Spatial load prediction method based on clustering and nonlinear autoregression

Abstract

The invention discloses a spatial load prediction method based on clustering and nonlinear autoregression. The spatial load prediction method comprises the steps of collecting power distribution areadata; dividing a to-be-predicted region into square units with the same area; standardizing the historical load curve of each unit, generating a plurality of clusters according to the similarity clustering units of the historical load curves, and independently forming a cluster by the outlier units; taking the load curves of a plurality of units in one cluster as training data, and training an improved nonlinear autoregressive neural network; independently executing a trend extrapolation method for each unit in the outlier unit cluster to predict the load of the next year; and selecting a prediction model of a cluster where each unit is located, and inputting historical data of the prediction model to obtain load prediction of the next year. According to the method, a large number of unitsshare one model, sufficient training data are provided for the model, through an improved nonlinear autoregressive neural network algorithm, the model is made to well fit the plot power utilization development form, and the recent space load prediction precision is effectively improved.

Description

technical field [0001] The invention relates to a power system planning and space load forecasting technology, in particular to a space load forecasting method based on clustering and nonlinear auto-regression. Background technique [0002] Space load forecasting is divided into short-term forecasting and long-term forecasting. The short-term space load forecast can provide the project scheduling basis for the planning and implementation department. Traditionally, trend extrapolation is used for near-term forecasts for feeder areas. However, due to the development of the region and the transformation and construction of the power sector, the traditional grid and cell division often change, and the coverage of the branch line will also change, causing the meter of the branch line to be inaccurately corresponding to the area. At the same time, the trend extrapolation method often produces large errors due to the limited amount of data. [0003] Recently, spatial load foreca...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the deficiencies in the prior art, such as large errors in power system planning and spatial load forecasting, and insufficient analysis, the problem to be solved by the present invention is to provide a clustering and nonlinear autoregressive algorithm based on clustering and nonlinear autoregressive that can effectively improve the accuracy of recent spatial load forecasting. Space Load Forecasting Method

Images