Short-term power load prediction method based on multilayer improved GRU neural network

Abstract

The invention discloses a short-term power load prediction method based on a multilayer improved GRU neural network. The method comprises the following steps: periodically establishing a sample data set; performing abnormal point identification and missing value processing on input data in the samples, performing standardized transformation on the processed data, and dividing the data set into a training set, a verification set and a to-be-predicted set; constructing the improved GRU neural network, and inputting the data of the training set in the network for training the same for multiple times to obtain a trained network, verifying the test learning result of the verification set by using the network for multiple times, and recording and storing the model weight of the optimal verification result; and inputting the to-be-predicted set in the trained optimal GRU model, calculating a standardized prediction result and performing reverse standardized transformation to obtain a final prediction result. By adoption of the short-term power load prediction method, the training speed and the training efficiency are improved.

Description

technical field [0001] The invention relates to a short-term power load forecasting method based on a multi-layer improved GRU neural network. Background technique [0002] Power load forecasting refers to the forecasting of electricity demand within a certain period of time in the future. As an important work in the power sector, accurate load forecasting can promote dispatch and power supply companies to economically and rationally arrange power generation plans and unit maintenance plans for internal generator sets in the power grid, maintain the safety and stability of power grid operation, and ensure the normal production and life of society. In this case, the short-term power load forecast is based on the intraday load data at intervals of 30 minutes or 1 hour as the forecast object. For the power sector, the accuracy of short-term load forecasting directly affects the dispatching of the next day's power generation plan, which is conducive to the stable operation of t...

AI Technical Summary

Problems solved by technology

However, support vector machines are difficult to handle large-scale training samples; wavelet transform algorithms usually need to be combined with artificial neural networks; fuzzy systems do not have self-learning ability, and the establishment of fuzzy rules relies more on expert experience

Images