An Intelligent Prediction Method of Electricity Sales Amount Based on Deep Recurrent Neural Network

Abstract

The invention relates to big data processing, and aims to provide an intelligent prediction method for electricity sales amount based on a deep recurrent neural network. Including: reading the historical data of sales flow and electricity consumption in the power sector, conducting information mining and analysis after preprocessing, evaluating the relationship between the time when the amount arrives and the time when the user pays, and obtaining distribution information; organizing the historical data structure, extracting The normalized n-day data is used as input, and the multi-layer recurrent neural network (GRU) is used to learn high-dimensional features, and the high-dimensional features are input into the softmax discriminator to classify the sales amount in a certain period in the future; The hyperparameters of the network model are traversed, the best hyperparameters are recorded after multiple experiments, and the final deep cycle neural network model for amount prediction is constructed, and it is used to intelligently predict the amount of electricity sales. The present invention is more accurate and reasonable, requires less manual intervention, has more robust results, is more adaptable to big data, and can learn automatically.

Description

technical field [0001] The invention relates to big data processing, in particular to an intelligent prediction method for electricity sales amount based on a deep cyclic neural network. Background technique [0002] Sales forecasting refers to the sales forecasting model obtained through mathematical modeling based on the past sales situation and the analysis of the future form, and on the basis of fully considering various influencing factors, so as to realize the sales of all products or specific products within a certain period of time in the future. Estimates of quantities and sales amounts. Sales forecast is very important for the enterprise's development planning, strategic deployment, production management, import, export and effective control of each link of the supply chain. There are many factors that affect the sales forecast, including market demand, development status of related enterprises, policy changes and seasonal changes, etc. Among these many factors, ...

AI Technical Summary

Problems solved by technology

In fact, it is difficult to apply deep recurrent neural network technology to the daily sales forecast of the power system. The main reason is that the historical data used for forecasting only have a single feature of sales amount and sales date, and the feature dimension is too large. Less makes the model overfitting The data distribution seen cannot predict the unknown data well

Images