Cooling Load Measurement Model Based On Support Vector Regression And K-Means Clustering

With the improvement of residents' living standards,the number of air conditioners gradually increases,and the cooling load shows a trend of increasing year by year.The reasonable and effective calculation of the cooling load by the construction model will greatly improve the short-and medium-term load forecasting accuracy and provide certain support for the grid operation arrangement.The calculation of basic load is the basis of the calculation of cooling load.In recent years,due to the more complicated changes of the basic economic environment,the timing distribution characteristics of the basic load have changed to some extent,mainly reflected in the increase of the monthly difference and the increase of intraday fluctuation,etc.The traditional cooling load measurement method is difficult to adapt to this complex environment,and the calculation has great limitations.Based on this background,this paper conducts relevant research on the effective measurement and verification of the cooling load under the condition that the basic load changes are more complicated.This paper firstly analyzes the main influencing factors of load,and analyzes the factors that have a great influence on the cooling load,such as meteorology,and typical load changes.On this basis,the complex environment faced by the current cooling load measurement work is further analyzed.The monthly difference of the basic load during the measurement process is excessively analyzed,and the intra-month and intra-day fluctuations are also analyzed,as well as the economic development and extreme weather's impact.Secondly,based on the problem of the variation of the basic load time series distribution characteristics,a combined calculation model of cooling load based on support vector regression and K-means clustering is constructed.The model includes a single-stripping model of cooling load based on SVR(Support Vector Regression)-Winters and a secondary stripping model based on EMD(Empirical Mode Decomposition)-Kmeans,which respectively solves the problem that the basic load is difficult to adapt to the traditional method.And intraday fluctuations and other issues.Again,the traditional method can only judge the reasonableness of the cooling load curve based on experience.The two-dimensional measurement verification and correction model based on load timing distribution characteristics is constructed,including the verification and correction model of the cooling load based on time series decomposition and power comparison.The probability density calculation model of the cooling load of the quantile regression is used to verify the rationality of the cooling load curve from the angle of the temperature deviation of the cooling and the probability density distribution of the point of the cooling load.Finally,based on the actual data of a province in Northwest China,the two models proposed in this paper are used to calculate the cooling load of the province,and the corresponding calculation and correction model is used to verify and correct.The results show that the proposed support vector regression-K-means clustering cooling load measurement model has better adaptability to the complex environment faced by the cooling load measurement.Compared with the traditional method,the model has higher precision.