Kernel extreme learning machine electricity sales prediction method based on generalized maximum correlation entropy criterion

Abstract

The method comprises the following steps: correcting abnormal data of historical daily electricity consumption, constructing a training sample set, selecting model input by using a Pearson correlationcoefficient, selecting a kernel extreme learning machine model to predict daily electricity consumption, and predicting the daily electricity consumption by using a kernel extreme learning machine model. Aiming at the non-Gaussian characteristic of the electricity selling quantity prediction error; a generalized maximum correlation entropy criterion is used as a cost function of a prediction model, online sequence learning is introduced to enable the model to perform rolling prediction, and K-break cross validation and grid optimization are introduced to optimize key parameters sigma, lambdaand alpha of a generalized maximum correlation entropy kernel extreme learning machine model. And predicting the electricity selling quantity by using the generalized maximum correlation entropy kernel extreme learning machine prediction model to obtain a prediction result. Compared with an existing method, the electricity selling quantity prediction method has good performance under the conditionof large outlier and non-Gaussian, non-Gaussian nonlinear data can be better predicted, and the prediction effect is better.

Description

technical field [0001] The invention belongs to the technical field of power system electricity forecasting, and relates to a nuclear extreme learning machine electricity sales forecasting method based on the generalized maximum correlation entropy criterion. [0002] technical background [0003] Power forecasting is a forecast of power consumption for a period of time in the future on the premise of analyzing historical data and influencing factors. In the current power market, whether it is the power generation side or the power sales side, power forecasting is a very important task. Especially for electricity sales companies, due to the promulgation of the deviation assessment mechanism, the accuracy of electricity sales forecast directly affects the evaluation of deviations, and the greater the deviation, the higher the penalty. However, it is difficult to accurately predict electricity sales due to the large impact of random factors. Therefore, the research on high-pre...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to propose a nuclear extreme learning machine electricity sales forecast method based on the generalized maximum correlation entropy criterion, which solves the problem that the accuracy of the electricity sales forecast in the prior art is not high, and it is difficult to meet the requirements of the electricity sales transaction for the forecast accuracy of the electricity sales

Images