Search

Background: The traditional state-based non-intrusive load monitoring method mainly deploys the aggregate load as the characteristic to identify the states of every electrical appliance. Each identification is relatively independent, and there is no correlation between the identification results. Objective: This paper combines the event detection results with the state-based non-intrusive load identification algorithm to improve accuracy. Methods: Firstly, the load recognition model based on an artificial neural network is constructed, and the state-based recognition results are obtained. An event recognition and detection model is then built to identify electrical state transitions, that is, the current moment based on the event recognition results obtained from the previous moment. Finally, a reasonable decision method is constructed to determine the identification result of the electrical states. Result: Experimental results on the public data set REDD show that in the Long Short-Term Memory (LSTM) fusion model, the macro-F1 is increased by an average of 6%, and the macro-F1 of the Artificial Neural Network (ANN) fusion model is increased by an average of 5.3% compared with LSTM and ANN. Conclusion: The proposed model can effectively improve the accuracy of identification compared with the state-based load identification method.

Background : The day-ahead load forecasting is an essential guideline for power generating, and it is of considerable significance in power dispatch. Objective: Most of the existing load probability prediction methods use historical data to predict a single area, and rarely use the correlation of load time and space to improve the accuracy of load prediction. Methods: This paper presents a method for day-ahead load probability prediction based on spacetime correction. Firstly, the kernel density estimation (KDE) is employed to model the prediction error of the long short-term memory (LSTM) model, and the residual distribution is obtained. The correlation value is then used to modify the time and space dimensions of the test set's partial period prediction values. Results: The experiment selected three years of load data in 10 areas of a city in northern China. The MAPE of the two modified models on their respective test sets can be reduced by an average of 10.2% and 6.1% compared to previous results. The interval coverage of the probability prediction can be increased by an average of 4.2% and 1.8% than before. Conclusion: The test results show that the proposed correction schemes are feasible.

Background: Electricity price forecasting is still a challenging issue as it plays an essential role in balancing electricity generation and consumption. Probabilistic electricity price forecasting not only provides deterministic price forecasts but also effectively quantifies the uncertainty of electricity price. Methods: This paper introduces a new short-term electricity price forecasting approach called SASVQR, which is based on support vector quantile regression (SVQR) optimized by simulated annealing algorithm. In this study, SVQR is employed to obtain the conditional quantiles of the electricity under different quantile points, while the simulated annealing algorithm is applied to optimize each SVR model. Then the kernel density estimation takes these conditional quantiles as inputs and generates the probability density functions for future electricity prices. Result: The proposed algorithm is assessed in three datasets: the GEFCom 2014, two real electricity price datasets from the PJM market and the Singapore market. Three popular probabilistic forecasting criteria, namely prediction interval coverage probability (PICP), prediction interval normalized average width (PINAW), and coverage width-based criterion (CWC), are utilized to evaluate the numerical experiment results..It shows the promising forecasting performance, robustness, and effectiveness of SASVQR on different datasets. Conclusion: The SASVQR method can effectively forecast the short-term electricity price compared with other methods.

Evolutionary computations are kinds of random searching algorithms derived from natural selection and biological genetic evolution behavior. Evaluating the performance of an algorithm is a fundamental task to track and find the way to improve the algorithm, while visualization technique may play an important act during the process. Based on current existing algorithm performance evaluation criteria and methods, a Python-based programming tracking strategy, which employs 2-D graphical library of python matplotlib for online algorithm performance evaluation, is proposed in this paper. Tracking and displaying the performance of genetic algorithm (GA) and particle swarm optimization (PSO) optimizing two typical numerical benchmark problems are employed for verification and validation. Results show that the tracking strategy based on Python language for online performance evaluation of evolutionary algorithms is valid, and can be used to help researchers on algorithms’ performance evaluation and finding ways to improve it.

In order to improve the predict precision of GM(1,1) when it is applied to the short term load forecasting (STLF) problem, the meteorological information is taken into consideration. Firstly, an improved multi-strategy is used to organize the origin load data. Secondly, this chapter proposes meteorological analyzing and correcting algorithm to recognize the weather sensitive data and amend them. Then GM(1,1) is taken as the basic method to do the prediction. Finally, nearby trend extrapolation amending and similar-day replacing method is proposed to adjust the result and clear the mutation in it. Through the test, it is found that such method has a far more better precision than origin GM(1,1) when there is weather mutation in history days or predict days. The highest variety of accuracy can be up to 7 % and there is an average increase in predicting accuracy by almost 2 %. It can be concluded that such methods can not only take care of the social and climate affect but also considers the weakness of GM itself.

Power load forecasting has always been a hotspot. Recently, Artificial intelligence and computational intelligence methods have been widely used in the power load forecasting field. SVR (Support Vector Regression), one of computational intelligence methods, has been paid more and more attention for its ability of solving none-liner problem and its high prediction accuracy. Most predicting methods based on SVR prefer researching how to optimize argument of SVR model. for the aim of downsizing the training sample or improve the accuracy, some literatures proposed to get optimal subset from the whole training set or reduce attributes of each sample by using mathematical models. but the result of attribute auto reduction can't intuitive show the relationship between various attributes. Moreover it is difficult to deal with the relation between many attributions which may lead to retain or abandon the attributes improperly. This paper proposed a method to construct training set by not only analyzing the relation between the load data and attributes such as weather factor, but also analyzing the load data self-similarity. the result of load forecasting experiment adopting our method shows that the accuracy of short-term load forecasting can be improved effectively.