Electric power system short-term load prediction method based on hybrid kernel function adaptive fusion

Abstract

The invention belongs to the technical field of power systems. The invention discloses an electric power system short-term load prediction method based on hybrid kernel function adaptive fusion. Kernel functions are selected from the local kernel function library and the global kernel function library respectively, weight variables are allocated to form a mixed kernel function, the weight variables and parameters of the kernel functions are put together to be combined into a new parameter state vector, and a nonlinear parameter estimation model is established; and then estimating a parameter state by using high-order cubature Kalman filtering based on a parameter estimation model, so that a local kernel function and a global kernel function can be adaptively fused, and a trained neural network is used to predict a power load. According to the power system short-term load prediction method based on hybrid kernel function adaptive fusion, the optimal neural network hybrid kernel functionfusion coefficient is selected, and the load prediction precision is improved.

Description

technical field [0001] The invention belongs to the technical field of power systems, and in particular relates to a short-term load forecasting method of power systems based on hybrid kernel function self-adaptive fusion. Background technique [0002] At present, the existing technologies commonly used in the industry are as follows: [0003] Accurate forecasting of short-term power load is the premise of safe and economical operation of the power system. It can effectively reduce power generation costs and improve efficiency. It is also the basis for scheduling, power supply and transaction planning. The change of power load is affected by many factors such as natural weather, social economy, etc., and its change is dynamic, random, and complex. Therefore, high-precision short-term load forecasting has become an important content of power system research. [0004] Traditional load forecasting methods mainly include regression analysis and time series methods, etc. These m...

AI Technical Summary

Problems solved by technology

[0008] (1) When the existing neural network predicts the short-term load of the power system, there is a problem that the weight of the mixed kernel function cannot be adaptively assigned according to the sample characteristics

[0009] (2) Existing state estimation algorithms such as extended Kalman filter, unscented Kalman filter and volumetric Kalman filter are still not high enough to estimate the state

[0010] (3) Most of the existing methods calculate the weights from the perspective of neural network learning and fitting, and the accuracy of the weights obtained in this way is not high

[0012] Aiming at the problem of low short-term forecasting accuracy of electric power, existing algorithms cannot accurately estimate network connection weights by using neural networks;

[0013] The present invention has important significance to calculate the weight from the perspective of filter estimation, and uses the weight as a variable of state estimation. However, how to design a high-precision high-order volumetric Kalman filter algorithm to achieve adaptive calculation of the weight is a challenging problem. question

Images