The Research And Implementation Of Communication Base Station Energy Consumption Prediction Based On Data Mining

With the expansion of the communication network and the rapid increase of communication equipment,the supervision and administration of communication energy consumption gradually receives widespread social attention.The research object is energy consumption time series data in communication base station,and the new prediction models are put forward,which include trend and event prediction,on the basis of existing time series data mining algorithms.The existing time delay neural network(TDNN)simply models with historical data,which has certain limitations.Therefore,the energy consumption prediction model(NARX)with exogenous series is put forward.Its main idea is to obtain information by not only the historical data but also the information external to historical data,which leads to better performance.There are two important parameters of NARX model:first one is the effective and meaningful exogenous series,second one is the input delay value of the model.Both of them are determined by correlation theories,so that exogenous series can provide more effective and meaningful information and minimize the redundant information at the same time.The accuracy and stability of the NARX model is verified by experiments.An energy consumption event prediction model is put forward,which includes three steps:(1)Distinguishing time series feature extraction,which reduces the data dimension to two-dimensional space;(2)Feature clustering with the feature set obtained by the previous step;(3)Event prediction pattern identifying,which use clustering feature vector as the input data of the classification model.The realization and parameter tuning are described in each step.The slope threshold 8 in the distinguishing piecewise linear extraction method,the clustering number K in the clustering algorithm based on distinguishing feature distance,as well as the observation time M in the classification algorithm,are all selected by experimental comparison results.The experiments verify the good performance of the model.With the existing energy analysis system in the laboratory,the two energy consumption prediction models are encapsulated as a prediction analysis module,and integrated into the system,which makes up the lack of future data forecasting function in the system,and provides more scientific basis to the energy management agencies and personnel.The integration is done with the software engineering theory,including module needs analysis,logic process design,database design,static class design,etc.Finally,the integration module interface and operation results are displayed.