Energy Efficient Scheduling in Smart Home using Deep Reinforcement Learning

In a smart home, the scheduling of the period of time that household appliances are allowed to be operational necessarily requires the adjustment of multiple parameters in accordance with the amount of available energy. Nevertheless, the scheduling of the operational time of multiple appliances in a smart home itself is a difficult problem, and as a result, it requires an intelligent, heuristic method in order to be solved in polynomial time. In this piece of research, we propose scheduling of household appliances based on a well-known value iterative reinforcement learning technique called Quality learning. This technique is used to learn values over time. The proposed method will be carried out in two stages. The first step in the Q learning process involves the agents interacting with the environment of the smart home in order to earn a reward for their efforts. The value of the reward is then used to schedule the operating times of various household appliances in the subsequent state so that the total amount of energy consumed is kept to a minimum. In the second phase, the user's dissatisfaction is maintained due to the scheduling of the household appliances. This is accomplished by classifying the household appliances into two groups: shiftable and non-shiftable. In addition, by making use of the phenomenon of shared memory synchronisation, the agents that are connected to each individual appliance in a smart home become synchronised. The simulations are carried out in a model of a smart home that consists of a single person and a number of different types of appliances. It has come to our attention that, in contrast to manual scheduling algorithm and scheduling that was based on a demand-response strategy, the operational time of the household appliances has been revealed to be effectively scheduled in order to reduce the amount of energy that is consumed.