Neural Network Electricity Price Forecasting Based On Grey Data Preprocessing

With the reform of domestic electric power system,the monopoly barriers of electricity market have been broken.Electricity selling companies as the representative of the main body of electricity sales mushroomed,electricity price as the core competitive factor has become the competition orientation of all parties in the market.How to accurately forecast electricity price has become another important research direction.In this paper,the grey model is used to deal with the historical electricity price and load value to generate the fitting value.Then the fitting value is used as the input value of the neural network,and the actual electricity price is used as the target value to train the neural network model.When the training times reach the maximum value or the precision reaches the set interval,the training is stopped,and the weights and thresholds at this time are determined as the optimal solution,The final model parameters are obtained.The gray model is good at getting the intrinsic correlation of historical data by accumulation,and deeply mining the endogenous law of historical data.By using the gray model to process historical electricity price data and load data,we can effectively mine the effect of historical electricity price and load on electricity price,which is the main criterion of the regular factors in the model.In this paper,the neural network model is used as the core mathematical model to simulate and express the irregular factors that affect the electricity price.Neural network model is a non-linear model in essence,which has strong mapping and imitation ability for input and output data.In this paper,neural network model is used to train and predict the input gray preprocessing data,and the advantages of gray model are used to optimize the prediction results of neural network model,while avoiding the shortcomings of gray model.In the process of building prediction model,this paper also focuses on the inherent defects of grey model and neural network model,and makes improvement.For the grey model,single or several fluctuating data will seriously affect the development coefficient and grey action of the grey model,interfere with the accuracy of the model.The grey model is good at predicting smooth data.In this paper,the equal dimensional recurrence,moving average and inverse hyperbolic cosine change method are used to improve the GM(1,2)model.In view of the inflexible way to determine the weights and thresholds of neural network model,it is easy to make the training process of neural network model converge to local maximum and minimum.In this paper,genetic algorithm is used to determine the weights and thresholds of neural network model and determine the approximate optimal space,Then,the optimal weights and thresholds are obtained by a large number of neural network training.