Research On Electrical Load Type Recognition Method Based On Decision Tree And Bayesian

For the power department,understanding the composition of power user's load through power load type identification can not only enhance the management of load side,but also guide users to consume reasonably.At the same time,it can also reduce network loss and adjust peak-valley difference,improve the accuracy of forecasting power load,and provide more accurate data parameters for power system simulation and power system planning.For consumers,through the identification of electrical load to achieve the understanding of load switching time,power consumption and other parameters,reasonable reduction of unnecessary power consumption,so as to achieve the purpose of energy saving.For managers such as student apartments,understanding the use of electrical devices in student apartments through electrical load identification can avoid some potential safety hazards caused by irregular use of electrical devices.In view of the difficulty of field data sampling for various electrical loads in use,simulation circuits for different types of electrical loads are built by using simulation software to obtain sample data.The start-up current characteristics of electrical load are selected as the start-up time,the maximum start-up current time and the maximum start-up current.Meanwhile,the steady working current of electrical load is processed by fast Fourier transform(FFT)to obtain harmonic parameters as the current spectrum characteristics.A data preprocessing method based on interval of discrete sample eigenvalues is proposed.According to the different parameters such as power,load and current,we collect multiple sets of feature data,and select some of them as the sample data constructed by the classifier,and the rest as the test data of the classifier.The decision tree classifier and Bayesian classifier are used to classify the types of electrical loads respectively.On this basis,a design method of combined classifier is proposed,which uses overlapping thresholds of parameters as the critical value and combines decision tree with Bayesian classifier.The overlapping thresholds of parameters with the highest classification accuracy are the optimal overlapping thresholds of combined classifiers.The accuracy of combined classifier is 22.5% higher than that of Bayesian classifier and 5.65% higher than that of decision tree classifier.The results show that the combined classifier has better classification effect and more accurate classification.