Research On Multi-Scale Refined Electricity Forecast And Its Influencing Factors

The medium and long-term power consumption can be divided into monthly power consumption,quarterly power consumption and annual power consumption according to different time scales.The annual power consumption time series is often used as an important economic indicator to participate in the research and analysis of econometric problems;Due to its complex characteristics,the monthly power consumption time series makes it difficult for a single prediction method to achieve the desired prediction accuracy.This article explores the relevant factors that affect the growth of electricity consumption to guide the forecasting of monthly electricity consumption and improve the accuracy of monthly electricity consumption forecasting.The main research contents are as follows:First,in this paper,the Apriori algorithm is used to mine the economic and social development factors that affect the growth of electricity consumption.In order to apply the Apriori algorithm to the growth rate of electricity consumption,the membership function in fuzzy mathematics theory is used to preprocess the data.Aiming at the problem that the critical value is too dependent on subjective experience when establishing the membership function,the optimal partition algorithm is used to select the critical value of the membership function.Combining fuzzy mathematics theory with Apriori algorithm to realize the mining of economic and social development factors that affect the growth of electricity consumption,and find the main factors that affect the growth of electricity consumption.Second,in the large scope of the time series method,this paper proposes a refined power prediction method based on multi-scale decomposition.The X12 seasonal decomposition method is used to decompose the monthly power consumption time series into multi-scale decomposition,and the originally complex time series is decomposed into trend cycle components,seasonal components and irregular components.After finding out the economic and social development factors that affect the growth of electricity consumption,select stepwise regression method to screen these influencing factors again.And then selecting the influencing factors that have a higher contribution to the forecast of electricity consumption as independent variables to establish regression equations int order to realize the prediction of trend components.Aiming at the problem that the monthly electricity consumption in the first quarter of each month is affected by the Spring Festival,combined with the irregular components decomposed by X12,a model is established to predict the monthly values of the irregular components in the first quarter.Using the previous value as the quarterly component prediction value,the prediction results of the three components are combined to obtain the final prediction result.Compared with the single prediction model,the prediction accuracy is improved,and the disturbance of the Spring Festival effect on the monthly electricity consumption forecast is solved.identified the influencing factors that have a higher contribution to the electricity consumption forecast.Third this article introduces artificial intelligence algorithms to explore the application in monthly electricity consumption forecast.Empirical mode decomposition algorithm is used to eliminate the influence of the Spring Festival effect on the training neural network.In order to solve the problem of unstable training results of neural networks under poor sample conditions,this paper uses the ability of genetic algorithms to search for optimal solutions,combining genetic algorithms with neural networks to improve the stability of neural networks.In order to solve the problem that genetic algorithms are easy to come into premature,this paper introduces two populations for genetic iteration,one population searches for global optimal solutions,and one population searches for local optimal solutions to improve the genetic algorithm’s ability to search for optimal solutions.Finally,due to the strong anti-interference ability of the quarterly power consumption,the forecast results are easy to obtain.In this paper,when the forecast accuracy of quarterly electricity consumption is higher than the forecast accuracy of monthly electricity consumption,the adjustment model of monthly electricity consumption forecast value based on the forecast value of quarterly electricity consumption is proposed.After obtaining the quarterly electricity consumption forecast value and the monthly electricity consumption forecast value,an equation constraint model for solving the adjustment amount of the monthly electricity consumption forecast value is established to realize the correction of the monthly electricity consumption forecast value and improve the forecast accuracy.