Research On Load Forecast Based On The Trend Of Three Peaks And Three Valleys

With the rapid development of social economy and the advancement of science and technology,the emergence of a large number of sophisticated instruments and the growing needs of the people,put forward higher requirements for the reliability of power load.If the power load is too high,there will be a situation of oversupply,which will waste a lot of energy and not meet the strategy of sustainable development.If the power load is too small,the situation of less than demand will not meet the daily production and life of the people,which is not conducive to The rapid development of social economy,in order to rationally and fully utilize electric energy,greatly improve the utilization rate of electric energy,so that electric energy can effectively reduce the waste of electric energy under the premise of meeting economic development,so that social resources can be fully utilized,and electric load must be accurately predicted.Firstly,the definition and classification of electric load are introduced.It is generally divided into motor power load,electric arc furnace load,power electronic rectification load,electrolysis load,residential lighting load and system loss load.The definition and steps of electric load forecasting are introduced.The requirements of the dispatching department analyzed four conditions and seven implementation steps that the power load forecasting should meet;introduced the characteristics of the electric load,including randomness,time,regionality and diversity;introduced the level of electric load,according to The requirements for power supply reliability and the degree of impact on the socio-economic,residential property,and political situation after the failure of power supply are divided into primary load,secondary load and tertiary load.Secondly,the prediction methods of electric load are introduced,including traditional prediction methods and modern prediction methods.The traditional prediction methods are based on existing mathematical tools,including exponential smoothing method,grey theory,regression model,time series and load derivation method.The principles and characteristics of various prediction methods are introduced respectively.Modern prediction methods are based on the emergence of artificial intelligence,which generally includes similar Japanese methods,neural networks,expert systems,fuzzy theory and combined prediction.Models,etc.,introduce the principles and characteristics of various prediction methods.Thirdly,the causes,treatment process and processing methods of load dead pixels are analyzed,and the principle and improvement method of linear interpolation method are given.The trend of power load with time is analyzed,including daily change trend,weekly change trend,monthly change trend and quarterly change trend.The power load in this region shows the phenomenon of “three peaks and three valleys”,which is characterized by the morning peak time concentrated at 10:30.The afternoon peak time is concentrated at 15 o’clock,the evening peak time is concentrated at 21:30;the afternoon trough time is concentrated at 12 o’clock,the late trough time is concentrated at 18 o’clock,and the night trough time is concentrated at 4 o’clock;the electric load is analyzed with climate change.The trend mainly includes the trend with temperature,the trend with humidity and the trend with the weather.The electric load in this region is positively correlated with temperature and has a strong correlation,that is,it increases with the increase of temperature and decreases with the decrease of temperature.The electrical load and humidity in this area are negatively correlated but the correlation is small,decreasing with the increase of humidity,and increasing with the decrease of humidity;the electric load in this area is affected by the change of weather type,but generally affects in summer and winter.More obvious,the impact is less in spring and autumn.Finally,the electric load period based on the trend of three peaks and three valleys was divided.Study the manifestation and characteristics of the accumulated temperature effect and establish a temperature correction model considering the accumulated temperature effect.The electric load multivariate regression method(MIR)based on temperature and humidity factor,the electric load forecasting method based on trend change(LD),the BP neural network electric load forecasting method based on temperature and humidity factor(BP)and the temperature and humidity factor based Algorithm Implementation and Prediction Results of Four Methods for Support Vector Machine Power Load Forecasting Method(SVM).Considering the power load in different stages of three peaks and three valleys,showing different characteristics of change,Based on the trend of combined forecasting,the three peaks and three valleys adopt the combined forecasting method and solve the optimal combination parameters based on the least squares method.In this paper,a power load correction prediction model based on the trend of three peaks and three valleys is proposed,and the prediction accuracy is improved,reaching 3.36%,which can meet the actual production requirements.