This paper explores the possibility of using machine learning to improve the profits generated by an energy management system for so called prosumer households. Which are households with their own energy production, consumption, and storage. The commonly used deterministic algorithms only take the current data into account when deciding on how much power to buy or sell. But, if there are flexible energy costs and future energy consumption and production data is available, better choices can be made. For example, you could
store energy at a cheaper time for later usage when the prices are higher. To make these decisions, a new algorithm was created. In this paper different machine learning models are evaluated for generating the future data. The final simulations will use the improved algorithm use the data predicted by these models. For a prosumer with a 10 kWh battery and which produces 70% of its own total energy consumption, the machine learning algorithm decreased the total energy cost by around 7%. While this does show that machine learning
is a viable option to increase the efficiency of an energy management system, there is still much improvement possible to get closer to the theoretical 24% reduction.