Research On Short-term Load Forecasting Of Power System Based On ELM

Short-term Load Forecastin(STLF)plays a major role in my country’s power energy economy by coordinating market demand and stabilizing people’s livelihood development.In recent years,with the development of industrial diversification,the load has shown the characteristics of various types and complex structures.The development characteristics of load pose a severe challenge to STLF,and designing a prediction scheme with high prediction accuracy and strong generalization ability has become an important research topic in this field.Based on the Extreme Learning Machine(ELM)model,this paper studies the power load conditions in different regions with strong and weak meteorological sensitive loads according to the characteristics of the temperature in the electricity consumption area,and deeply mines the influence of meteorological information on the load change trend.Influence,main work and innovation points are as follows:(1)In view of the insufficient consideration of temperature and humidity in traditional methods in areas with strong meteorological sensitivity,a short-term power load forecasting method based on temperature form strategy and whale algorithm is proposed.Firstly,the correlation analysis of meteorological factors is carried out by using Maximum Information Correlation(MIC);Then,based on the highly correlated meteorological factors,three different forms of variables are constructed as the input characteristics of the model;Then,the traditional whale algorithm is improved by introducing adaptive weight,Variable Helix strategy and Cauchy inverse accumulation distribution function;Finally,the Improved Whale Optimization Algorithm(IWOA)is used to determine the number of neurons in the ELM and output the final result.The load data provided by New South Wales,Australia were used for simulation,and the experimental simulation showed that the proposed scheme has higher prediction accuracy than the comparison model.(2)For the purpose of the problems of low prediction accuracy and weak generalization ability caused by insufficient reflection of the characteristics of a single model,a short-term power load forecasting model based on similar day selection and multi integration combination is proposed.Firstly,the maximum information criterion is used to select high correlation variables as the model input;Then,considering the mic and grey correlation degree,the similar days of load in areas with weak meteorological sensitivity are selected;Finally,a framework combination multi-core kernel ELM integrating random subspace,adaptive enhancement and stack neural network is introduced to perform regression fitting on the predicted day and output the final prediction result.The load data provided by Johor power supply company in Malaysia were used for example analysis.Similarly,the experimental simulation showed that the proposed scheme has high prediction accuracy and strong generalization ability.(3)In order to facilitate power scientific research institutions,electricity sales companies and other relevant units to formulate plans,analyze needs,and build a forecast data visualization interface.First,the forecast data obtained by the load forecasting algorithm is stored in the Mongo DB database,and the database is placed on the server side.The server side adopts Express framework to monitor database requests;Nginx technology is used to provide highperformance HTTP and reverse proxy services.On the browser side,the Vue framework is used as the overall architecture,and the Echarts data analysis visualization tool and the Element UI plug-in are used for rendering and display.The browser side includes functions such as data display,data validation and corresponding image saving,which is convenient for research and development and learning by the user.