Client-side multi-energy system source load prediction method and device based on feature engineering

Abstract

The invention discloses a client-side multi-energy system source load prediction method and device based on feature engineering. The method comprises the steps of preprocessing original observation data of a client-side multi-energy system; performing candidate feature extraction based on the preprocessed data; screening the candidate features, and extracting an optimal feature subset; and performing source load prediction based on the extracted optimal feature subset. According to the invention, domain knowledge and statistical knowledge are combined, and relatively systematic and complete feature engineering is carried out on the data, so that the feature set for model construction can more completely and accurately express information implied in the original data. By considering the characteristics of a prediction target and a feature item, a proper algorithm is selected to construct a prediction model, so that the prediction precision and speed are better guaranteed.

Description

technical field [0001] The invention belongs to the technical field of client-side source-load prediction, and in particular relates to a method and device for client-side multi-energy system source-load prediction based on feature engineering. Background technique [0002] In recent years, with the increasing shortage of energy supply and the advancement of policies related to energy conservation and emission reduction, the proportion of renewable energy such as wind power and photovoltaics in the customer-side power structure has continued to increase. Through accurate source-load forecasting, it can provide support for the economic dispatch of multi-energy systems on the client side, coordinate and controllable resources, and realize mutual benefit and complementarity of various energy sources. However, since the source-load data of the client-side multi-energy system is affected by weather, environment, energy consumption behavior, etc., it has strong randomness and vola...

AI Technical Summary

Problems solved by technology

Although the data is simply processed, considering the challenge and complexity of feature engineering, it is obvious that these operations cannot guarantee the quality of features

With the integration of distributed energy such as photovoltaics and wind power, energy storage equipment, and various types of loads, the characteristics of source-load data of multi-energy systems on the client side are becoming more and more complex. Simple processing of data will obviously lead to larger predictions error

[0005] In general, there are two main problems in the research on source-load forecasting of customer-side multi-energy systems: on the one hand, there is a lack of general modeling methods, and each independent forecasting target needs to be analyzed and modeled one by one; On the one hand, less attention has been paid to the link of feature extraction, which is not conducive to the improvement of the accuracy of source and load prediction in the field of comprehensive energy

Images