Short-term bus load prediction method based on RBF hidden layer parameter optimization

Abstract

The invention discloses a short-term bus load prediction method based on RBF hidden layer parameter optimization. The method comprises the following steps: training a short-term bus load prediction model based on an RBF neural network by using a training sample; and inputting an input quantity into the trained short-term bus load prediction model to obtain a short-term bus load. According to the method, the center vector of the hidden layer node of the RBF neural network is selected based on the affinity propagation method, the overfitting problem caused by too many hidden layer neurons is avoided, and the problem of low prediction precision caused by too few hidden layer neurons is also avoided. According to the method, the minimum sum of squares of absolute errors of network fitting is taken as a target function, and the base width parameter of the hidden layer of the RBF neural network is optimized based on the genetic algorithm, so that the rationality and the adaptivity of prediction model parameters are ensured.

Description

technical field [0001] The invention relates to a short-term bus load forecasting method based on RBF hidden layer parameter optimization, and belongs to the technical field of power system load forecasting. Background technique [0002] The load is seasonal, regional, industrial, policy-based, random and adjustable. The study of power load forecasting is the basis for the development planning and power generation plan of the power system. High-precision load forecasting is conducive to reducing energy consumption, balancing power generation investment, and arranging power generation plans, thereby reducing costs and improving power supply reliability. With the further promotion of electricity marketization, the spot market has put forward higher requirements for short-term bus load forecasting. [0003] Traditional forecasting mostly uses methods such as trend extrapolation and gray forecasting GM(1,1), only considering the development trend of the sequence itself; later, ...

AI Technical Summary

Problems solved by technology

[0003] Traditional forecasting mostly uses methods such as trend extrapolation and gray forecasting GM(1,1), only considering the development trend of the sequence itself; later, it gradually develops into methods such as multivariate gray forecasting and multiple linear regression that consider external factors; The further deepening of refined management, the development of information collection, data mining and other technologies provide a more complete data basis for load forecasting, and provide an application platform for artificial intelligence methods that consider multiple influencing factors. , Artificial intelligence algorithms have a large amount of calculations, and a series of over-fitting problems need to be solved urgently

Images