Short-term load forecasting system and method for cooling, heating and power based on granular support vector machine

Abstract

The invention discloses a system and method for short-term load prediction of cooling, heating and electricity based on granularity support vector machine. For the sample data, the shared nearest neighbor similarity is used to divide the sample into several information granules, and the k nearest neighbor connectivity is used to extract the key support vector points of each information granule, and the final decision model is obtained after the support vector machine training is performed on the support vector points. Then, the decision model is used to forecast the cooling, heating and power load to improve the short-term load forecasting accuracy and reduce the time complexity. The method of the invention reduces the complexity of the load training samples while ensuring that the key granulation information is not lost, overcomes the randomness of the traditional granularity support vector machine for the granulation of the short-term load historical sample data, so as to obtain better prediction accuracy and model generalization. transformation ability.

Description

technical field [0001] The invention relates to the technical field of comprehensive energy system load forecasting, in particular to a short-term load forecasting system and method for cooling, heating and electricity based on granular support vector machines. Background technique [0002] Due to the influence of factors such as building envelope elements, outdoor weather, user behavior and holidays, the short-term load demand of cooling, heating and power in the integrated energy system is fluctuating and random. The error of short-term load forecasting of cooling and heating power will bring many problems to the safe, reliable and stable operation and dispatching management of power grid and heating network. The commonly used short-term load forecasting methods for cooling, heating and power mainly include the time series method and the neural network method. The principles and models of the time series method are relatively simple, and it is difficult to meet the current...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to propose a short-term load forecasting system and method for cooling, heating and electricity based on granular support vector machines, aiming at avoiding the problems of over-fitting and under-fitting in conventional machine learning, while ensuring the key The granulation information is not lost, the complexity of the load training samples is reduced, and at the same time, the randomness of the granulation of the short-term load history sample data by the traditional granular support vector machine is overcome, so as to obtain better prediction accuracy and model generalization ability

Images