Machine Learning Based Sizing of a Renewable Battery System for Grid Connected Homes With Fast Charg

This paper develops a sizing model of solar photovoltaic (SPV), small wind turbine (SWT) and battery storage system (BSS) for a grid-connected home with a fast-charging plug-in electric vehicle (PEV). The home trades energy with the main grid under time-of-use tariffs for selling and purchasing electricity that affects the energy management. In this paper, a practical rule-based operation strategy is developed for the grid-connected home with fast-charging PEV that enables efficient and cheap energy management. The sizing problem is solved using a supervised machine learning algorithm, which is a feed forward neural network, by minimizing the cost of electricity. While the developed renewable-battery sizing model is general, it is examined using actual data of insolation, wind speed, temperature, load, grid constraints, as well as technical and economic data of BSS, SPV, SWT, and PEV in Australia. Uncertainty analysis is investigated based on ten scenarios of data for wind speed, temperature, load, insolation, and PEV. The effectiveness of the proposed model with fast-charging PEV is verified by comparing to slow charging and uncontrolled fast-charging models, as well as two other machine learning methods and a metaheuristic algorithm. It is found that the proposed model decreases the cost of electricity by 10.1% and 19.6% compared to slow charging and uncontrolled fast-charging models for the grid-connected home with PEV.