No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Load forecasting in the user side using wavelet-ANFIS
Abstract
At present, the intelligent energy management systems (IEMS) are used to maximiz the relation between productivity and cost using a variety of energy sources. In this work, we present a method of short-time load forecasting, using the ANFIS model and a component of preprocessing based in the discrete wavelet transform; the models was implemented in the user-side, analyzing real data of a factory in order to test the proposed algorithm.
... The prediction based on the artificial neural networks, was widely accepted by the scientific and engineering spheres, becoming the most widespread technique for the load forecasting. There are several scientific publications that prove the quality and robustness of predictions based on neural networks (Chen et al., 1996;Giacometto et al., 2012;Hippert et al., 2001;Lino et al., 2016). ...
... The period under review belongs to winter of 2014 and was: One day (4 th of February), one week (3 rd to 9 th of February) and two weeks (3 rd to 16 th of February). The results are satisfactory considering that the maximum value of MAPE in the forecast studies of this type should not exceed 5% (Giacometto et al., 2012;Hippert et al., 2001).at this point was small: 0.70%. ...
Currently, load forecasting is a fundamental task for planning, operation and exploration of the electric power systems. The importance of forecasting has become more evident with the restructuring of the national energy sector, thus, promoting projects linked to smart grids, namely in Portugal-InovGrid. This study proposes the computational forecast model of the load diagram based on the Levenberg-Marquardt algorithm of Artificial Neural Networks. The used data are the time series of active power, recorded by EDP Distribution Telemetry System, and the climatic time series of the Portuguese Institute of the Sea and Atmosphere, collected on the city of Évora. The forecast horizon is short term: from one hour to a week. The results showed that main statistical error parameter (mean absolute percentage error) was not exceed 5%.
... In layer three, the normalization of the strengths is performed, while in layer four the output of each node is calculated by the mathematical product of the normalized strengths and polynomials. Finally in layer 5, the output of the ANFIS is computed as the summation of all incoming signals from the fourth layer [12]. In the presented methodology the algorithm is implemented to calculate the different mathematical models of the energy demand of the plant. ...
This paper presents an energy optimization methodology applied on industrial plants with multiple energy carriers. The methodology combines an adaptive neuro-fuzzy inference system to calculate the short-term load forecasting of a plant, and the sequential quadratic programming algorithm to optimize its energy flow. Furthermore, the mathematical models of the plant's equipment are considered into the optimization process, in order to calculate the dynamic system response and the equipment's inertias. The final algorithm optimizes the operation of the plant in order to satisfy the energy demand, minimizing several optimization criteria. The methodology has been tested and evaluated in an automotive factory plant using real production and consumption data. Keywords—energy optimization; energy hub; energy prediction; adaptive neuro-fuzzy inference system; sequential quadratic programming algorithm.
The accurate modelling and prediction of air temperature values is an exceptionally important meteorological variable that affects in many areas. The present study is aimed at developing models for the prediction of monthly mean air temperature values in Turkey using ANN, ANFIS and SVMr methods. In developing the models, the monthly data derived from eight stations of the TSMS for the 1963–2015 period were used, including latitude, longitude, elevation, month, and minimum, maximum and mean air temperatures. The performances of the ANN, ANFIS and SVMr models were compared using R², MSE, MAPE and RRMSE. In order to verify the differences between the predicted temperature values provided by the ANN, ANFIS and SVMr models and the observed temperature values derived from the stations, a t-test analysis was conducted, and the best ANN, ANFIS and SVMr models were determined according to the statistical performance values. These models were then used to make air temperature predictions for the cities. Manova was carried out to determine the effects of the differences temperature predictions and RRMSE values of the models. Generally, the statistical performance values of the ANFIS models were found to be slightly better than those of the ANN and SVMr models.
Load forecast is becoming currently more fundamental in planning, operating and controlling of modern electric power systems. Nowadays the load peak forecast is also important, because it is of great interest in economy stability and improvement in the electrical systems. This paper presents an approach for load forecasting in the medium voltage distribution network in Portugal. The forecast horizon is short term, from 24 hours up to a week. The forecast method is based on the use of artificial neural networks. The study was done with the time series of telemetry data of EDP medium voltage substations, applied for the urban areas of Evora and Oporto. The performance of the proposed methodology is illustrated by graphical results and evaluated with statistical indicators. The mean absolute percentage error was lower than 5%, meaning that chosen methodology clearly validate the feasibility of the test.
Electrical energy consumption is affected by many parameters. These includes the variables related to power system itself, weather and climatic factors and socio-economic being of the energy consumers. In this paper, two components of load forecasting are classified. The parameters that influence the energy consumption and the methods used to forecast the energy consumption are reviewed. It is observed that, many factors have great influence on the energy consumption, and the forecasting accuracy depends on the amount of data used. Also the methods applied contribute in the forecasting accuracy and complexity of the method. It is therefore important to use large data, and apply an appropriate method (technique) while forecasting electrical energy. A lot of methods are reviewed, from time series method to artificial intelligence with varying parameters, most of which are weather related, demography of the area, economy class of the consumers and the history of electrical energy consumed.
Short term load forecast provides market participants the opportunity to balance their generation and/or consumption needs and contractual obligation one day in advance. It also helps to determine reference price for electricity energy and provide system operator a balanced system. This paper presents a comparative study of ANFIS and ANN methods for short term load forecast. Using the load, season and temperature data of Turkey between years of 2009-2011, the prediction is carried out for 2012. The mean absolute percentage errors for ANFIS and ANN models are found as 1.85 and 2.02, respectively in all days except holidays of 2012.
A novel hybrid approach, combining wavelet transform, particle swarm optimization, and adaptive-network-based fuzzy inference system, is proposed in this paper for short-term electricity prices forecasting in a competitive market. Results from a case study based on the electricity market of mainland Spain are presented. A thorough comparison is carried out, taking into account the results of previous publications. Finally, conclusions are duly drawn.
Load forecast has been a central and an integral process in the planning and operation of electric utilities. Many techniques and approaches have been investigated to tackle this problem in the last two decades. These are often different in nature and apply different engineering considerations and economic analyses. In this paper a comparative evaluation of five short-term load forecasting techniques is presented. These techniques are: 1. Multiple Linear Regression; 2. Stochastic Time Series; 3. General Exponential Smoothing; 4. State Space Method; and 5. Knowledge-Based Approach. The authors have applied these algorithms to obtain hourly load forecasts (for up to 24 hours) during the winter and summer peaking seasons. Thus the five forecasting methodologies have been applied to the same database and their performances are directly compared. The forecast error analyses are provided in Tables 1 and 2 for the winter and summer days respectively. As these results are based on forecasts of two single days, these should be used for comparative purposes only. Some interesting observations are made about the results presented in Tables 1 and 2. For example, for the peak summer day the transfer function (TF) approach gave the best result, whereas for the peak winter day the TF approach resulted in the next to the worst accuracy. During the peak summer day the temperature profile was typical whereas during the peak winter day the profile was unseasonal.
Multiresolution representations are effective for analyzing the
information content of images. The properties of the operator which
approximates a signal at a given resolution were studied. It is shown
that the difference of information between the approximation of a signal
at the resolutions 2<sup>j+1</sup> and 2<sup>j</sup> (where j is an
integer) can be extracted by decomposing this signal on a wavelet
orthonormal basis of L <sup>2</sup>( R <sup>n</sup>), the
vector space of measurable, square-integrable n -dimensional
functions. In L <sup>2</sup>( R ), a wavelet orthonormal
basis is a family of functions which is built by dilating and
translating a unique function ψ( x ). This decomposition
defines an orthogonal multiresolution representation called a wavelet
representation. It is computed with a pyramidal algorithm based on
convolutions with quadrature mirror filters. Wavelet representation lies
between the spatial and Fourier domains. For images, the wavelet
representation differentiates several spatial orientations. The
application of this representation to data compression in image coding,
texture discrimination and fractal analysis is discussed
The artificial neural network (ANN) technique for short term load forecasting (STLF) has been proposed by several authors, and gained a lot of attention recently. In order to evaluate ANN as a viable technique for STLF, one has to evaluate the performance of ANN methodology for practical considerations of STLF problems. This paper makes an attempt to address these issues. The paper presents the results of a study to investigate whether the ANN model is system dependent, and/or case dependent. Data from two utilities were used in modeling and forecasting. In addition, the effectiveness of a next 24 hour ANN model in predicting 24 hour load profile at one time was compared with the traditional next one hour ANN model.
This paper presents the architecture and learning procedure underlying ANFIS (Adaptive-Network-based Fuzzy Inference System), a fuzzy inference system implemented in the framework of adaptive networks. By using a hybrid learning procedure, the proposed ANFIS can construct an input-output mapping based on both human knowledge (in the form of fuzzy if-then rules) and stipulated input-output data pairs. In our simulation, we employ the ANFIS architecture to model nonlinear functions, identify nonlinear components on-linely in a control system, and predict a chaotic time series, all yielding remarkable results. Comparisons with artificail neural networks and earlier work on fuzzy modeling are listed and discussed. Other extensions of the proposed ANFIS and promising applications to automatic control and signal processing are also suggested. I. Introduction System modeling based on conventional mathematical tools (e.g., differential equations) is not well suited for dealing with ill-defin...
A mathematical tool to build a fuzzy model of a system where fuzzy implications and reasoning are used is presented. The premise of an implication is the description of fuzzy subspace of inputs and its consequence is a linear input-output relation. The method of identification of a system using its input-output data is then shown. Two applications of the method to industrial processes are also discussed: a water cleaning process and a converter in a steel-making process.
A mathematical tool to build a fuzzy model of a system where fuzzy implications and reasoning are used is presented. The premise of an implication is the description of fuzzy subspace of inputs and its consequence is a linear input-output relation. The method of identification of a system using its input-output data is then shown. Two applications of the method to industrial processes are also discussed: a water cleaning process and a converter in a steel-making process.
This paper presents two hybrid neural networks derived from fuzzy neural networks (FNN): wavelet fuzzy neural network (WFNN) using the fuzzified wavelet features as the inputs to FNN and fuzzy neural network (FNCI) employing the Choquet integral as the outputs of FNN. The learning through FNCI is simplified by the use of q-measure and the speed of convergence of the parameters is increased by reinforced learning. The underlying fuzzy models of these hybrid networks are a modified form of fuzzy rules of Takagi-Sugeno model. The number of fuzzy rules is found from a fuzzy curve corresponding to each input-output by counting the total number of peaks and troughs in the curve. The models can forecast hourly load with a lead time of 1 h as they deal with short-term load forecasting. The results of the two hybrid networks using Indian utility data are compared with ANFIS and other conventional methods. The performance of the proposed WFNN is found superior to all the other compared methods.
Short-term load forecast (STLF) is a key issue for operation of both regulated power systems and electricity markets. In spite of all performed research in this area, there is still an essential need for more accurate and robust load forecast methods. In this paper, a new hybrid forecast method is proposed for this purpose, composed of wavelet transform (WT), neural network (NN) and evolutionary algorithm (EA). Hourly load time series usually consists of both global smooth trends and sharp local variations, i.e. low- and high-frequency components. WT can efficiently decompose the time series into its components. Each component is predicted by a combination of NN and EA and then by inverse WT the hourly load forecast is obtained. The proposed method is examined on three practical power systems and compared with some of the most recent STLF methods.
Both planning and design of municipal solid waste management systems (MSWMS) require accurate prediction of waste generation (WG). In this study, the hybrid of wavelet transform-adaptive neuro-fuzzy inference system (WT-ANFIS) and wavelet transform-artificial neural network (WT-ANN) is used to predict the weekly WG in Tehran, concerning complexity and dynamic MSWMS. In order to input variables preprocessing is done by WT and then new variables entered to ANFIS and ANN models. Consequently, output uncertainty of WT-ANFIS and WT-ANN models is done. The results achieved in this research indicate the positive effect of input variables preprocessing by WT in the prediction of weekly WG in Tehran, and it has led to noticeable increase in the accuracy of two model calculations. However, WT-ANFIS model had better results than WT-ANN model, because of the smaller uncertainty than WT-ANN model.
The artificial neural network (ANN) technique for short-term load forecasting (STLF) has been proposed previously. In order to evaluate ANNs as a viable technique for STLF, one has to evaluate the performance of ANN methodology for practical considerations of STLF problems. The authors make an attempt to address these issues. The results of a study to investigate whether the ANN model is system dependent, and/or case dependent, are presented. Data from two utilities are used in modeling and forecasting. In addition, the effectiveness of a next 24 h ANN model in predicting 24 h load profile at one time was compared with the traditional next 1 h ANN model
This work presents an electricity consumption-forecasting framework configured automatically and based on an Adaptative Neural Network Inference System (ANFIS). This framework is aimed to be implemented in industrial plants, such as automotive factories, with the objective of giving support to an Intelligent Energy Management System (IEMS). The forecasting purpose is to support the decision-making (i.e. scheduling workdays, on–off production lines, shift power loads to avoid load peaks, etc.) to optimize and improve economical, environmental and electrical key performance indicators. The base structure algorithm, the ANFIS algorithm, was configured by means of a Multi Objective Genetic Algorithm (MOGA), with the aim of getting an automatic-configuration system modelling. This system was implemented in an independent section of an automotive factory, which was selected for the high randomness of its main loads. The time resolution for forecasting was the quarter hour. Under these challenging conditions, the autonomous configuration, system learning and prognosis were tested with success.
This paper aims to develop a load forecasting method for short-term load forecasting, based on an adaptive two-stage hybrid network with self-organized map (SOM) and support vector machine (SVM). In the first stage, a SOM network is applied to cluster the input data set into several subsets in an unsupervised manner. Then, groups of 24 SVMs for the next day's load profile are used to fit the training data of each subset in the second stage in a supervised way. The proposed structure is robust with different data types and can deal well with the nonstationarity of load series. In particular, our method has the ability to adapt to different models automatically for the regular days and anomalous days at the same time. With the trained network, we can straightforwardly predict the next-day hourly electricity load. To confirm the effectiveness, the proposed model has been trained and tested on the data of the historical energy load from New York Independent System Operator.
This paper proposes a novel technique to forecast day-ahead electricity prices based on the wavelet transform and ARIMA models. The historical and usually ill-behaved price series is decomposed using the wavelet transform in a set of better-behaved constitutive series. Then, the future values of these constitutive series are forecast using properly fitted ARIMA models. In turn, the ARIMA forecasts allow, through the inverse wavelet transform, reconstructing the future behavior of the price series and therefore to forecast prices. Results from the electricity market of mainland Spain in year 2002 are reported.
The importance of short-term load forecasting has been increasing lately. With deregulation and competition, energy price forecasting has become a big business. Bus-load forecasting is essential to feed analytical methods utilized for determining energy prices. The variability and nonstationarity of loads are becoming worse, due to the dynamics of energy prices. Besides, the number of nodal loads to be predicted does not allow frequent interactions with load forecasting experts. More autonomous load predictors are needed in the new competitive scenario. This paper describes two strategies for embedding the discrete wavelet transform into neural network-based short-term load forecasting. Its main goal is to develop more robust load forecasters. Hourly load and temperature data for North American and Slovakian electric utilities have been used to test the proposed methodology.
Load forecasting has become one of the major areas of research in
electrical engineering, and most traditional forecasting models and
artificial intelligence techniques have been tried out in this task.
Artificial neural networks (NNs) have lately received much attention,
and a great number of papers have reported successful experiments and
practical tests with them. Nevertheless, some authors remain skeptical,
and believe that the advantages of using NNs in forecasting have not
been systematically proved yet. In order to investigate the reasons for
such skepticism, this review examines a collection of papers (published
between 1991 and 1999) that report the application of NNs to short-term
load forecasting. Our aim is to help to clarify the issue, by critically
evaluating the ways in which the NNs proposed in these papers were
designed and tested
Forecasting the short-term load entails the construction of a
model, and, using the information available, estimating the parameters
of the model to optimize the prediction performance. It follows that the
more closely the chosen model approximates the actual physical
generating process, the higher the expected performance of the
forecasting system. In this paper it is postulated that the load can be
modeled as the output of some dynamic system, influenced by a number of
weather, time and other environmental variables. Recurrent neural
networks, being members of a class of connectionist models exhibiting
inherent dynamic behavior, can thus be used to construct empirical
models for this dynamic system. Because of the nonlinear dynamic nature
of these models, the behavior of the load prediction system can be
captured in a compact and robust representation. This is illustrated by
the performance of recurrent models on the short-term forecasting of the
nation-wide load for the South African utility, ESKOM. A comparison with
feedforward neural networks is also given
The use of a weighted least squares procedure when training a
neural network to solve the short-term load forecasting (STLF) problem
is investigated. Our results indicate that a neural network that
implements the weighted least squares procedure outperforms a neural
network that implements the least squares procedure during the on-peak
period for the two performance criteria specified; MAE% and COST, during
the entire period for the COST criterion. It is therefore, recommended
that the weighted least squares procedure be further studied by electric
utilities which use neural networks to forecast their short-term load,
and experience large variabilities in their hourly marginal energy costs
during a 24-hour period
The ability to solve the short-term load forecasting (STLF)
problem with artificial neural networks (ANNs) is investigated by
conducting a fractional factorial experiment. The results of the
experiment are analyzed, and the factors and factor interactions that
affect forecast errors are identified and quantified. From the analysis,
we derive rules for building a `quasi optimal' neural network to solve
the STLF problem. A comparison study demonstrates the superior
performance of the `quasi optimal' neural network over an automated
Box-Jenkins seasonal ARIMA model in solving the STLF problem
The authors present an artificial neural network (ANN) model for
forecasting weather-sensitive loads. The proposed model is capable of
forecasting the hourly loads for an entire week. The model is not fully
connected; hence, it has a shorter training time than the fully
connected ANN. The proposed model can differentiate between the weekday
loads and the weekend loads. The results indicate that this model can
achieve greater forecasting accuracy than the traditional statistical
model. This ANN model has been implemented on real load data. The
average percentage peak error for the test cases was 1.12%
Existing studies on 1-24 hr load forecasting algorithms are
reviewed, and an expert-system-based algorithm is presented as an
alternative. The logical and syntactical relationships between weather
and load as well as the prevailing daily load shapes have been examined
to develop the rules for this approach. Two separate, but similar,
algorithms have been developed to provide 1-6 hr and 24 hr forecasts.
These forecasts have been compared with observed hourly load data for a
Virginia electric utility for all seasons of the year. The 1 hr and 6 hr
forecast errors (absolute average) ranged from 0.869% to 1.218% and from
2.437% to 3.48% respectively. The 24 hour forecast errors (absolute
average) ranged from 2.429% to 3.300%
An expert system based algorithm
- S Rahman
- R Bhatnagar
S. Rahman, and R. Bhatnagar, "An expert system based algorithm