Optimal Energy Supply Scheduling for a Single Household: Integrating Machine Learning for Power Forecasting
In this paper an optimal scheduling for energy resources and consumers for a single household is presented with the integration of renewable energy generation and battery storage systems. Based on the user's preference, three different optimization objectives are defined. The objectives are cost savings, reduction of CO2 emissions and user comfort. Two objectives are modeled as mixed integer linear programming (MILP) problems. The last one is modeled as a non-linear optimization problem. The approach furthermore utilizes several machine learning (ML) algorithms to forecast power generation and small load aggregation. The output of the ML forecast algorithms is therefore used as input for the optimization. The results show that the proposed approach is capable of optimizing energy supply at the level of a single household, based on user preferences for the objective of the optimization. In addition, better day-ahead planning of generation and demand is made possible by the use of a Convolutional Neural Network (CNN) together with other ML forecasting algorithms.