Generating And Analysis Of Energy Consumption Benchmark For Telecommunication Base Station Based On Data Mining

With the increasing demand of power consumption in base stations,the management of energy consumption becomes more important,due to the method which used to generate benchmarks for public buildings has been successfully applied,the methods:multiple linear regression,cluster analysis and artificial neural network,are referenced in generating TBBs’benchmark.First,multiple linear regression method was used to selected the significant features which influence the power consumption of base stations;next,clustering analysis was used to cluster the consumption data into 8 clusters which represented 8 typical pattern of energy consumption in base station;and suggestions about energy management were given after the analysis of energy benchmarks.Then artificial neural network method was used to predict the power consumption of base stations,and had better prediction accuracy than when it was used in public buildings,which had reached 7.55%of relative error.Then the analysis of feature that affect consumption were conducted and shown that the power of main device,PUE and weather were the most important features.Also,suggestions about energy management were given based on these three features after the analysis of energy benchmarks.At the end,the three methods were compared and advices about application scenarios were given.Multiple linear regression was the most economical and simplest method,cluster analysis was the best method for mining more information and knowledge,artificial neural network performed best at predicting.Then,the results about energy patterns are summed up to get the advices about energy management.