Marquardt algorithm with the Bayesian-Regularization algorithm for efficient localization in wireless sensor network. Based on this, Jazayeri [16] describes the comparison of the Levenberg-Marquardt algorithm with the Bayesian-Regularization algorithm for power estimation of photovoltaic modules. Both works describe better results of the Bayesian-Regularization algorithm at longer computation time. ...
... Both works describe better results of the Bayesian-Regularization algorithm at longer computation time. It is recommended to use the Levenberg-Marquardt algorithm for time-critical applications, otherwise the Bayesian-Regularization algorithm [15,16]. ...
... The training of the neural network is done using the MATLAB's Deep Learning Toolbox. As in [16], the performance metrics of the algorithms are compared in Table 1. The values highlighted in green are the best. ...
The increasingly noticeable effects of climate change require action to reduce greenhouse gas emissions. In the transport sector, the increase in electromobility is an important tool for reducing CO2 emissions. In order to meet the essential requirements, a particularly detailed system knowledge, for example on energy efficiency in the powertrain, is needed for the development and improvement of electric vehicles.
The present work contributes to this by mapping the efficiency of a planetary gearbox for an electric vehicle. The mapping is done by experimental modelling with machine learning methods based on a standard efficiency description. The data necessary for the neural network training is generated by efficiency experiments on the powertrain test rig. For suitable results the data was pre-processed by applying relevant low-pass filters. A comparison of the Bayesian-Regularization algorithm, the Levenberg-Marquardt algorithm and the scaled-conjugate-gradient algorithm for training exhibited the strengths and weaknesses of the individual algorithms for the efficiency mapping. By comparing each algorithms performance metrics, the one matching the requirements best is chosen for the efficiency mapping.
The result is a continuous function that determines the efficiency based on speed- and torque-inputs. Based on a key performance indicator, statistical validation by evaluating the standard deviations is used to ensure the quality of the results. In this paper the suitable algorithm for the given use case was determined. It can be applied for further research due to the significant results shown.
... Not only is the aim to minimize the mean square error, but also to achieve it with the lowest possible weights. The objective function takes the following form [39,40]: Figure 2. Example of structure of the feedforward multilayer perceptron (MLP) network. ...
... Not only is the aim to minimize the mean square error, but also to achieve it with the lowest possible weights. The objective function takes the following form [39,40]: ...
... The α factor enforces low weight values which greatly reduces the tendency of the network to over-fit. This modification also provides greater immunity to noise and incorrect input data, but is more time-consuming [39,40]. ...
Permanent magnet synchronous motors (PMSMs) are becoming more popular, both in industrial applications and in electric and hybrid vehicle drives. Unfortunately, like the others, these are not reliable drives. As in the drive systems with induction motors, the rolling bearings can often fail. This paper focuses on the possibility of detecting this type of mechanical damage by analysing mechanical vibrations supported by shallow neural networks (NNs). For the extraction of diagnostic symptoms, the Fast Fourier Transform (FFT) and the Hilbert transform (HT) were used to obtain the envelope signal, which was subjected to the FFT analysis. Three types of neural networks were tested to automate the detection process: multilayer perceptron (MLP), neural network with radial base function (RBF), and Kohonen map (self-organizing map, SOM). The input signals of these networks were the amplitudes of harmonic components characteristic of damage to bearing elements, obtained as a result of FFT or HT analysis of the vibration acceleration signal. The effectiveness of the analysed NN structures was compared from the point of view of the influence of the network architecture and various parameters of the learning process on the detection effectiveness.
... BRBNN is a robust form of Artificial Neural Network (ANN) that minimises overfitting or overtraining while providing an efficient model. The network utilises Back Propagation Networks (BPNs) to establish connections and adjust weights, resulting in minimised errors (Burden and Winkler 2009;Jazayeri, Jazayeri, and Uysal 2016;Fiorentini, Pellegrini, and Losa 2023). ...
... Our methodology revolves around the training phase, which involves using a BRBNN. We chose this model because it has been proven effective in mitigating overfitting the training data and reducing biases in neural networks by preventing excessive complexity (Burden and Winkler 2009;Jazayeri, Jazayeri, and Uysal 2016;Fiorentini, Pellegrini, and Losa 2023;Sariev and Germano 2020), which is crucial for managing passenger demand data's complex and dynamic nature. The BRBNN architecture consists of an input layer tailored to our dataset's specific features, the hidden layer that captures the nonlinear relationships, and an output layer that predicts passenger demand. ...
Autonomous Taxis (ATs) have seen remarkable global proliferation in recent years owing to the widespread adoption and advancements in Artificial Intelligence (AI) across various domains. ATs play a crucial role in Intelligent Transportation Systems (ITS) in smart cities. However, the effectiveness of ITS relies heavily on accurately forecasting the passenger demand for ATs, which poses a significant challenge. Precise prediction of passenger demand is essential for minimising waiting times and unnecessary cruising of ATs in metropolitan areas, which helps conserve energy. To address this issue, this study proposed an adaptive Bayesian Regularisation Backpropagation Neural Network (BRBNN) augmented with a Machine Learning (ML) model to predict passenger demand in different regions of metropolitan cities specifically for ATs. The study conducted extensive simulations using a real‐world dataset collected from 4781 taxis in Bangkok, Thailand. Using MATLAB2022b, the proposed model compared various state of art methods and existing research. The results indicate that proposed model outperforms existing methods in terms of performance metrics such as Root Mean Square Error (RMSE) and R‐squared (R2) for passenger demand forecasting. These findings validated the effectiveness of the prediction model and its ability to accurately forecast passenger demand for ATs, thereby contributing to the advancement of efficient transportation systems in smart cities.
... For the purpose of this study, a total of 258 experimental data were collected from the literature (see Appendix A). These data are taken from the tests conducted by Janss (1974), Lin (1988), Luksha and Nesterovich (1991), O'Shear and Bridge (1994, 1996, 2000, Kato (1995Kato ( , 1996, Saisho et al. (1999), Huang et al. (2002), Kang et al. (2002), Johansson (2002), Yamamoto et al. (2002), Yu et al. (2002Yu et al. ( , 2007, Yao (2003, 2004), Giakoumelis and Lam (2004), Gu et al. (2004), Sakino et al. (2004), Gardner and Jacobson (1967), Zhang and Wang (2004), Han et al. (2005), Tan (2006) Table 2. ...
... Bayesian regularization divides the datasets into two subsets: training and test cases. This algorithm eliminates network weights with small effects on the solution and shows super performance by avoiding local minimums (Jazayeri et al. 2016). In this analysis, 258 datasets of the CCFT columns were divided into two parts, in which 220 datasets are used for training (85%) and 38 datasets are used for testing (15%). ...
This paper presents an efficient approach to generate a new empirical formula to predict the axial compression capacity (ACC) of circular concrete-filled tube (CCFT) columns using the artificial neural network (ANN). A total of 258 test results extracted from the literature were used to develop the ANN models. The ANN model having the highest correlation coefficient (R) and the lowest mean square error (MSE) was determined as the best model. Stability analysis, sensitivity analysis, and a parametric study were carried out to estimate the stability of the ANN model and to investigate the main contributing factors on the ACC of CCFT columns. Stability analysis revealed that the ANN model was more stable than several existing formulae. Whereas, the sensitivity analysis and parametric study showed that the outer diameter of the steel tube was the most sensitive parameter. Additionally, using the validated ANN model, a new empirical formula was derived for predicting the ACC of CCFT columns. Obviously, a higher accuracy of the proposed empirical formula was achieved compared to the existing formulae.
... To enhance the existing LDA/GSVD algorithm, the Artificial Neural Network (ANN) can be utilized. The used of ANNs in several real-world purposes is because of their capability concerning resiliency and stableness even in noisy data and fault tolerance [59]. Thus, the widely employed method is Back Propagation Neural Network (BPNN). ...
... BRBP offers robust approximation for difficult and noisy inputs. Thus, it works excellently by removing network weights which have no impact on the problem solving and presents improvements on evading the problems of local minima [59]. Furthermore, it delivers weights into a training function while advancing the simplification performance of the old BPNN automatically [60]. ...
Managing the bandwidth in campus networks becomes a challenge in recent years. The limited bandwidth resource and continuous growth of users make the IT managers think on the strategies concerning bandwidth allocation. This paper introduces a mechanism for allocating bandwidth based on the users' web usage patterns. The main purpose is to set a higher bandwidth to the users who are inclined to browsing educational websites compared to those who are not. In attaining this proposed technique, some stages need to be done. These are the preprocessing of the weblogs, class labeling of the dataset, computation of the feature subspaces, training for the development of the ANN for LDA/GSVD algorithm, visualization, and bandwidth allocation. The proposed method was applied to real weblogs from university's proxy servers. The results indicate that the proposed method is useful in classifying those users who used the internet in an educational way and those who are not. Thus, the developed ANN for LDA/GSVD algorithm outperformed the existing algorithm up to 50% which indicates that this approach is efficient. Further, based on the results, few users browsed educational contents. Through this mechanism, users will be encouraged to use the internet for educational purposes. Moreover, IT managers can make better plans to optimize the distribution of bandwidth.
... It is also used in several mathematical calculations [36]. ANNs are applied in several real-world purposes because of their capability concerning resiliency and stableness even in noisy data and for its fault tolerance [37]. Thus, the widely employed method is Back Propagation Neural Network (BPNN). ...
... BRBP offers robust approximation for difficult and noisy inputs. Thus, it works excellently by removing network weights which have no impact on the problem solving and presents improvements on evading the problems of local minima [37]. Furthermore, it delivers weights into a training function while advancing the simplification performance of the old BPNN automatically [38]. ...
Artificial Neural Networks (ANN) form a dynamic architecture for machine learning and have attained significant capabilities in various fields. It is a combination of interrelated calculation elements and derives outputs for new inputs after being trained. This study introduced a new mechanism utilizing ANN which was trained using Bayesian Regularization Back Propagation (BRBP) to improve the computational cost problem of the existing algorithm of the Generalized Singular Value Decomposition-based Linear Discriminant Analysis (LDA/GSVD). The proposed approach can minimize the number of iterations and mathematical processes of the existing LDA/GSVD algorithm which suffers time complexity. Through simulation using BLE RSSI Dataset from UCI which has 105 classes and 13 dimensions with 1420 instances, it was found out that ANN improved the computational cost during the classification of the data up to 57.14% while maintaining its accuracy. This new technique is recommended when classifying big data, and for pattern analysis as well. © International Journal on Advanced Science, Engineering and Information Technology.
... It can also be integrated with conventional control techniques, utilizing ANN linguistic systems capable of resolving complex relationships between independent and dependent variables. On the other hand, ANN boasts parallel computing capabilities through pattern learning [8]. It can generate a knowledge pattern through self-regularization or learning abilities. ...
In order to attain equilibrium between energy supply and demand, reliance on conventional methods for precise long-term electricity demand forecasting is no longer viable. The utilization of artificial intelligence, such as fuzzy logic and artificial neural network (ANN) models, emerges as a prospective solution in the current dynamic scenario. This research explores long-term electricity demand forecasting within the Jakarta distribution grid system, employing various fuzzy logic and ANN approaches including Sugeno, Mamdani, Bayesian Regularization, and the Levenberg algorithm. The analysis incorporates time series data spanning 2016 to 2019, encompassing electricity load demand, economic factors, and demographic variables, processed using MATLAB. The outcomes of the four forecasting methods reveal an average error range of 2 to 3%. The findings indicate that employing fuzzy logic and ANN methods for long-term electricity demand forecasting can yield a forecast error of less than 3%. The study recommends future research enhancements through the inclusion of additional time series data and a more refined system.
... The results of the ANN model developed were used to simulate new data to perform a sensitivity analysis. Each data input varied between its lowest (L), mean (M) and highest (H ) values (Jazayeri et al., 2016). On the contrary, the other input parameters have not been changed with respect to their reference value. ...
The majority of existing findings regarding expansion risks in concretes containing waste glass stem from experimental studies. There is a need for rapid assessment methods to ensure safer recycling of glass waste in cementitious composites. In this study, an artificial neural network (ANN) model was developed to accurately predict ASR expansion/mitigation resulting from the integration of glass waste in mortars. The analysis considered glass incorporation either separately as waste glass powder (WGP) and waste glass aggregates (WGA), or in combination, at contents of up to 100% for WGA and 30% for WGP. A set of 175 mixtures was analyzed, considering five distinct variables, which encompassed different mix proportions, involving varying components of cement, natural aggregates, WGP, and WGA, in addition to the duration of environmental exposure. The results show that the expansion of WGA-mortars decreased with the increased incorporation of WGP. The expansion values obtained from validation and experience confirm the high accuracy of the developed ANN model, with validation coefficients reaching up to 98.061% and small value of the mean squared error (MSE).
... The LMB model combines the gradient descent method with the Gauss-Newton method, resulting in higher convergence rates (Sapna et al., 2012;Singh et al., 2007). Both LMB and BRB are effective in handling noisy data (Mahapatra and Sood, 2012;Kayri, 2016;Payal et al., 2013;Wali and Tyagi, 2020;Jazayeri et al., 2016), contributing to their superior performance compared to the other models. On the other hand, the RB and CGBPR models exhibit the least accurate predictions among the 8 models, with R 2 values of 0.806 and 0.814, respectively. ...
Higher Himalayan catchments are often poorly monitored for hydrological activities involving flood flow prediction for the safety of riverside communities and the successful operation of hydropower projects. This study aimed to estimate the comparative performance of artificial neural network (ANN) based flow prediction models using 10 years of daily river flow data of Kaligandaki catchment at Kotagaun, Nepal, which is a snow-fed catchment in the Himalayan region. The flow prediction models were trained and tested at a hydrological station using the previous 3 days' river flow data to predict the 1-day ahead flow data. Eight different training functions were employed in an ANN model for comprehensive statistical assessment of accuracy and precision of each training function. The most significant and validated result obtained in this study is the comprehensive comparison of various training functions' performance, and identification of the most efficient training function for the study case. Among the training functions investigated, the Levenberg-Marquardt backpropagation function exhibits the best performance for the model having Nash-Sutcliffe efficiency, root mean square error and mean absolute error values of 0.866, 209.578 and 75.422, respectively. This study provides a fundamental basis for accurate flow prediction of topographically challenged catchments where hydrological monitoring and data collection may be limited. In particular, this model will help to improve early warning system, hydrological planning, and the safety of riverside communities in the Himalayan region.
... The Levenberg-Marquardt and Bayesian algorithms for regulation have shown robustness and effectiveness in predicting highly nonlinear characteristics, such as solar radiation and wind speed [54]. Moreover, they provided precise estimates of photovoltaic power where the BR showed good accuracy while the LM offered rapid performance [55]. NARX has proven high SOH estimation precision with untrained data [56]. ...
The lithium-ion battery (LiB) has become the most widely used energy storage system for electric vehicles (EVs) due to its many advantages. The EV battery pack needs a battery management system (BMS) to estimate the state of charge (SOC) and balance the energy capacity through the cells. Apart from the fact that it is still challenging to accurately solve, the SOC forecasting represents an important concern in the study sector. This research proposes an effective battery SOC forecasting approach utilizing the non-linear autoregressive exogenous model (NARX) time’s series optimized Levenberg-Marquardt training algorithm, and Bayesian-Regularization (BR). The suggested technique is well-known for its resilience and high performance in nonlinear and complex system prediction, and it is extensively used in a wide range of disciplines. Also, the precision of the NARX technique has been investigated as a function of training data sets, error classifications based on experimental data of LiB. Both algorithms were evaluated with experimental data. Discharging followed by resting process was conducted on a 2.6 Ah LiB. They demonstrate good convergence in the low error and regression. In an effort to address a gap in the field, this paper offers a comparison between NARX-LM and NARX-BR algorithms for the LiB SOC prediction. Both algorithms are optimized the ANN using times series analysis based in the same training data. The results show that NARX-BR is more rapid and accurate with a low mean square error (MSE) of 2.39 10-5 than NARX-LM, which achieved an MSE of 1.11. Thus, it shows NARX-BR as an effective technique for LiB SOC prediction.
... There have been several regression-oriented approaches with various training functions used 22,23 . The Levenberg-Marquardt backpropagation 24 and Bayesian regularization backpropagation 25 algorithms are regarded as the best options for such nonlinear dynamics due to their fast computation because fast backpropagation algorithms are highly recommended as first-choice supervised algorithms. It has been shown that ANNs are effective when physical processes are obscure or complicated 26 . ...
Offshore wind energy is getting increasing attention as a clean alternative to the currently scarce fossil fuels mainly used in Europe’s electricity supply. The further development and implementation of this kind of technology will help fighting global warming, allowing a more sustainable and decarbonized power generation. In this sense, the integration of Floating Offshore Wind Turbines (FOWTs) with Oscillating Water Columns (OWCs) devices arise as a promising solution for hybrid renewable energy production. In these systems, OWC modules are employed not only for wave energy generation but also for FOWTs stabilization and cost-efficiency. Nevertheless, analyzing and understanding the aero-hydro-servo-elastic floating structure control performance composes an intricate and challenging task. Even more, given the dynamical complexity increase that involves the incorporation of OWCs within the FOWT platform. In this regard, although some time and frequency domain models have been developed, they are complex, computationally inefficient and not suitable for neither real-time nor feedback control. In this context, this work presents a novel control-oriented regressive model for hybrid FOWT-OWCs platforms. The main objective is to take advantage of the predictive and forecasting capabilities of the deep-layered artificial neural networks (ANNs), jointly with their computational simplicity, to develop a feasible control-oriented and lightweight model compared to the aforementioned complex dynamical models. In order to achieve this objective, a deep-layered ANN model has been designed and trained to match the hybrid platform’s structural performance. Then, the obtained scheme has been benchmarked against standard Multisurf-Wamit-FAST 5MW FOWT output data for different challenging scenarios in order to validate the model. The results demonstrate the adequate performance and accuracy of the proposed ANN control-oriented model, providing a great alternative for complex non-linear models traditionally used and allowing the implementation of advanced control schemes in a computationally convenient, straightforward, and easy way.
... Nevertheless, this training method has its own validation built into its algorithm, where the validation was performed in the form of regularization [22]. This result was analogous to those reported by Jazayeri et al. [28] whereby no data were available on the validation performance when they estimated the output power of a photovoltaic (PV) module. ...
In this study, the response surface methodology (RSM) and artificial neural network (ANN) were employed to study the adsorption process of 2,4-dichlorophenoxyacetic acid (2,4-D) by using modified hydrogel, i.e., activated carbon poly(dimethylaminoethyl methacrylate) (AC/PDMAEMA hydrogel). The effect of pH, the initial concentration of 2,4-D and the activated carbon content on the removal of 2,4-D and adsorption capacity were investigated through the face-centered composite design (FCCD), optimal design and two-level factorial design. The response surface plot suggested that higher removal of 2,4-D and adsorption capacity could be achieved at the higher initial concentration of 2,4-D and lower pH and activated carbon content. The modeling and optimization for the adsorption process of 2,4-D were also carried out by different design methods of RSM and different training methods of ANN. It was found that among the three design methods of RSM, the optimal design has the highest accuracy for the prediction of 2,4-D removal and adsorption capacity (R2 = 0.9958 and R2 = 0.9998, respectively). The numerical optimization of the optimal design found that the maximum removal of 2,4-D and adsorption capacity of 65.01% and 65.29 mg/g, respectively, were obtained at a pH of 3, initial concentration of 2,4-D of 94.52 mg/L and 2.5 wt% of activated carbon. Apart from the optimization of process parameters, the neural network architecture was also optimized by trial and error with different numbers of hidden neurons in the layers to obtain the best performance of the response. The optimization of the neural network was performed with different training methods. It was found that among the three training methods of the ANN model, the Bayesian Regularization method had the highest R2 and lowest mean square error (MSE) with the optimum network architecture of 3:9:2. The optimum condition obtained from RSM was also simulated with the optimized neural network architecture to validate the responses and adequacy of the RSM model.
... Wilamowski and Yu introduced the Hessian approximation, as demonstrated in (11). The LM algorithm update rule is provided as (12) [23]. Where J: Jacobian matrix (matrix of first derivatives with respect to weight vector), μ: combination coefficient, and I: identity matrix. ...
A related input parameter is used in this case study to forecast solar thermal systems (STS) capabilities and to compare which artificial neural network (ANN) algorithms and other artificial intelligence (AI) methods have the most reliable predictor for STS performance. In order to gauge the performance of the STS, this research aims to implement AI for predicting STS performance by comparing the ANN technique with other methods. Three different training algorithms which are Levenberg-Marquardt (LM), scaled conjugate gradient (SCG) and Bayesian regularization (BR) are considered in this research. This research will identify acceptable parameters and the best AI technique to use in predicting the STS performance. Previous research on STS demonstrates that the efficiency of STS has been estimated using different input parameters. The results show that the prediction of the LM training algorithm is the best for STS performance.
... BESN has an accuracy that meets the operational requirements of electricity supply feasibility of more than 90% [27], which is in line with [28], which also results in the overall performance of the Levenberg Marquardt and Bayesian Regularization neural network models in a different time and input intervals showing the best trade-off performance in estimating the power. ...
Weather forecasting has become very urgent in various fields of human life, including in big cities. The need for weather forecasting accuracy will be effective and efficient in managing the quality of civilization flexibly. Bayesian regularization is one of the techniques used to obtain accurate results and development of artificial neural networks. The training process achieves the smallest epoch using a general processing unit to solve big data and high resolution. Scenarios performed via dataset partitioning and MSE enhancement. The addition of training data will improve system performance which indicates a significant increasing accuracy. Likewise, the decrease in MSE can increase the system accuracy to achieve a convergence stability point. Weather forecasting can recommend work units within the city and its surroundings, even between provinces or countries.
... Each input is assigned a weight that represents its relative importance. Herein, the Levenberg-Marquardt backpropagation (BP) training algorithm is adopted for supervised learning [47]. It computes the gradient of the loss function and updates weights to minimize the error as learning proceeds. ...
The size of the steel section in concrete encased steel (CES) structure of bridge engineering is much larger than that of the building construction, accompanied by great shear demands at the steel–concrete interface. To enhance the interfacial shear performance, perforated web connection (PWC), which perforates the web of steel section with perforating rebars passing through the web openings, is proposed in this research for bridge application. Push-out tests and refined numerical analysis were performed on CES specimens with PWC, aiming to evaluate its effectiveness and to reveal the shear transfer mechanism. The effect of various parameters on the shear resistance was also investigated through a parametric study using the validated modeling method. Eventually, an Artificial Neural Network (ANN) was developed to predict the ultimate shear resistance of PWC. The results indicate that the interfacial shear-slip response especially the post-peak behavior is significantly improved by PWC with perforating rebars, which transforms the failure manner from brittle to ductile. The failure process of the PWC with perforating rebars experiences the loss of bonding, activation of perforating rebars, as well as shear failure of perforating rebars and concrete dowels. The parametric analysis shows that the PWC with large web openings exhibits great ultimate shear resistance and sensitivity to the variation in other parameters owing to good compatibility between the natural bond and reinforced concrete dowels. The well trained ANN model is a powerful tool for efficiently predicting the ultimate shear resistance of such connection.
... As the heat of the panels increases due to high temperature, the panels provide low energy, thereby decreasing the efficiency of the power generated by the panel, as the temperature is inversely proportional to PV power. Conversely, as the temperature of panels rises, the effectiveness of the PV power falls, and efficiency increases with the cooler panels [10]. In large grid, PV module fault detection is also very crucial [17]. ...
... The irradiance is defined as the density of the solar radiation power received on a given surface [8]. For the present work, a 32° inclined and south oriented pyranometer is used for in-situ measurements. ...
This paper deals with the predictive capabilies of a NARX-based forecaster used to predict the output power converted with a PV panel in isolated conditions. A timeseries based NARX model is proposed and the influence of the meteorological data such as irradiance, ambient temperature and wind speed, and the impact of the training algorithm on the performance of the NARX based forecaster model is studied. The results show that for the studied area, the NARX model trained by three meteorological data as inputs, the output power as output using Bayesian Regularization algorithm gives best performance with a mean squared error of 2.10414e-2.
... The performance of an algorithm used for training is to a large extent dependent on how well such an algorithm can learn and train the pattern of the input and the targeted parameters [41]. Although Bayesian regularization offers an advantage of a higher generalization ability but often requires a longer time to converge compared to the Levenberg-Marquardt trained ANN [42]. In a comparative analysis by Kayri et al. [29]. ...
Thermo-catalytic methane decomposition is a prospective route for producing COx free hydrogen. In this study, Bayesian regularization and Levenberg-Marquardt trained multilayer perceptron neural networks were employed in predictive modeling of hydrogen production by thermo-catalytic methane decomposition. Based on the non-linear relationship between the reaction temperature, weight of the catalysts, time of stream, calcination temperature, calcination time, specific volume, and the hydrogen yield, the various topology was configured for the neural network and tested to determine the artificial neuron that would result in the best model performance. The Levenberg-Marquardt trained neural network displayed the best performance with the model topology of 7–16-1 compared with the Bayesian regularization trained network. The model topology of 7–16-1 represents the input units, hidden neuron, and the output unit. The predicted hydrogen yield by the 7–16-1 configured neural network was in strong agreement with the observed value, evidenced by the coefficient of determination (R²) of 0.953 and mean square error of 0.03. A predicted hydrogen yield of 86.56 vol.% was obtained at the reaction temperature of 700 °C, 0.5 g catalyst weight, calcination temperature of 600 °C, calcination time of 240 min, catalyst specific surface area of 24.1 m²/g, the pore volume of 0.03 cm³/g, and 160 min time on stream which is at proximity with the observed value of 84 vol.%. The sensitivity analysis revealed that all the input parameters have varying levels of importance on the model output. However, the intrinsic properties of the catalysts (specific surface area, and the pore volume) have the most significant influence on the predicted hydrogen yield.
... The irradiance is defined as the density of the solar radiation power received on a given surface [29]. In the present work, the global irradiance incident on a 32° inclined and southoriented panel is considered. ...
Photovoltaic energy is volatile in nature since it depends on weather conditions. It is important to have an idea about the reliability and the economic feasibility of any new project to decide whether it is right to proceed with the installation of such a project. Hence, it is becoming fundamental to know renewable energy state and production that can be combined with other less variable and more predictable sources to justify the choice of regions for the new photovoltaic projects installation. The current research investigates the forecasting abilities of a NARX based approach. The influence of the meteorological data, such as irradiance, ambient temperature, and wind speed, and the impact of training algorithms on the performance of the NARX-based forecaster is studied. For this purpose, four models are discussed, each model is trained based on three training algorithms. The NARX model using a Bayesian Regularization algorithm, trained by the three meteorological data as inputs and the converted power output as output, outperforms the other models. It consists of a simple architecture with one input layer, a hidden layer containing 1O neurons, and an output layer, with a mean square error of 0.0085 W2 for the training phase and 0.0043 W2 testing phase, and the overall regression of 95.48%. This simplified architecture and low values of the mean square error and the regression coefficient suggest that they are promising photovoltaic output prediction tools, particularly in locations where few meteorological parameters are monitored.
... This framework has been implemented in the function "trainbr" of MATLAB software [62,63]. This algorithm eliminates network weights with small effects on the solution and shows super performance by avoiding local minimums [64]. ...
This study aims to investigate the performance of three advanced data-driven models, namely multivariate adaptive regression spline (MARS), artificial neural network (ANN), and adaptive neural fuzzy inference system (ANFIS), for predicting the axial compression capacity of circular concrete-filled double-skin steel tube (CFDST) columns. For this purpose, 125 experimental data sets collected from the literature were used to develop the MARS, ANN, and ANFIS models. In this regard, the column length, the outer diameter, the outer thickness and yield strength of the outer steel tube, the inner diameter, the inner thickness and yield strength of the inner steel tube, and the compressive strength of concrete were considered as input variables, meanwhile, axial compression capacity was considered as the output variable. The performance of the three data-driven models was compared with six equations proposed by design codes and other authors. The comparisons showed that three data-driven models achieved more accuracy than previous equations, of which, the ANN model has an advantage over the ANFIS and MARS models. Finally, a graphical user-friendly interface (GUI) was developed to make the MARS, ANN, and ANFIS models become more attractive for practical use.
... The developed ANN model was used to simulate new data in order to conduct a sensitivity analysis. In other words, each input of ANN model varied between its lowest (L), mean (M), and highest (H) values [72] while other input parameters were kept constant at their reference value (a value close to their mean value). Table 6 represents the reference values having to do with each input parameter. ...
Alkali-Silica-Reaction (ASR) is one of the most deteriorating phenomena in concrete structures. This study uses a machine learning approach (i.e. Artificial Neural Network) to provide further insight into ASR. The approach combines chemo-mechanical and kinetics-based approaches to develop a time- and temperature-dependent model of ASR, which is eventually used in generating user-friendly charts to conveniently assess existing concrete structures. To reach a higher degree of confidence in the precision of the model, an experimental dataset was developed from the laboratory and was combined with a dataset from the literature. A comparison between the developed model and a chemo-mechanical one (Gao's model) showed higher accuracy for the developed model. This higher accuracy was more obvious regarding the specimen with fine single-size aggregate grading. This study also reveals a varying thickness of connected porosity (tc) for fine single-size aggregate. Based on the results, aggregate size and tc have a coupled effect on the ASR-induced expansion.
... In general, the backpropagation principle is the most widely used algorithm in the field of neural networks. It is used in about 80-90% of applications [49]. Backpropagation optimization is used for forward neural networks with hidden layers and is designed for data classification that is generally not linearly separable [50]. ...
This article describes the design and verification of the indirect method of predicting the course of CO2 concentration (ppm) from the measured temperature variables Tindoor (°C) and the relative humidity rHindoor (%) and the temperature Toutdoor (°C) using the Artificial Neural Network (ANN) with the Bayesian Regulation Method (BRM) for monitoring the presence of people in the individual premises in the Intelligent Administrative Building (IAB) using the PI System SW Tool (PI-Plant Information enterprise information system). The CA (Correlation Analysis), the MSE (Root Mean Squared Error) and the DTW (Dynamic Time Warping) criteria were used to verify and classify the results obtained. Within the proposed method, the LMS adaptive filter algorithm was used to remove the noise of the resulting predicted course. In order to verify the method, two long-term experiments were performed, specifically from February 1 to February 28, 2015, from June 1 to June 28, 2015 and from February 8 to February 14, 2015. For the best results of the trained ANN BRM within the prediction of CO2, the correlation coefficient R for the proposed method was up to 92%. The verification of the proposed method confirmed the possibility to use the presence of people of the monitored IAB premises for monitoring. The designed indirect method of CO2 prediction has potential for reducing the investment and operating costs of the IAB in relation to the reduction of the number of implemented sensors in the IAB within the process of management of operational and technical functions in the IAB. The article also describes the design and implementation of the FEIVISUAL visualization application for mobile devices, which monitors the technological processes in the IAB. This application is optimized for Android devices and is platform independent. The application requires implementation of an application server that communicates with the data server and the application developed. The data of the application developed is obtained from the data storage of the PI System via a PI Web REST API (Application Programming Integration) client.
... As for the number of hidden layer nodes, 70% of the input values were allocated for training samples, and 30% for the testing. Moreover, Bayesian Regularization Back Propagation (BRBP) [35,36] was employed to train the network. ...
Due to the limited bandwidth and continuous growth of users make the IT managers focus on the strategies concerning bandwidth allocation. Thus, managing the bandwidth in campus networks has become one of the challenges in recent years. This paper introduces a mechanism for bandwidth allocation based on the users' web usage patterns. The main purpose is to set a higher bandwidth to the users who are inclined to browsing educational websites compared to those who are not. In attaining this proposed technique, a hybrid data mining approach were done which composed of preprocessing of the weblogs, class labeling of the dataset, computation of the feature subspaces, training for the development of the ANN for LDA/GSVD algorithm, and visualization. The proposed method was applied to real weblogs from university's proxy servers which include 104 active users. The results indicate that the proposed method is useful in classifying those users who used the internet in an educational way and those who are not. Thus, the developed ANN for LDA/GSVD algorithm outperformed the existing algorithm up to 50% which indicates that this approach is efficient. Based on the results, 35.06% of the users browsed educational contents. Through this technique, users will be encouraged by using the internet for educational purposes. Moreover, IT managers can make better decisions to optimize the allocation of bandwidth resources.
... The "trainbr" uses Bayesian Regularisation Backpropagation (BRB) training algorithm, providing strong estimation for noisy and difficult inputs. BRB algorithm effectually eliminates network weights with small effects on the problem's solution and shows superb performance by avoiding local minimums [40]. ...
The main objective of this study is to introduce a novel numerical approach, based on Artificial Neural Network (ANN), to predict the shear strength of Perfobond rib shear connector (PRSC). For this purpose, 90 records were extracted from the literature and were used to develop a number of Bayesian neural network models for predicting the shear strength of PRSC. An accurate ANN model was attained with a high value of correlation coefficient for the train and test subsets. Having a reliable ANN, a parametric study on the shear strength of PRSC was carried out to establish the trend of main contributing factors. The majority of assumptions, considered by empirical equations, were predicted by the developed ANN. Moreover, a sensitivity analysis of input variables was conducted; the outcomes revealed that the area of concrete dowels had the strongest influence on the shear strength of PRSC. Eventually, using the validated ANN, an abundant number of curves (Master Curves) were generated to introduce a user-friendly equation. According to the results, both the ANN model and the proposed equation reflect a higher accuracy than other existing empirical equations.
The adaptive neuro-fuzzy inference system (ANFIS), central composite experimental design (CCD)-response surface methodology (RSM), and artificial neural network (ANN) are used to model the oxidation of benzyl alcohol using the tert-butyl hydroperoxide (TBHP) oxidant to selectively yield benzaldehyde over a mesoporous ceria-zirconia catalyst. Characterization reveals that the produced catalyst has hysteresis loops, a sponge-like structure, and structurally induced reactivity. Three independent variables were taken into consideration while analyzing the ANN, RSM, and ANFIS models: the amount of catalyst (A), reaction temperature (B), and reaction time (C). With the application of optimum conditions, along with a constant (45 mmol) TBHP oxidant amount, (30 mmol) benzyl alcohol amount, and rigorous refluxing of 450 rpm, a maximum optimal benzaldehyde yield of 98.4% was obtained. To examine the acceptability of the models, further sensitivity studies including statistical error functions, analysis of variance (ANOVA) results, and the lack-of-fit test, among others, were employed. The obtained results show that the ANFIS model is the most suited to predicting benzaldehyde yield, followed by RSM. Green chemistry matrix calculations for the reaction reveal lower values of the E-factor (1.57), mass intensity (MI, 2.57), and mass productivity (MP, 38%), which are highly desirable for green and sustainable reactions. Therefore, utilizing a ceria-zirconia catalyst synthesized via the inverse micelle method for the oxidation of benzyl alcohol provides a green and sustainable methodology for the synthesis of benzaldehyde under mild conditions.
The quest for energy and environmental sustainability necessitates an increasing interest in the photocatalytic conversion of wastewater to biohydrogen. However, the complexity of the photocatalytic conversion and the low productivity of the biohydrogen produced has become a major concern in the scale-up of the process. This study employs a data-driven approach to model biohydrogen production from the photocatalytic conversion of wastewater. Having ascertained the influence of five different parameters namely catalyst size, reaction temperature, catalyst among, irradiation time, and radiation intensity on the biohydrogen production through parametric analysis, the data were employed to model the process using multilayer perceptron neural network (MLPNN) and nonlinear autoregressive neural network (NARX). Both the MLPNN and NARX models were trained using Levenberg-Marquardt (LM), Bayesian regularization (BR), and scaled conjugate gradient (SCG) algorithms. The performance of 20 network architectures was tested for MLPNN-LM, MLPNN-BR, MLPNN-SCG, NARX-LM, NARX-BR, and NARX-SCG. The analysis revealed that the best network architectures of 5-14-1, 5-11-1, 5-7-1, 5-14-1, 5-15-1, and 5-7-1 were obtained for the MLPNN-LM, MLPNN-BR, MLPNN-SCG, NARX-LM, NARX-BR, and NARX-SCG, respectively. All the models demonstrated a good predictability of the biohydrogen production as evidenced by the coefficient of determination (R²) > 0.9 and low root mean square error (RMSE) values. The best performance was displayed by MLPNN-BR model with R² of 0.999 and RMSE of 0.138. The independent variable analysis shows that all the factors significantly influence the predicted biohydrogen production. The catalyst size has the most significant effect on the predicted hydrogen production as indicated by the importance value of 0.329.
Renewable energy induced by wind and wave sources is playing an indispensable role in electricity production. The innovative hybrid renewable offshore platform concept, which combines Floating Offshore Wind Turbines (FOWTs) with Oscillating Water Columns (OWCs), has proven to be a promising solution to harvest clean energy. The hybrid platform can increase the total energy absorption, reduce the unwanted dynamic response of the platform, mitigate the load in critical situations, and improve the system's cost efficiency. However, the nonlinear dynamical behavior of the hybrid offshore wind system presents an opportunity for stabilization via challenging control applications. Wind and wave loads lead to stress on the FOWT tower structure, increasing the risk of damage and failure, and raising maintenance costs while lowering its performance and lifespan. Moreover, the dynamics of the tower and the platform are extremely sensitive to wind speed and wave elevation, which causes substantial destabilization in extreme conditions, particularly to the tower top displacement and the platform pitch angle. Therefore, this article focuses on two main novel targets: (i) regressive modeling of the hybrid aero-hydro-servo-elastic-mooring coupled numerical system and (ii) an ad-hoc fuzzy-based control implementation for the stabilization of the platform. In order to analyze the performance of the hybrid FOWT-OWCs, this article first employs computational Machine Learning (ML) techniques, i.e., Artificial Neural Networks (ANNs), to match the behavior of the detailed FOWT-OWCs numerical model. Then, a Fuzzy Logic Control (FLC) is developed and applied to establish a structural controller mitigating the undesired structural vibrations. Both modeling and control schemes are successfully implemented, showing a superior performance compared to the FOWT system without OWCs. Experimental results demonstrate that the proposed ANN-based modeling is a promising alternative to other intricate nonlinear NREL 5 MW FOWT dynamical models. Meanwhile, the proposed FLC improves the platform's dynamic behavior, increasing its stability under a wide range of wind and wave conditions.
Offshore wind energy is getting increasing attention as a clean alternative to the currently scarce fossil fuels mainly used in Europe’s electricity supply. The further development and implementation of this kind of technology will help fighting global warming, allowing a more sustainable and decarbonized power generation. In this sense, the integration of Floating Offshore Wind Turbines (FOWT) with Oscillating Water Column (OWCs) devices arise as a promising solution for hybrid renewable energy production. In these systems, OWC modules are employed not only for wave energy generation but also for FOWT stabilization, cost-efficiency and prognosis. Nevertheless, analyzing and understanding the aero-hydro-servo-elastic floating structure control performance composes an intricate and challenging task. Even more given the dynamical complexity increase that involves the incorporation of OWCs within the FOWT platform. In this regard, although some time and frequency domain models have been developed, they are complex, computationally inefficient and not suitable for neither real-time nor feedback control. In this context, this work presents a novel control-oriented regressive model for hybrid OWC-FOWT platforms. The main objective is to take advantage of the predictive and forecasting capabilities of the deep-layered artificial neural networks (ANN), jointly with their computational simplicity, to develop a feasible control-oriented and lightweight model compared to the aforementioned complex dynamical models. In order to achieve this objective, a deep-layered ANN model has been designed and trained to match the hybrid platform’s structural performance. Then, the obtained scheme has been benchmarked against standard Multisurf-Wamit-FAST 5MW FOWT output data for different challenging scenarios in order to validate the model. The results demonstrate the adequate performance and accuracy of the proposed ANN control-oriented model, composing a great alternative for complex non-linear models traditionally used and allowing the implementation of advanced control schemes in a computationally convenient, straightforward and easy way.
In this study, thermal/hydraulic performance (THP) of a heated bottom surface channel, on which hollow straight, elliptic, and wavy top surface trapezoidal baffles are attached, have been investigated. Artificial neural network modeling (ANN) and 3D ANSYS numerical valuations are presented and validated using experimental data obtained by Sahin et al. (2019). Six parameters are studied: baffle height (H), baffle width (S), baffle length (L), corner angle (α), inclination angle (β), and Reynolds number (Re). This study is carried for both aligned and staggered baffles configuration. Air is considered to be the working fluid at a range of Reynolds number (5 × 10³ ≤ Re ≤ 2.3 × 10⁴). Backpropagation technique used to determine the optimal ANN size with the Bayesian regularization algorithm. Results show that the most effective parameters on THP are Re, α, and H. Based on the thermal hydraulic efficiency values, the optimum condition is obtained at α= 16°, β = 0°, and baffle dimensions of H = 20 mm, L = 25 mm, S = 26 mm. The average value of thermal enhancement factor (TEF) shows that the staggered arrangement is superior to aligned one for all baffle shapes at the optimum condition. The largest TEF is 1.2 which is captured for the case of the wavy top surface trapezoidal baffles in staggered arrangement at Re = 5000.
In the middle-stream petroleum industry, the concurrent flow of oil and water in pipes is a typical occurrence. Several factors have been reported to influence the flowability of two-phase water-oil in pipeline. However, the non-linear relationship between these factors and the flowability of the two-phase water-oil is not yet known. The understanding of these relationships could help in developing models that might be employed in predicting flow behavior. In this study, two-phase water-oil flowability was experimentally investigated and modeled using supervised machine learning algorithms. The performance of each of the models was optimized by evaluating the model with the best-hidden neuron. For the Levenberg-Marquardt backpropagation, Bayesian regularization backpropagation, and scaled conjugate gradient backpropagation networks, the best topologies of 4-8-4, 4-9-4, and 4-10-4 were obtained, and coefficient of determination (R²) of 0.998, 0.999, 0.996 respectively. The predicted kinematic viscosity, dynamic viscosity, pressure drop, and power consumption were consistent with the observed values. The sensitivity analysis revealed that the temperature has the most significant effect on the predicted output. The robustness of using the supervised machine learning technique in modeling the flowability of two-phase water-crude oil mixes in pipelines based on the relationship between the predictors and the targeted variables was demonstrated.
Experimental results of free convection thermal performance on the elliptical-tube outer surface at various inclination angles (θ) and heat fluxes (q) with artificial neural network modelling (ANN) and 3D ANSYS numerical evaluations are presented. The considered parameters include: axial distance (0.1 ≤ X/L ≤ 0.9), (0° ≤ θ ≤ 90°), and Rayleigh number (2.6 × 10⁶ ≤ Ra ≤ 2.05 × 10⁸). Bayesian regularization algorithm as a method of back-propagation technique is used for selecting the optimal ANN size. Results revealed that Nu increases with the increase in θ at constant Ra. Furthermore, θ = 90° gives higher values of Nu compared to all other elliptical-tube inclination angles at the considered Ra range. The elliptical-tube shape with major to minor diameter of (≈1.8) and at θ = 90° can be approximated as a vertical flat plate case. So, the elliptical-tube with θ = 90° is equivalent to two vertical-plates. The recapitulation of the ANN results indicates that, the highest Nu values exist at the tube two peripheral ends with (405 W/m² ≤ q ≤ 885 W/m²) and Nu achieves the highest values at 555 W/m². The current experimental data is correlated and compared with the ANN model output and other reported work in the literature.
This paperexplores theapplicationof artificial neural networksfor online identification of a multimachinepower system.Arecurrent neural networkhas been proposedas the identifier of the two area, four machinesystemwhich is a benchmark system for studying electromechanical oscillations in multimachine powersystems. This neural identifier is trained using the static Backpropagation algorithm. The emphasis of thepaper is on investigating the performance of the variants of the Backpropagation algorithm in training theneural identifier. The paper also compares the performances of the neural identifiers trained usingvariantsof the Backpropagation algorithmover a wide range of operating conditions.The simulation resultsestablish a satisfactory performance of the trained neural identifiers in identification of the test powersystem
This paper presents a solar power modelling method using artificial neural networks (ANNs). Two neural network structures, namely, general regression neural network (GRNN) feedforward back propagation (FFBP), have been used to model a photovoltaic panel output power and approximate the generated power. Both neural networks have four inputs and one output. The inputs are maximum temperature, minimum temperature, mean temperature, and irradiance; the output is the power. The data used in this paper started from January 1, 2006, until December 31, 2010. The five years of data were split into two parts: 2006-2008 and 2009-2010; the first part was used for training and the second part was used for testing the neural networks. A mathematical equation is used to estimate the generated power. At the end, both of these networks have shown good modelling performance; however, FFBP has shown a better performance comparing with GRNN.
Two popular techniques for generalizing artificial neural network (ANN) models are the Bayesian regularization (BR) and the cross-validated early stopping (CVES) methods. The comparison between the two methods and the no-stop training method (NST, in which the training is stopped after 1000 epochs) for forecasting one-step ahead discharges based on univariate daily and monthly streamflow time series of three rivers (the Yellow River in China, and the Rhine and the Danube rivers in Europe), shows that both methods perform generally better than NST when the ratio (R) of training sample size to the number of weights in the network is less than 20. But the advantage of the two techniques over NST is not guaranteed. In several cases they fail even when the ratio R < 10, especially the CVES method. On the other hand, comparison between the BR and CVES method shows that BR outperforms CVES in most cases, and it even outperforms the NST method in several cases when the ratio R is larger than 30. Furthermore, the performance of BR is very stable. whereas the performance of CVES is highly unstable. The evaluation result for making 1- to 5-day ahead streamflow forecasts for the upper Yellow River based on multiple explanatory variables further confirms the advantage of the BR method.
The paper illustrates an adaptive approach based on different topologies of artificial neural networks (ANNs) for the power energy output forecasting of photovoltaic (PV) modules. The analysis of the PV module's power output needed detailed local climate data, which was collected by a dedicated weather monitoring system. The Department of Energy, Information Engineering, and Mathematical Models of the University of Palermo (Italy) has built up a weather monitoring system that worked together with a data acquisition system. The power output forecast is obtained using three different types of ANNs: a one hidden layer Multilayer perceptron (MLP), a recursive neural network (RNN), and a gamma memory (GM) trained with the back propagation. In order to investigate the influence of climate variability on the electricity production, the ANNs were trained using weather data (air temperature, solar irradiance, and wind speed) along with historical power output data available for the two test modules. The model validation was performed by comparing model predictions with power output data that were not used for the network's training. The results obtained bear out the suitability of the adopted methodology for the short-term power output forecasting problem and identified the best topology.
The present study describes a neural network approach for modeling and making short-term predictions on the total solar radiation
time series.
The future hourly values of total solar radiation for several years are predicted, by extracting knowledge from their past
values, using feedforward backpropagation neural networks. The results are tested using various sets of non training measurements,
the findings are very encouraging and the model is found able to simulate the future values of total solar radiation time
series based on their past values. “Multi-lag” output predictions are performed using the predicted values to the input database
in order to model future total solar radiation values with sufficient accuracy. Furthermore, an autoregressive model is developed
for analysing and representing the total solar radiation time series. The predicted values of solar radiation are compared
with the observed data series and it was found that the neural network approach leads to better predictions than the AR model.
A quantitative and practical Bayesian framework is described for learning of mappings in feedforward networks. The framework makes possible (1) objective comparisons between solutions using alternative network architectures, (2) objective stopping rules for network pruning or growing procedures, (3) objective choice of magnitude and type of weight decay terms or additive regularizers (for penalizing large weights, etc.), (4) a measure of the effective number of well-determined parameters in a model, (5) quantified estimates of the error bars on network parameters and on network output, and (6) objective comparisons with alternative learning and interpolation models such as splines and radial basis functions. The Bayesian "evidence" automatically embodies "Occam's razor," penalizing overflexible and overcomplex models. The Bayesian approach helps detect poor underlying assumptions in learning models. For learning models well matched to a problem, a good correlation between generalization ability and the Bayesian evidence is obtained.
CD-ROM with databases and software included CD-Rom avec bases de données et logiciels inclus
This book is edited by Presses des Mines. See at https://www.pressesdesmines.com/produit/the-european-solar-radiation-atlas-vol-1/
The present study utilizes the radial basis functions technique for the estimation of monthly mean daily values of solar radiation falling on horizontal surfaces and compares its performance with that of the multilayer perceptrons network and a classical regression model. In this work, we use solar radiation data from 41 stations that are spread over the Kingdom of Saudi Arabia. The solar radiation data from 31 locations are used for training the neural networks and the data from the remaining 10 locations are used for testing the estimated values. However, the testing data were not used in the modeling or training of the networks to give an indication of the performance of the system at unknown locations. Results indicate the viability of the radial basis for this kind of problem.
Wireless sensor networks (WSNs) have many applications in the field of disaster management, military, healthcare and environmental monitoring. Capability of WSNs is further enhanced by the efficient localization algorithms. Localization is the process by which a sensor node determines its own location after deployment. Neural approaches are gaining popularity in evolving new localization algorithms that are capable of optimizing various parameters of WSNs. In this paper, we analyse two backpropagation algorithms based on multi-layer Perceptron (MLP) neural network. The network is trained using static sensor nodes placed in a grid with their coordinates known. The input values are distances from each anchor nodes to a particular sensor node. The output is the actual coordinates of the sensor nodes. After training, the network will be able to predict the coordinates of unknown sensor nodes. This MLP model is analyzed for Bayesian regularization and Levenberg-Marquardt training algorithm. Both algorithms are tested for the robustness and cross-validation. The simulation results demonstrate the effectiveness of the proposed model on localization error.
In this work, a novel approach using an artificial neural network was used to develop a model for analyzing the relationship between the Global Radiation (GR) and climatological variables, and to predict GR for locations not covered by the model's training data. The predicted global radiation values for the different locations (for different months) were then compared with the actual values. Results indicate that the model predicted the Global Radiation values with a good accuracy of approximately 93% and a mean absolute percentage error of 7.30. In addition, the model was also tested to predict GR values for the Seeb location over a 12 month period. The monthly predicted values of the ANN model compared to the actual GR values for Seeb produced an accuracy of 95% and a mean absolute percentage error of 5.43. Data for these locations were not included as part of ANN training data. Hence, these results demonstrate the generalization capability of this novel approach over unseen data and its ability to produce accurate estimates. Finally, this ANN-based model was also used to predict the global radiation values for Majees, a new location in north Oman.
Many computer simulation models which predict growth, development, and yield of agronomic and horticultural crops require daily weather data as input. One of these inputs is daily total solar radiation, which in many cases is not available owing to the high cost and complexity of the instrumentation needed to record it. The aim of this study was to develop a neural network model which can predict solar radiation as a function of readily available weather data and other environmental variables. Four sites in the southeastern USA, i.e. Tifton, GA, Clayton, NC, Gainesville, FL, and Quincy, FL, were selected because of the existence of longterm daily weather data sets which included solar radiation. A combined total of 23 complete years of weather data sets were available, and these data sets were separated into 11 years for the training data set and 12 years for the testing data set. Daily observed values of minimum and maximum air temperature and precipitation, together with daily calculated values for daylength and clear sky radiation, were used as inputs for the neural network model. Daylength and clear sky radiation were calculated as a function of latitude, day of year, solar angle, and solar constant. An optimum momentum, learning rate, and number of hidden nodes were determined for further use in the development of the neural network model. After model development, the neural network model was tested against the independent data set. Root mean square error varied from 2.92 to 3.64 MJ m−2 and the coefficient of determination varied from 0.52 to 0.74 for the individual years used to test the accuracy of the model. Although this neural network model was developed and tested for a limited number of sites, the results suggest that it can be used to estimate daily solar radiation when measurements of only daily maximum and minimum air temperature and precipitation are available.
Most algorithms for the least-squares estimation of non-linear parameters have centered about either of two approaches. On the one hand, the model may be expanded as a Taylor series and corrections to the several parameters calculated at each iteration on the assumption of local linearity. On the other hand, various modifications of the method of steepest-descent have been used. Both methods not infrequently run aground, the Taylor series method because of divergence of the successive iterates, the steepest-descent (or gradient) methods because of agonizingly slow convergence after the first few iterations. In this paper a maximum neighborhood method is developed which, in effect, performs an optimum interpolation between the Taylor series method and the gradient method, the interpolation being based upon the maximum neighborhood in which the truncated Taylor series gives an adequate representation of the nonlinear model. The results are extended to the problem of solving a set of nonlinear algebraic e
The improved computation presented in this paper is aimed to optimize the neural networks learning process using Levenberg-Marquardt (LM) algorithm. Quasi-Hessian matrix and gradient vector are computed directly, without Jacobian matrix multiplication and storage. The memory limitation problem for LM training is solved. Considering the symmetry of quasi-Hessian matrix, only elements in its upper/lower triangular array need to be calculated. Therefore, training speed is improved significantly, not only because of the smaller array stored in memory, but also the reduced operations in quasi-Hessian matrix calculation. The improved memory and time efficiencies are especially true for large sized patterns training.
Modeling of global solar radiation data from sunshine duration and temperature using the radial basis function networks
- A Mellit
- M Benghanem
- A Hadj Arab
- A Guessoum