Hourly Load Forecasting Research Articles

Probabilistic short-term load forecasting models based on a parametric approach

The forecasting of electric load plays an essential role in the effective management of electric power systems. Specifically, short-term models, which predict hourly load over the 24 hours of the subsequent day, hold significant value for applications within the realm of electricity markets. In this context, research efforts have predominantly concentrated on the development of point load forecasting models. These models provide solely the estimated value of hourly load, omitting any information regarding its associated uncertainty. Probabilistic forecasting models aim to address this limitation by offering comprehensive information on forecasted values, including their associated uncertainty, thereby enabling their more effective utilization in risky decision-making environments. This paper presents a parametric probabilistic model designed for hourly load forecasting. The model is refined through a multi-objective genetic algorithm optimization process that identifies explanatory variables from a specified set. The selected variables are combined linearly to predict the parameters of a probability distribution function for the hourly load. The process also selects the type of distribution from among those characterized by two parameters. The model is applied to data from a real distribution substation, yielding superior forecasting evaluation indexes compared to those achieved by two benchmark probabilistic load forecasting models.

Short‐term load and spinning reserve prediction based on LSTM and ANFIS with PSO algorithm

AbstractWith the increase in population and the growth of technology, the load demand has increased and major changes in spinning reserve are unavoidable. Short‐term forecasting to hourly predict the required load and spinning reserve is of great importance. All of the power system studies in planning and operation fields are depend on short‐term hourly load forecasting. In this work, the problem of load forecasting and spinning reserve based on deep learning (DL) algorithms and traditional methods is investigated with the help of the proposed information combination system. The proposed method tries to reduce the weaknesses of the stated methods and increase the accuracy of the predicted signal. First, short‐term predicting of load and spinning reserve is performed using a combination of adaptive network‐based fuzzy inference system (ANFIS) and meta‐heuristic algorithms including differential evolution (DE), genetic algorithm (GA), and particle swarm optimization (PSO). The ANFIS‐PSO is selected as the best ANFIS combination in load and spinning reserve prediction with a lower error criterion than other methods. Also, the long short‐term memory (LSTM) network can provide good accuracy for load and spinning reserve forecasting. Therefore, the combination of ANFIS‐PSO and LSTM is used to reduce the average error and error variance.

Demand response modeling in a day-ahead wholesale electricity market in Japan, considering the impact of customer risk aversion and dynamic price elasticity of demand

This study aims to develop a mathematical modeling approach to maximize the welfare of the price-responsive customers (CUs) in a wholesale electricity market in Tokyo, Japan. The contributions made by this paper in the quest to determine the role of CUs in a demand response market are twofold. First, the aversion of the CUs to the risk of choosing the Demand Response Programs (DPRs) is taken into account by considering their expected utility from consuming electricity. The proposed model is founded on the customer theory in microeconomics, using the concept of the expected utility function to model the behavior of the risk-averse CUs in response to different DPRs. Second, it introduces an hourly-based model for the short-term price elasticity of demand, considering the day-ahead price mechanism defined in the Japan Electric Power Exchange (JEPX) market. The estimated price elasticities are used in a price elasticity matrix of demand (PEMD) to precisely reflect the different response strategies, including flexible, in-flexible forward-shifting, backward shifting, and optimizing responses. An accurate day-ahead hourly load forecasting is performed, using the Seasonal Autoregressive Integrated Moving Average (SARIMA) model, which is trained on four years of data provided by the Tokyo Electric Power Company (TEPCO), with a mean absolute percentage error (MAPE) of 0.94%. The developed model is used to analyze the CUs’ behaviors with different response strategies in the JEPX market in Tokyo. By applying the Time-of-Use (TOU) and Real-Time-Pricing (RTP) programs, the results reveal a peak reduction potential of 10.7% and 7.3%, respectively, for the flexible CUs. Applying the RTP program to the curtailable loads can achieve a 7.7% reduction in daily peak demand and a 1.6% reduction in daily electricity consumption.

A data-driven knowledge-based system with reasoning under uncertain evidence for regional long-term hourly load forecasting

Reliable long-term hourly load forecasting is imperative for electricity utilities and planners' decision-making, especially in generation-transmission expansion planning. This paper develops a data-driven knowledge-based system for regional hourly load forecasting in the long-term horizon with contributions in both knowledge base and inference engine. In this regard, in the knowledge base, two types of driving factors are taken into account to determine the forecasting model: long-term trend-related features such as macro-economic factors and short-term factors such as temperature. The knowledge base is decomposed into three-step-related parts to deal with these groups' substantial differences in frequency. Firstly, the long-term trends are identified by the tendency variables. Then, the short-term variability is considered by temperature-related variables in addition to day type. Machine learning regression methods, support vector machines, random forest, and artificial neural networks are compared to determine the non-linear relationship of variables in these steps. The results of these steps are combined in the third step. Then, in the inference engine, a new reasoning method based on fuzzy logic is represented to forecast the regional long-term hourly load under uncertain evidence of temperature as an effective non-predictable feature in the long run. To evaluate the performance of the proposed system, a comparison against five systems is conducted. The results show the superiority of this system compared to the other systems for a publicly available dataset (ISO New England electricity market) as well as Iran’s residential, commercial and agricultural electricity load.

Short-Term Net Load Forecasting with Singular Spectrum Analysis and LSTM Neural Networks

Short-term electricity load forecasting is key to the safe, reliable, and economical operation of power systems. An important challenge that arises with high-frequency load series, e.g., hourly load, is how to deal with the complex seasonal patterns that are present. Standard approaches suggest either removing seasonality prior to modeling or applying time series decomposition. This work proposes a hybrid approach that combines Singular Spectrum Analysis (SSA)-based decomposition and Artificial Neural Networks (ANNs) for day-ahead hourly load forecasting. First, the trajectory matrix of the time series is constructed and decomposed into trend, oscillating, and noise components. Next, the extracted components are employed as exogenous regressors in a global forecasting model, comprising either a Multilayer Perceptron (MLP) or a Long Short-Term Memory (LSTM) predictive layer. The model is further extended to include exogenous features, e.g., weather forecasts, transformed via parallel dense layers. The predictive performance is evaluated on two real-world datasets, controlling for the effect of exogenous features on predictive accuracy. The results showcase that the decomposition step improves the relative performance for ANN models, with the combination of LSTM and SAA providing the best overall performance.

A Method of Hourly Load Forecasting by Time Series by Recycle of Predicted Values

A Method of Hourly Load Forecasting by Time Series by Recycle of Predicted Values

Power Load Forecasting using Back Propagation Algorithm

The day today operation and scheduling activities of a power generation unit requires the forecast of power demand to meet the needs of the consumers. Load forecasting is categorized into short, medium and long term forecasts. The short term prediction refers to hourly load forecast for the period lasting from an hour to more than a few days. The medium term forecasts is prediction of power load ranging from one to several months ahead. Finally, the long term forecast indicates prediction of power load ranging from one to many years in near future. Excellence of short term hourly power load forecast has a considerable effect on cost effective functioning of power plants because numerous assessments are dependent on these forecasts. These assessments comprise cost effective scheduling of power plants, scheduling of coal acquisition, power plant security consideration, and planning for energy related business. The significance of precise load forecasts will reduce the wastage of power in the future because of remarkable changes taking place in the structure of power generation industries due to deregulation and heavy competition. This situation creates thrust on the power generation sectors to function at maximum possible efficiency, which leads to precise forecasting of the power load.

A New Input Selection Algorithm Using the Group Method of Data Handling and Bootstrap Method for Support Vector Regression Based Hourly Load Forecasting

Electric load forecasting is indispensable for the effective planning and operation of power systems. Various decisions related to power systems depend on the future behavior of loads. In this paper, we propose a new input selection procedure, which combines the group method of data handling (GMDH) and bootstrap method for support vector regression based hourly load forecasting. To construct the GMDH network, a learning dataset is divided into training and test datasets by bootstrapping. After constructing GMDH networks several times, the inputs that appeared frequently in the input layers of the completed networks were selected as the significant inputs. Filter methods based on linear correlation and mutual information (MI) were employed as comparison methods, and the performance of hybrids of the filter methods and the proposed method were also confirmed. In total, five input selection methods were compared. To verify the performance of the proposed method, hourly load data from South Korea was used and the results of one-hour, one-day and one-week-ahead forecasts were investigated. The experimental results demonstrated that the proposed method has higher prediction accuracy compared with the filter methods. Among the five methods, a hybrid of an MI-based filter with the proposed method shows best prediction performance.

Optimization of neural network parameters by Stochastic Fractal Search for dynamic state estimation under communication failure

A powerful metaheuristic technique is proposed that uses a mathematical concept called Stochastic Fractal Search (SFS) to ensure fast convergence along with accuracy. This paper intends to determine the optimal set of multilayer perceptron neural network (MLP) parameters (weights and thresholds) to improve the performance of MLP by using the SFS technique. The SFS is used because of its effective search in finding the global minima and therefore, it avoids the MLP neural network trapped in local minima. The hybrid approach (MLP-SFS) is applied to solve the dynamic state estimation (DSE) problem at the filtering stage. DSE amalgamates forecasting procedure with measurement data to precisely assess the system state. The approach classifies the process into three stages. In the first stage, a short term hourly load forecasting is applied using support vector machine (STLF-SVM) for time series to forecast the unavailable load data due to communication failure from the previous hourly historical data load. The second stage constitutes an optimal power flow (OPF) that is used to determine the minimum cost generation dispatch to serve the given load and convert the obtained loads, and generations into measurement data. The third stage has a filtering process, which uses SFS technique to optimize the MLP neural network parameters (weights and thresholds) to estimate the system state. The hybrid MLP-SFS is used to find the optimal connection weights and thresholds for the MLP neural network. Following this, a simple backpropagation neural network (BPN) will adjust the final parameters. The approach is tested on IEEE 14-and 118-bus systems using realistic load patterns from the New York Independent System Operator (NYISO) under several scenarios of measurement error and communication failure. The mean absolute percentage error index of the system state (phase and magnitude voltage) is used to determine the accuracy of the approach (MLP-SFS). Results of the proposed approach (MLP-SFS) are compared with non-optimized MLP (random weights and thresholds) and other methods, such as, optimized MLP based on genetic algorithm (MLP-GA) and Particle Swarm Optimization (MLP-PSO) individually. The results indicate that the hybrid (MLP-SFS) increases the precision by about 20%–50% and reduces the computational time around by 30%–50%, which is good for real-time applications, such as, security assessment and contingency evaluation. Details of the models of generation and distribution level are not part of the state estimation in the high voltage transmission problems.

Volt-VAR Multiobjective Optimization to Peak-Load Relief and Energy Efficiency in Distribution Networks

This paper addresses the integrated volt-var control for distribution network operation via multiobjective optimization. This paper seeks to explore the problem of energy savings and peak demand relief through the voltage reduction procedure. Currently, due to the emergence of the distribution smart grids, these procedures are gaining renewed interest and attention. The proposal presented here is for the operation phase, with a strategy based on an hourly load forecast for the next day/week, taking into account the active power intake reduction, and the voltage deviation. Therefore, the result is a set of nondominated optimal solutions, and then one may decide when, where, and how to apply them to meet different goals. The obtained solutions, for two typical distribution networks, describe relevant economic and technical benefits.

Short-term load forecasting using fuzzy logic and ANFIS

This paper presents short-term load forecasting models, which are developed by using fuzzy logic and adaptive neuro-fuzzy inference system (ANFIS). Firstly, historical data are analyzed and weekdays are grouped according to their load characteristics. Then, historical load, temperature difference and season are selected as inputs. In general literature, fuzzy logic hourly load forecasts are tested in the range a few days or a few weeks. Unlike previous studies, the hourly load forecast is carried out for 1Â year. This paper shows that fuzzy logic can give good results in very large test data sets for 1Â year. Besides, for countries with large areas, the temperature data taken from only one point would lead to increase the forecasting errors. Therefore, the average of temperature for six cities having the maximum power consumption is weighted average. The mean absolute percentage errors of the fuzzy logic and ANFIS models in terms of prediction accuracy are obtained as 2.1 and 1.85, respectively. The results show that the proposed fuzzy logic and ANFIS models are capable of load forecasting efficiently and produce very close values to the actual data and are the alternative way for short-term load forecasting in Turkey.

Short-term hourly load forecasting of a hospital using an artificial neural network

Short-term hourly load forecasting of a hospital using an artificial neural network

The 24 Hourly Load Forecasting of the Election Day Using the Load Variation Rate

Short-term electric load forecasting of power systems is essential for the power system stability and the efficient power system operation. An accurate load forecasting scheme improves the power system security and saves some economic losses in power system operations. Due to scarcity of the historical same type of holiday load data, most big electric load forecasting errors occur on load forecasting for the holidays. The fuzzy linear regression model has showed good accuracy for the load forecasting of the holidays. However, it is not good enough to forecast the load of the election day. The concept of the load variation rate for the load forecasting of the election day is introduced. The proposed algorithm shows its good accuracy in that the average percentage error for the short-term 24 hourly loads forecasting of the election days is 2.27%. The accuracy of the proposed 24 hourly loads forecasting of the election days is compared with the fuzzy linear regression method. The proposed method gives much better forecasting accuracy with overall average error of 2.27%, which improved about average error of 2% as compared to the fuzzy linear regression method.

Short-term load forecasting of power systems by combination of wavelet transform and neuro-evolutionary algorithm

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.

An efficient linear model and optimisation algorithm for multi-site combined heat and power production

Combined heat and power (CHP) production is an important energy production technology which can help to improve the efficiency of energy production and to reduce the emission of CO 2. Cost-efficient operation of a CHP system can be planned using an optimisation model based on hourly load forecasts. A long-term planning model decomposes into hourly models, which can be formulated as linear programming (LP) problems. Such problems can be solved efficiently using the specialized Power Simplex algorithm. However, Power Simplex can only manage one heat and one power balance. Since heat cannot be transported over long distances, Power Simplex applies only for local CHP planning. In this paper we formulate the hourly multi-site CHP planning problem with multiple heat balances as an LP model with a special structure. We then develop the Extended Power Simplex (EPS) algorithm for solving such models efficiently. Even though the problem can be quite large as the number of demand sites increases, EPS demonstrates very good efficiency. In test runs with realistic models, EPS is from 29 to 85 times faster than an efficient sparse Simplex code using the product form of inverse (PFI). Furthermore, the relative efficiency of EPS improves as the problem size grows.

An efficient linear programming model and optimization algorithm for trigeneration

Trigeneration is a booming technology for efficient and clean provision of energy. It has potential for reducing pollution emissions dramatically. Similar to combined heat and power (CHP) production, cost-efficient operation of a trigeneration system can be planned using an optimization model based on hourly load forecasts. A long-term planning model decomposes into thousands of hourly models, which can be solved separately. In this paper, we model the hourly trigeneration problem as a linear programming (LP) model with a joint characteristic for three energy components to minimize simultaneously the production and purchase costs of three energy components, as well as CO 2 emissions costs. Then we explore the structure of the problem and propose the specialized Tri-Commodity Simplex (TCS) algorithm that employs this structure efficiently. The speed of TCS is based on extremely fast basis inverse operations and reuse of old basic solutions from previously solved hourly models. We compare the performance of TCS with realistic models against an efficient sparse Simplex code using the product form of inverse. In test runs, TCS is from 36 to 58 times faster when starting from the initial basis and from 43 to 179 times faster when reusing the old basis.

Hourly cooling load prediction by a combined forecasting model based on Analytic Hierarchy Process

Predicting the next-24-hour load in a building is essential for the optimal control of heating, ventilating and air-conditioning (HVAC) systems that use thermal/cool storage technology and also for cost and energy reduction of the non-storage systems. To fully integrate the advantages of several models and improve the accuracy of forecasting load, the application of the combined forecasting method to hourly load forecasting is presented in this paper. The method of Analytic Hierarchy Process (AHP) is employed to deduce the weights of each model. A case study shows that the combined forecasting model based on AHP may be better than the individual ones in predicting the building's hourly load for the future hours.

An efficient linear programming algorithm for combined heat and power production

Combined heat and power (CHP) production is an increasingly important energy production technology. CHP production is usually applied in back pressure plants, where the heat and power generation follows a joint characteristic. A CHP system may also comprise separate heat and power production facilities. Cost-efficient operation of a CHP system can be planned using an optimisation model based on hourly load forecasts. A long-term optimisation model decomposes into thousands of hourly models, which can be formulated as linear programming (LP) problems. We model the hourly CHP operation as an LP problem with a special structure and present the specialised Power Simplex algorithm that utilises this structure efficiently. The basis can be organised as an identity matrix and a small block of non-zero coefficients. There are only a few different types of non-zero blocks, and extremely fast inversion procedures have been designed for each type. The performance of Power Simplex is compared with realistic models against a non-sparse tabular Simplex algorithm and the LP2 software based on the sparse Revised Simplex algorithm using the product form of inverse. At its best, Power Simplex performs from 21 to 190 times faster than the tabular Simplex. Power Simplex has been implemented as part of the EHTO NEXUS energy optimisation system, which is in commercial use at several Finnish energy companies.

Short Term Energy Forecasting with Neural Networks

Artificial neural networks are beginning to be used by electric utilities, to forecast hourly system loads on a day ahead basis. This paper discusses the neural network specification in terms of conventional econometric language, providing parallel concepts for terms such as training, learning, and nodes in the, hidden layer. It is shown that these models are flexible nonlinear equations that can be estimated using nonlinear least squares. It is argued that these models are especially well suited to hourly load forecasting, reflecting the presence of important nonlinearities and variable interactions. The paper proceeds to show how conventional statistics, such as the BIC and MAPE statistics can be used to select the number of nodes in the hidden layer. It is concluded that these models provide a powerful, robust and sensible approach to hourly load forecasting that will provide modest improvements in forecast accuracy relative to well-specified regression models.

A temperature match based optimization method for daily load prediction considering DLC effect

This paper presents a unique optimization method for short-term load forecasting. The new method is based on the optimal template temperature match between the future and past temperatures. The optimal error reduction technique is a new concept introduced in this paper. Two case studies show that for hourly load forecasting, this method can yield results as good as the rather complicated Box-Jenkins transfer function method, and better than the Box-Jenkins method. For peak load prediction, this method is comparable in accuracy to the neural network method with backpropagation, and can produce more accurate results than the multilinear regression method. The direct load control (DLC) effect on system load is also considered in this method.