Power load prediction method, device and apparatus based on echo state network

Abstract

The invention discloses a power load prediction method based on an echo state network. The method comprises the following steps: dividing a power load time sequence data set into a time sequence training data set and a time sequence test data set, inputting the time sequence training data set into a plurality of echo state networks with different dependency relationships, and splicing obtained storage pool states of different time scales to obtain a spliced storage pool state matrix; training the spliced reserve pool state matrix by using ridge regression to obtain an output layer weight; andinputting the time sequence test data set into each echo state network, and predicting the power load value of the time sequence data set at the next preset moment in combination with the weight of the output layer. The robustness of the model used in the prediction process is improved, and the prediction precision of the power load value is greatly improved. The invention also discloses a power load prediction device and equipment based on the echo state network, and a storage medium, which have corresponding technical effects.

Description

technical field [0001] The present invention relates to the technical field of power load forecasting, in particular to a power load forecasting method, device, equipment and computer-readable storage medium based on an echo state network. Background technique [0002] Time series refers to a group of statistical data arranged in chronological order for the same phenomenon observed or recorded. Through the analysis of time series, according to the development trend reflected in the time series, the level reached in the next stage or several stages can be predicted. This technique is called time series forecasting technique. [0003] In recent years, time series forecasting technology based on neural network has shown a rapid development momentum, and each new neural network method has promoted the enrichment and improvement of forecasting technology. Among them, in electric load forecasting, multi-layer forward network and recursive neural network are the two most basic for...

AI Technical Summary

Problems solved by technology

These prediction methods are all based on the space of the original sequence to capture the relationship between each sequence, and cannot model the dynamics of the time series, and the accuracy of the prediction of the power load value is low.

Images