Deep deterministic policy gradient-based energy storage control and distribution between two microgrids

• A central energy controller using the deep deterministic policy gradient (DDPG) strategy as an artificial intelligence technique to control the ESU energy distribution between two adjacent SMGs. • The installation of multiple ESUs in the multi-SMG region is eliminated by using individual energy storage units. Consequently, this reduces energy costs and makes the improvement of an energy management system easier. • The advantage of the decision gradient in the DDPG method (continuity) is to exploit the surplus energy remaining in the battery for the benefit of the other microgrid instead of other methods that give discrete decisions through which the controller cannot obtain the remaining energy in those batteries • Depending on DDPG, energy mismatch can be overcome by forecasting by abandoning the need to build a system to predict the behavior of the system by forecasting the dynamics of the load demand, energy prices, and actual need for battery energy for two networks for a full day before the day of use. Recently, energy storage units (ESUs) have become indispensable energy sources in the power supply market. So, the roll-out of an ESU necessitates autonomous energy control systems that can use real-time data at a high energy distribution level in the smart microgrids (SMGs). This aims to decrease users' energy costs and address the uncertainty in the stochastic behavior of these microgrids. This paper proposes a central energy controller using the deep deterministic policy gradient (DDPG) strategy as an artificial intelligence technique to control the ESU energy distribution between two adjacent SMGs. Using a central controller eliminates the need for sub-controllers, which increases the system's complexity and thus degrades its efficiency. The proposed control strategy's superior efficiency in decreasing the energy cost is confirmed by applying the proposed strategy to the training and testing data and validated by comparing the results with those of the traditional artificial bee colony (ABC) algorithm. The simulation results illustrate that the introduced control strategy achieves an error rate less than 2% compared with the conventional method (ABC). In addition, the proposed system structure eliminates the need to install an additional, expensive energy storage unit in the second SMG.