Short-term power load prediction method based on dimensionality reduction and improved neural network

Abstract

The invention discloses a short-term power load prediction method based on dimensionality reduction and an improved neural network, and relates to the technical field of power load prediction, and themethod comprises the steps: firstly enabling high-dimensional load related meteorological data to be mapped to a low dimension through affine transformation through employing an SNE algorithm; and then taking the output weight and threshold of the ELM as optimization variables through a moth flame optimization algorithm, taking the mean square error value of the load prediction result as an optimization result, finding out and feeding the training weight and threshold corresponding to the optimal prediction result back to the ELM, and obtaining the improved neural network prediction model; and jointly inputting the dimension-reduced meteorological data and the power load data into an improved neural network for data training and load prediction. According to the invention, the short-termpower load prediction process of the neural network is improved, and the precision of power load prediction is greatly improved.

Description

technical field [0001] The invention relates to the technical field of power load forecasting, in particular to a short-term power load forecasting method based on dimension reduction and improved neural network. Background technique [0002] Load forecasting is of great significance to the reliable and economical operation of power systems. In all aspects of the energy sector, load forecasting is an indispensable tool and a prerequisite for ensuring reliable power supply and safe operation of power systems. With the arrival of big data in the power grid, the interference of multi-dimensional meteorological factors will affect the accuracy of load forecasting. The continuous growth of data dimensions has led to many problems such as the curse of dimensionality. How to represent high-dimensional data in low-dimensional spaces has become more and more important. SNE can better map the data in high-dimensional space to low-dimensional space and maintain the data structure of h...

AI Technical Summary

Problems solved by technology

The traditional prediction network BP neural network backpropagates errors, the training speed is slow and the prediction accuracy is not high, and the ELM prediction speed is fast but the stability is not good, which will lead to a substantial increase in the operating cost of electricity

Images