Power Management and Optimization for a Residential Smart Microgrid Using Stochastic Methods

摘要

In this paper, energy usage and its associated price for a residential smart microgrid are analyzed using three different forecasting methods: Auto-Regressive Integrated Moving Average (ARIMA), Support Vector Machines (SVM), and Polynomial Regression. Energy demand and its price are then forecast while taking into account the effect of demand response. The accuracy of the forecast values are evaluated using Mean Absolute Percentage Error (MAPE) and Mean Squared Error (MSE) criteria. Numerical results, based on the data acquired for a residential microgrid from San Diego Gas & Electric (SDGE), are used for model validation purposes. Such data are employed to assess the performance, demonstrate the effectiveness and verify the reliability of the proposed optimization and forecasting methods. Our analyses show that the ARIMA method is more accurate in forecasting the demand as well as the price of energy for the smart migrogrid compared to other methods.