Using Action Dependent Heuristic Dynamic Programming and Genetic Algorithms in the Energy Resource Scheduling Problem

Abstract

Energy management in smart buildings and homes has become an important issue. Proper energy management is judged upon the amount of consumed electrical energy as well as the total electricity cost. In this master thesis, two optimization algorithms, namely Action Dependent Heuristic Dynamic Programming (ADHDP) and Genetic Algorithms (GA) are used for the energy resource scheduling problem. The main objective of the renewable energy resource scheduling problem is to decrease the electricity cost over a fixed time period while meeting demand. In this work, ADHDP and GA were trained and evaluated on different simulation scenarios with various amounts of available renewable energy. It was demonstrated by computer simulations that both ADHDP and GA are effective in cost minimization compared to the baseline method. A correlation between optimization improvement and available renewable energy was also confirmed by computer simulation in all scenarios.