No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
A computational intelligence approach for solar photovoltaic power generation forecasting
Abstract
This article describes an approach applying computational intelligence methods for the problem of forecasting solar photovoltaic power generation at country level. Precise forecast of power generation plays a vital role in designing a dependable photovoltaic power generation system. The computed predictions enable the implementation of efficient planning, management, and distribution strategies for the generated power, ultimately enhancing the performance and efficiency of the system. The study analyzes and compares artificial neural network approaches for a specific case study using real solar photovoltaic power generation data from Uruguay in the period 2018 to 2022. Several artificial neural network architectures are evaluated for forecasting. The main results indicate that the approach applying a combination of Encoder-Decoder and Long Short Term Memory artificial neural networks is the most effective method for the addressed forecasting problem. The approach yielded promising results, with an average mean error value of 0.09, improving over the other artificial neural network architectures. Even better results were obtained for sunny days. The generated forecasts hold significant value for its application in planning and scheduling processes, aiming to enhance the overall quality of service of the electricity grid.
This study analyzes the technical and economic feasibility of hybrid photovoltaic/thermal (PVT) solar energy systems, comparing them with independent flat plate solar thermal collectors (FP) and photovoltaic (PV) modules. Using TRNSYS software, different configurations were simulated under the climatic conditions of Mexicali, Baja California, Mexico. The results indicate that, although independent FP and PV systems exhibit better energy performance in terms of energy production and solar fraction, PVT collectors offer higher specific global energy production (electrical + thermal). However, the economic feasibility of PVT systems is limited due to high investment and maintenance costs, coupled with the low cost of thermal and electrical energy in the study region. Independent FP and PV collectors also face similar economic challenges, with return on investment extending beyond the 20-year lifespan of the project. Additionally, the low thermal energy demand for domestic hot water in the studied region negatively impacts the efficient utilization of these systems. This analysis suggests that in contexts where available installation space is limited, PVT systems can be an efficient solution. However, in scenarios with sufficient space and high energy costs, energy generation with separate equipment remains more viable. Moreover, climatic conditions and DHW demand affect the efficiency of the systems. The discrepancy between solar energy generation and hourly DHW demand necessitates the implementation of TEST, which significantly increases costs and impacts the economic viability of the project.
In this paper, we present a comprehensive and innovative framework for optimizing planning in power distribution systems. Firstly, we introduce various types of planning involved in power distribution systems, and subsequently, we provide a detailed presentation of each planning type. The different types of Power Distribution System Planning (PDSP) can be classified as follows: Asset plan, Network plan, Process plan, Energy plan, Data plan, Facility plan, Economic plan, Regulation plan, and Customer relation plan. These classifications effectively demonstrate the influential factors, conditions, and objectives of PDSP. Furthermore, this paper investigates the management strategies for different PDSP classifications. Within each classification, we list and categorize relevant studies. By utilizing the framework proposed in this paper, readers can gain a comprehensive understanding of PDSP planning measures and make more optimal decisions within power distribution companies. Additionally, this paper identifies research gaps and presents a roadmap for future PDSP studies. Based on the findings of this paper, it is evident that the majority of research in this field has focused on expansion, while process-related studies have received relatively less attention. These findings should be taken into consideration when formulating the future research roadmap.
Global navigation satellite system (GNSS) signals are significantly affected by the ionosphere. An efficient way to assess the ionospheric effects on GNSS signals is by retrieving the vertical total electron content (VTEC). Here, we propose convolutional recurrent neural network architectures to forecast VTEC based on global ionosphere maps (GIMs) of the days before the prediction period. We proposed modifications to the encoder–decoder convolutional long short-term memory (ED-ConvLSTM) architecture by innovatively using GIMs from several days and exploring the previous day’s GIM. Three new architectures were tested: The first one uses a residual connection to force the network to learn the difference between the previous and the next-day GIM; the second one uses the memory from the encoder to improve the transformation from the previous to the next-day GIM; and a third one uses 3 × 3 kernels on the encoder and 1 × 1 kernels on the decoder. Two experiments were performed using the international GNSS service (IGS) final GIM product from the years 2014–2015 (high solar activity) and 2019–2020 (low solar activity). The proposed architectures obtained better results than the original version of ED-ConvLSTM and other baseline models. We evaluated the influence of the number of GIMs (from one to four days) in the next-day GIM predictions. The results suggest that providing more than one day of GIM to the proposed networks can lead to better prediction metrics.
A modern power system is a complex network of interconnected components, such as generators, transmission lines, and distribution subsystems, that are designed to provide electricity to consumers in an efficient and reliable manner [...]
Smart grids are able to forecast customers’ consumption patterns, i.e., their energy demand, and consequently electricity can be transmitted after taking into account the expected demand. To face today’s demand forecasting challenges, where the data generated by smart grids is huge, modern data-driven techniques need to be used. In this scenario, Deep Learning models are a good alternative to learn patterns from customer data and then forecast demand for different forecasting horizons. Among the commonly used Artificial Neural Networks, Long Short-Term Memory networks—based on Recurrent Neural Networks—are playing a prominent role. This paper provides an insight into the importance of the demand forecasting issue, and other related factors, in the context of smart grids, and collects some experiences of the use of Deep Learning techniques, for demand forecasting purposes. To have an efficient power system, a balance between supply and demand is necessary. Therefore, industry stakeholders and researchers should make a special effort in load forecasting, especially in the short term, which is critical for demand response.
The recent global warming effect has brought into focus different solutions for combating climate change. The generation of climate-friendly renewable energy alternatives has been vastly improved and commercialized for power generation. As a result of this industrial revolution, solar photovoltaic (PV) systems have drawn much attention as a power generation source for varying applications, including the main utility-grid power supply. There has been tremendous growth in both on- and off-grid solar PV installations in the last few years. This trend is expected to continue over the next few years as government legislation and awareness campaigns increase to encourage a shift toward using renewable energy alternatives. Despite the numerous advantages of solar PV power generation, the highly variable nature of the sun’s irradiance in different seasons of various geopolitical areas/regions can significantly affect the expected energy yield. This variation directly impacts the profitability or economic viability of the system, and cannot be neglected. To overcome this challenge, various procedures have been applied to forecast the generated solar PV energy. This study provides a comprehensive and systematic review of recent advances in solar PV power forecasting techniques with a focus on data-driven procedures. It critically analyzes recent studies on solar PV power forecasting to highlight the strengths and weaknesses of the techniques or models implemented. The clarity provided will form a basis for higher accuracy in future models and applications.
Flood prediction in ungauged catchments is usually conducted by hydrological models that are parameterized based on nearby and similar gauged catchments. As an alternative to this process-based modelling, deep learning (DL) models have demonstrated their ability for prediction in ungauged catchments (PUB) with high efficiency. Catchment characteristics, the number of gauged catchments, and their level of hydroclimatic heterogeneity in the training dataset used for model regionalization can directly affect the model's performance. Here, we study the generalization ability of a DL model to these factors by applying an Encoder-Decoder Long Short-Term Memory neural network for a 6-hour lead-time runoff prediction in 35 mountainous catchments in China. By varying the available number of catchments and model settings with different training datasets, namely local, regional, and PUB models, we evaluated the generalization ability of our model. We found that both quantity (i.e. number of gauged catchments available) and heterogeneity of the training dataset used for the DL model are important for improving model performance in the PUB context, due to a data synergy effect. The assessment of the sensitivity to catchment characteristics showed that the model performance is mainly correlated to the local hydro-climatic conditions; the more arid the region, the more likely it is to have a poor model performance for prediction in ungauged catchments. The results suggest that the regional ED-LSTM model is a promising method to predict streamflow from rainfall inputs in PUB, and outline the need for preparing a representative training dataset.
Energy forecasting plays a vital role in mitigating challenges in data rich smart grid (SG) systems involving various applications such as demand‐side management, load shedding, and optimum dispatch. Managing efficient forecasting while ensuring the least possible prediction error is one of the main challenges posed in the grid today, considering the uncertainty in SG data. This paper presents a comprehensive and application‐oriented review of state‐of‐the‐art forecasting methods for SG systems along with recent developments in probabilistic deep learning (PDL). Traditional point forecasting methods including statistical, machine learning (ML), and deep learning (DL) are extensively investigated in terms of their applicability to energy forecasting. In addition, the significance of hybrid and data pre‐processing techniques to support forecasting performance is also studied. A comparative case study using the Victorian electricity consumption in Australia and American electric power (AEP) datasets is conducted to analyze the performance of deterministic and probabilistic forecasting methods. The analysis demonstrates higher efficacy of DL methods with appropriate hyper‐parameter tuning when sample sizes are larger and involve nonlinear patterns. Furthermore, PDL methods are found to achieve at least 60% lower prediction errors in comparison to other benchmark DL methods. However, the execution time increases significantly for PDL methods due to large sample space and a tradeoff between computational performance and forecasting accuracy needs to be maintained.
CNN originates from image processing and is not commonly known as a forecasting technique in time-series analysis which depends on the quality of input data. One of the methods to improve the quality is by smoothing the data. This study introduces a novel hybrid exponential smoothing using CNN called Smoothed-CNN (S-CNN). The method of combining tactics outperforms the majority of individual solutions in forecasting. The S-CNN was compared with the original CNN method and other forecasting methods such as Multilayer Perceptron (MLP) and Long Short-Term Memory (LSTM). The dataset is a year time-series of daily website visitors. Since there are no special rules for using the number of hidden layers, the Lucas number was used. The results show that S-CNN is better than MLP and LSTM, with the best MSE of 0.012147693 using 76 hidden layers at 80%:20% data composition.
This article introduces a dataset containing electricity consumption records of residential households in Uruguay (mostly in Montevideo). The dataset is conceived to analyze customer behavior and detect patterns of energy consumption that can help to improve the service. The dataset is conformed by three subsets that cover total household consumption, electric water heater consumption, and by-appliance electricity consumption, with sample intervals from one to fifteen minutes. The datetime ranges of the recorded consumptions vary depending on the subset, from some weeks long to some years long. The data was collected by the Uruguayan electricity company (UTE) and studied by Universidad de la República. The presented dataset is a valuable input for researchers in the study of energy consumption patterns, energy disaggregation, the design of energy billing plans, among other relevant issues related to the intelligent utilization of energy in modern smart cities.
In the last decades, cities have increased the number of activities and services that depends on an efficient and reliable electricity service. In particular, households have had a sustained increase of electricity consumption to perform many residential activities. Thus, providing efficient methods to enhance the decision making processes in demand-side management is crucial for achieving a more sustainable usage of the available resources. In this line of work, this article presents an optimization model to schedule deferrable appliances in households, which simultaneously optimize two conflicting objectives: the minimization of the cost of electricity bill and the maximization of users satisfaction with the consumed energy. Since users satisfaction is based on human preferences, it is subjected to a great variability and, thus, stochastic resolution methods have to be applied to solve the proposed model. In turn, a maximum allowable power consumption value is included as constraint, to account for the maximum power contracted for each household or building. Two different algorithms are proposed: a simulation-optimization approach and a greedy heuristic. Both methods are evaluated over problem instances based on real-world data, accounting for different household types. The obtained results show the competitiveness of the proposed approach, which are able to compute different compromising solutions accounting for the trade-off between these two conflicting optimization criteria in reasonable computing times. The simulation-optimization obtains better solutions, outperforming and dominating the greedy heuristic in all considered scenarios.
Featured Application
The methodology described in this article is applicable to design proper management strategies for demand response in smart electricity grids to fairly select water heaters to intervene while guaranteeing the lower discomfort of users.
Abstract
Demand-response techniques are crucial for providing a proper quality of service under the paradigm of smart electricity grids. However, control strategies may perturb and cause discomfort to clients. This article proposes a methodology for defining an index to estimate the discomfort associated with an active demand management consisting of the interruption of domestic electric water heaters. Methods are applied to build the index include pattern detection for estimating the water utilization using an Extra Trees ensemble learning method and a linear model for water temperature, both based on analysis of real data. In turn, Monte Carlo simulations are applied to calculate the defined index. The proposed approach is evaluated over one real scenario and two simulated scenarios to validate that the thermal discomfort index correctly models the impact on temperature. The simulated scenarios consider a number of households using water heaters to analyze and compare the thermal discomfort index for different interruptions and the effect of using different penalty terms for deviations of the comfort temperature. The obtained results allow designing a proper management strategy to fairly decide which water heaters should be interrupted to guarantee the lower discomfort of users.
The increasing rate of penetration of non-conventional renewable energies is affecting the traditional assumption of controllability over energy sources. Power dispatch scheduling methods need to integrate the intrinsic randomness of some new sources, among which, wind energy is particularly difficult to treat. This work aims at the optimal construction of energy bands around wind energy forecasts. Complementarily, a remarkable fact of the proposed technique is that it can be extended to integrate multiple forecasts into a single one, whose band width is narrower at the same level of confidence. The work is based upon a real-world application case, developed for the Uruguayan Electricity Market, a world leader in the penetration of renewable energies.
Modern Smart Cities are highly dependent on an efficient energy service since electricity is used in an increasing number of urban activities. In this regard, Time-of-Use prices for electricity is a widely implemented policy that has been successful to balance electricity consumption along the day and, thus, diminish the stress and risk of shortcuts of electric grids in peak hours. Indeed, residential customers may now schedule the use of deferrable electrical appliances in their smart homes in off-peak hours to reduce the electricity bill. In this context, this work aims to develop an automatic planning tool that accounts for minimizing the electricity costs and enhancing user satisfaction, allowing them to make more efficient usage of the energy consumed. The household energy consumption planning problem is addressed with a multiobjective evolutionary algorithm, for which problem-specific operators are devised, and a set of state-of-the-art greedy algorithms aim to optimize different criteria. The proposed resolution algorithms are tested over a set of realistic instances built using real-world energy consumption data, Time-of-Use prices from an electricity company, and user preferences estimated from historical information and sensor data. The results show that the evolutionary algorithm is able to improve upon the greedy algorithms both in terms of the electricity costs and user satisfaction and largely outperforms to a large extent the current strategy without planning implemented by users.
Solar energy constitutes an effective supplement to traditional energy sources. However, photovoltaic power generation (PVPG) is strongly weather-dependent, and thus highly intermittent. High-precision forecasting of PVPG forms the basis of the production, transmission, and distribution of electricity, ensuring the stability and reliability of power systems. In this work, we propose a deep learning based framework for accurate PVPG forecasting. In particular, taking advantage of the long short-term memory (LSTM) network in solving sequential-data based regression problems, this paper considers the specific domain knowledge of PV and proposes a physics-constrained LSTM (PC-LSTM) to forecast the hourly day-ahead PVPG. It aims to overcome the shortcoming of recent machine learning algorithms that are applied based only on massive data, and thus easily producing unreasonable forecasts. Real-life PV datasets are adopted to evaluate the feasibility and effectiveness of the models. Sensitivity analysis is conducted for the selection of input feature variables based on a two-stage hybrid method. The results indicate that the proposed PC-LSTM model possesses stronger forecasting capability than the standard LSTM model. It is more robust against PVPG forecasting, and more suitable for PVPG forecasting with sparse data in practice. The PC-LSTM model also demonstrates superior performance with higher accuracy of PVPG forecasting compared to conventional machine learning and statistical methods.
Renewable energy is essential for planet sustainability. Renewable energy output forecasting has a significant impact on making decisions related to operating and managing power systems. Accurate prediction of renewable energy output is vital to ensure grid reliability and permanency and reduce the risk and cost of the energy market and systems. Deep learning's recent success in many applications has attracted researchers to this field and its promising potential is manifested in the richness of the proposed methods and the increasing number of publications. To facilitate further research and development in this area, this paper provides a review of deep learning-based solar and wind energy forecasting research published during the last five years discussing extensively the data and datasets used in the reviewed works, the data pre-processing methods, deterministic and probabilistic methods, and evaluation and comparison methods. The core characteristics of all the reviewed works are summarised in tabular forms to enable methodological comparisons. The current challenges in the field and future research directions are given. The trends show that hybrid forecasting models are the most used in this field followed by Recurrent Neural Network models including Long Short-Term Memory and Gated Recurrent Unit, and in the third place Convolutional Neural Networks. We also find that probabilistic and multistep ahead forecasting methods are gaining more attention. Moreover, we devise a broad taxonomy of the research using the key insights gained from this extensive review, the taxonomy we believe will be vital in understanding the cutting-edge and accelerating innovation in this field.
The management of clean energy is usually the key for environmental, economic, and sustainable developments. In the meantime, the energy management system (EMS) ensures the clean energy which includes many sources grouped in a small power plant such as microgrid (MG). In this case, the forecasting methods are used for helping the EMS and allow the high efficiency to the clean energy. The aim of this review paper is providing the necessary data about the basic principles and standards of photovoltaic (PV) power forecasting by stating numerous research studies carried out on the PV power forecasting topic specifically in the short-term time horizon which is advantageous for the EMS and grid operator. At the same time, this contribution can offer a state of the art in different methods and approaches used for PV power forecasting along with a careful study of different time and spatial horizons. Furthermore, this current review paper can support the tenders in the PV power forecasting.
The increased penetration of photovoltaic (PV) generation introduces new challenges for the stability of electricity grids. In this work, machine learning (ML) techniques were implemented to forecast PV power production up to 1-hour ahead with a 10-minute granularity. Three different input combinations were utilised: Model 1 (M1) using the AC power only, Model 2 (M2) using the elevation angle (α), azimuth angle (φ) and AC power and Model 3 (M3) using α, φ, the AC power and satellite observations (SAT) aiming to improve the forecasting performance. Historical PV operational data were used for the training and validation stages of intra-hour PV forecasting models for time t + 10 to 60 minutes ahead. The results obtained over the test set period (15% of the data, i.e. ≈ 110 days) have shown that M2 exhibits the best-performance with a normalised root mean square error (nRMSE) in the range of 7.6% to 14.2%, whereas the skill score (SS) ranged between 6.5% and 30.9% for the 10-to 60-minute ahead, respectively. Index Terms-artificial neural networks, forecasting, PV power.
The power prediction for photovoltaic (PV) power plants has significant importance for their grid connection. Due to PV power’s periodicity and non-stationary characteristics, traditional power prediction methods based on linear or time series models are no longer applicable. This paper presents a method combining the advantages of the wavelet decomposition (WD) and artificial neural network (ANN) to solve this problem. With the ability of ANN to address nonlinear relationships, theoretical solar irradiance and meteorological variables are chosen as the input of the hybrid model based on WD and ANN. The output power of the PV plant is decomposed using WD to separated useful information from disturbances. The ANNs are used to build the models of the decomposed PV output power. Finally, the outputs of the ANN models are reconstructed into the forecasted PV plant power. The presented method is compared with the traditional forecasting method based on ANN. The results shows that the method described in this paper needs less calculation time and has better forecasting precision.
The goal of precipitation nowcasting is to predict the future rainfall
intensity in a local region over a relatively short period of time. Very few
previous studies have examined this crucial and challenging weather forecasting
problem from the machine learning perspective. In this paper, we formulate
precipitation nowcasting as a spatiotemporal sequence forecasting problem in
which both the input and the prediction target are spatiotemporal sequences. By
extending the fully connected LSTM (FC-LSTM) to have convolutional structures
in both the input-to-state and state-to-state transitions, we propose the
convolutional LSTM (ConvLSTM) and use it to build an end-to-end trainable model
for the precipitation nowcasting problem. Experiments show that our ConvLSTM
network captures spatiotemporal correlations better and consistently
outperforms FC-LSTM and the state-of-the-art operational ROVER algorithm for
precipitation nowcasting.
With the increasing use of large-scale grid-connected photovoltaic system, accurate forecast approach for the power output of photovoltaic system has become an important issue. In order to forecast the power output of a photovoltaic system at 24-hour-ahead without any complex modeling and complicated calculation, an artificial neural network based approach is proposed in this paper. The improved back-propagation learning algorithm is adopted to overcome shortcomings of the standard back-propagation learning algorithm. Similar day selection algorithm based on forecast day information is proposed to improve forecast accuracy in different weather types. Forecasting results of a photovoltaic system show that the proposed approach has a great accuracy and efficiency for forecasting the power output of photovoltaic system.
This article presents an approach applying computational intelligence for detecting the use of air conditioners in households. The main objective is determining the intensive use of air conditioning with a high level of confidence. Unsupervised (-means) and supervised (Artificial Neural Networks) approaches are developed for classifying consumers for a case study in Uruguay, using data collected by a smart meters network and open weather data. Data from 29 Uruguayan cities were considered in the period from January 1, 2022, to December 31, 2022. Two thermal models are developed for estimating the temperature inside households. The main results indicate that the proposed approach is able to reach a high classification accuracy, up to 94.5% and a high classification recall, up to 95%, for the considered real case study. The final scope of the work is developing smart tools for classifying consumers, to design and suggest specific commercial products that promote energy efficiency.
In recent years, artificial intelligence methods have been widely applied to solve issues related to renewable energy because of their ability to solve nonlinear and complex data structures. In this paper, we provide a comprehensive bibliometric analysis to better understand the evolution of Artificial Intelligence in Renewable Energy (AI&RE) research from 2006 to 2022. This study is performed based on the Web of Science Core Collection Database, and a dataset of 469 publications have been retrieved. This paper uses VOS viewer, CiteSpace, and Bibliometrix to perform bibliometric analysis and science mapping. The analysis results show that China is the most productive and influential country/region, with the widest range of collaborative partners. The study reveals that AI-related technologies can effectively solve issues related to integrating renewable energy with power system, such as solar and wind forecasting, power system frequency analysis and control, and transient stability assessment. In addition, future research trends are discussed. This paper helps scholars to understand the evolution of AI&RE research from a bibliometric perspective and inspires them to think about the field through multiple aspects.
To combat the worsening global energy shortage, global photovoltaic (PV) installation capacity has been increasing rapidly every year. Since the instability and intermittence of PV power output have great impacts on utility grids, accurate PV power output prediction is crucial. This paper proposes the use of machine learning approaches, combined with a weather type classification method, to predict short-term PV power output. The datasets are collected from a commercial PV power station located in Yangjiang, Guangdong province of China (latitude 21.56 °N, longitude 112.09 °E). Firstly, daytime meteorological data from 07:30 to 18:00 are divided into six 2-h intervals, and then the meteorological conditions of each interval are divided into four categories using an Extremely randomized Trees Classification model according to the PV power generation in each period. Secondly, nine machine learning models are established based on the weather type classification to predict the PV power output. The results show that weather type classification is vital to the selection of appropriate machine learning models and the accurate prediction of PV power output because the characteristic correlation between the meteorological data and PV power output always changes. In general, the Lasso Regressor, Random Forest Regressor, Gradient Boosting Regressor, and Support Vector Regressor models show better performances than the other models. Furthermore, all the models’ accuracy is relatively high when the local meteorological conditions are relatively stable, such as in October, November, and December, during which time the Mean Relative Error values are 2.07, 1.07, and 1.73, respectively. During the period when the weather is unstable, the performance of the SVR model is better than that of the other models. The prediction accuracy can be significantly improved with integrating the accurate weather classification into the model. With regards to each daytime period, the prediction accuracy in the morning and evening is relatively high and the MREs for these times are small. This study provides a theoretical basis for selecting appropriate machine learning models to predict photovoltaic power generation under different weather conditions.
This article presents the evaluation of multicriteria planning heuristics for demand response in datacenters and supercomputing facilities. This is a relevant problem for science nowadays, when the growing application of cutting-edge technologies (numerical methods, big data processing, artificial intelligence, smart systems, etc.) has raised the energy demands in datacenters. The proposed approach involves a negotiation mechanism for colocation datacenters, where the datacenter operator agrees prices and quality of service with a group of tenants. Twelve different multicriteria heuristics are proposed for planning using both local and global information at tenants and datacenter operator levels. The proposed approach is evaluated applying simulations over realistic scenarios considering different tenant sizes and heterogeneity levels that model different business models for datacenters. Several metrics are computed and Pareto analysis is provided. The main results indicate that accurate trade-off values between the problem objectives are obtained, offering different options for decision making. The proposed approach provides a useful and applicable method for demand response planning in modern datacenters.
Alternative energy sources are becoming more and more common around the world. In order to reduce environmental pollution and CO 2 emissions, in addition to being an ideal solution to overcome the energy crisis. In this context, power energy stands out, as it is the most abundant and most widely available natural resource on the entire planet. Due to the high level of uncertainty of the factors that directly interfere in the generation of solar power, such as temperature and solar radiation, make predictions of solar power with high precision is a challenge. Thus, the objective of this article is to develop a forecasting model, through time series, that makes it possible to predict the production of power energy, using a database collected in a photovoltaic plant in Uruguay. For the development of the proposal, models (base-learners), pre-processing techniques and models (meta-learners) used in the Stacking-Ensemble Learnig (STACK) method were used, which were compared using the measurements of performance Relative Root Mean Square Error (RRMSE), Symmetric Mean Absolute Percentage Error (sMAPE) and Determination Coefficient (R 2 ) in addition to statistical tests. In the end, it can be concluded that the combination Correlation Matrix (CORR) and Language Model (LM), from Layer-0 obtained the best results, in the three performance measures and the combination of models (base-learners) and pre-processing techniques (Layer-0) presented the best results when compared to Layer-1, obtaining satisfactory values in all performance measures.
Integration of photovoltaics into power grids is difficult as solar energy is highly dependent on climate and geography; often fluctuating erratically. This causes penetrations and voltage surges, system instability, inefficient utilities planning and financial loss. Forecast models can help; however, time stamp, forecast horizon, input correlation analysis, data pre and post-processing, weather classification, network optimization, uncertainty quantification and performance evaluations need consideration. Thus, contemporary forecasting techniques are reviewed and evaluated. Input correlational analyses reveal that solar irradiance is most correlated with Photovoltaic output, and so, weather classification and cloud motion study are crucial. Moreover, the best data cleansing processes: normalization and wavelet transforms, and augmentation using generative adversarial network are recommended for network training and forecasting. Furthermore, optimization of inputs and network parameters, using genetic algorithm and particle swarm optimization, is emphasized. Next, established performance evaluation metrics MAE, RMSE and MAPE are discussed, with suggestions for including economic utility metrics. Subsequently, modelling approaches are critiqued, objectively compared and categorized into physical, statistical, artificial intelligence, ensemble and hybrid approaches. It is determined that ensembles of artificial neural networks are best for forecasting short term photovoltaic power forecast and online sequential extreme learning machine superb for adaptive networks; while Bootstrap technique optimum for estimating uncertainty. Additionally, convolutional neural network is found to excel in eliciting a model's deep underlying non-linear input-output relationships. The conclusions drawn impart fresh insights in photovoltaic power forecast initiatives, especially in the use of hybrid artificial neural networks and evolutionary algorithms.
This article describes the national initiative for installing and operating a collaborative scientific HPC infrastructure in Uruguay (Cluster-UY). The project was conceived as a mean to foster research and innovation projects that face complex problems with high computing demands. The main ideas and motivations of the Cluster-UY project are described. The technological decisions to install the platform are explained and the collaborative operation model to guarantee sustainability is introduced. In addition, the perspectives of the national scientific HPC initiative are highlighted and sample current projects are presented.
Renewable energy sources (RESs) and energy storage systems (ESSs) are the key technologies for smart grid applications and provide great opportunities to de-carbonize urban areas, regulate frequency, voltage deviations, and respond to severe time when the load exceeds the generation. Nevertheless, uncertainty and inherent intermittence of renewable power generation units impose severe stresses on power systems. Energy storage systems such as battery energy storage system enables the power grid to improve acceptability of intermittent renewable energy generation. To do so, a successful coordination between renewable power generation units, ESSs and the grid is required. Nonetheless, with the existing grid architecture, achieving the aforementioned targets is intangible. In this regard, coupling renewable energy systems with different generation characteristics and equipping the power systems with the battery storage systems require a smooth transition from the conventional power system to the smart grid. Indeed, this coordination requires not only robust but also innovative controls and models to promote the implementation of the next-generation grid architecture. In this context, the present research proposes a smart grid architecture depicting a smart grid consisting of the main grid and multiple embedded micro-grids. Moreover, a focus has been given to micro-grid systems by proposing a “Micro-grid Key Elements Model” (MKEM). The proposed model and architecture are tested and validated by virtualization. The implementation of the virtualized system integrates solar power generation units, battery energy storage systems with the proposed grid architecture. The virtualization of the proposed grid architecture addresses issues related to Photovoltaic (PV) penetration, back-feeding, and irregularity of supply. The simulation results show the effect of Renewable Energy (RE) integration into the grid and highlight the role of batteries that maintain the stability of the system.
Predicting the short-term power output of a photovoltaic panel is an important task for the efficient management of smart grids. Short-term forecasting at the minute scale, also known as nowcasting, can benefit from sky images captured by regular cameras and installed close to the solar panel. However, estimating the weather conditions from these images—sun intensity, cloud appearance and movement, etc.—is a very challenging task that the community has yet to solve with traditional computer vision techniques. In this work, we propose to learn the relationship between sky appearance and the future photovoltaic power output using deep learning. We train several variants of convolutional neural networks which take historical photovoltaic power values and sky images as input and estimate photovoltaic power in a very short term future. In particular, we compare three different architectures based on: a multi-layer perceptron (MLP), a convolutional neural network (CNN), and a long short term memory (LSTM) module. We evaluate our approach quantitatively on a dataset of photovoltaic power values and corresponding images gathered in Kyoto, Japan. Our experiments reveal that the MLP network, already used similarly in previous work, achieves an RMSE skill score of 7% over the commonly-used persistence baseline on the 1-min future photovoltaic power prediction task. Our CNN-based network improves upon this with a 12% skill score. In contrast, our LSTM-based model, which can learn the temporal dependencies in the data, achieves a 21% RMSE skill score, thus outperforming all other approaches.
This paper reviews state-of-the-art on wind speed/power forecasting and solar irradiance forecasting with ensemble methods. The ensemble forecasting methods are grouped into two main categories: competitive ensemble forecasting and cooperative ensemble forecasting. The competitive ensemble forecasting is further categorized based on data diversity and parameter diversity. The cooperative ensemble forecasting is divided according to pre-processing and post-processing. Typical articles are discussed according to each category and their characteristics are highlighted. We also conduct comparisons based on reported results and comparisons based on simulations conducted by us. Suggestions for future research include ensemble of different paradigms and inter-category ensemble methods among others.
Most of the major renewable power sources, such as solar, wind, and ocean wave, are variable in time. Variability arises from local turbulence, weather patterns, clouds, diurnal variations, and seasonal variations. The variable, non-dispatchable nature of renewable power sources requires an increase in the utility reserve requirement to maintain reliability. Power sources with greater minute-to-minute and hourly variation require greater reserves. This paper presents a methodology for quantifying the variability of renewable power sources based on calculating the utility reserve requirement over short and medium time scales.
A feedforward neural network which can account for nonlinear relationships was used to compare ARIMA and neural network price forecasting performance. Data used was monthly live cattle and wheat prices from 1950 through 1990. The experiment was repeated seven times for successive three year periods. This involved using a walk forward or sliding window approach from 1970 through 1990 which generated out of sample results. The neural network models achieved a 27 percent and 56 percent lower mean squared error than ARIMA model. The absolute mean error and mean absolute percent error were also lower for the neural network models. The neural network models were able to capture a significant number of turning points for both wheat and cattle, while the ARIMA model was only able to do so for wheat. Since this forecasting method is not problem specific and uses only past prices, it can be applied to other forecasting problems such as stocks and other financial prices.
Forecasting solar photovoltaic power generation: a real-world case study in Uruguay
- Sergio Nesmachnow
- S Risso