Smart Customer Demand Forecast Wıth Substantıal Energy Demand Varıabılıty

Abstract

One of the most challenging tasks in the emerging smart grid environment is Load forecasting. Load variations are highly unpredictable in smart grid environment and This paper focuses an innovative approach to forecast the smart customer load by making use of energy consumption data obtained from smart meters. The data obtained from smart meters is analytically linearized first by applying extended k-means clustering approach.Then linearized load profiles are smoothened and then linearized by using Taylor series linearization process. Case studies are reported using smart meter data of real system and then the proposed approach with artificial neural network. Four different cases are considered for forecasting. Results showed that, for the instance of high variability in energy demand, accuracy of forecasting is high using linearized profiles than using original non-linear profiles as the source of forecasting. Performance and robustness of the approach is verified by repeating the forecasting process several times and the results justified that accuracy of the forecast is further enhanced with the proposed approach