2-Stage Electric Load Forecasting Scheme for Day-Ahead CCHP Scheduling

Abstract

Smart grid technology has been gaining much attention as a solution for energy shortage and environmental pollution problems. For the deployment of the smart grid, among the various energy systems, CCHP (Combined Cooling, Heating and Power) has attracted much attention because it can reduce energy costs effectively by using the thermal energy generated by the power generation process for heating and cooling. In this paper, we propose a novel 2-stage load forecasting model and perform value-based CCHP operation scheduling based on the model. To construct our model, we first perform an hourly load forecasting using two popular algorithms for time series forecasting, XGBoost (Extreme Gradient Boosting) and Random Forest. And then, we combine their forecasting results using a sliding window-based Multiple Linear Regression to reflect the energy consumption pattern more accurately. The basic guideline of the CCHP operating schedule is to run CCHP only when using CCHP is more economical than using the public power system. We report some of the results.