Optimal operation by recurrent neural networks for a renewable energy power-grid

Because large demands for electricity due to rapid increasing of population growth, depletion of fossil fuels and reducing greenhouse gas emissions, power generation systems by renewable energy have widely been studied, and introduction of the power generation systems into many fields such as houses or buildings is accelerating. Essentially, supplying electric power by renewable energy often becomes unstable because the amount of the electric power generated by the renewable energy depends on the weather conditions. Then, we need to introduce a sophisticated control method to maintain supply systems stably. From this view point, M. E. Gamez et al. proposed an optimal control method using recurrent neural networks for a wind solar power energy generation system. In the conventional control method, optimization problems for the wind solar energy power generation system are regarded as the linear programming problems, and they solved the problems by the recurrent neural networks. Then, results indicate that the control method has much possibility to apply into the real power generation systems. However, only small sizes of the systems are evaluated for the control method. Then, we evaluated the control method using more realistic power generation systems in this paper. In this model, consumers which have the wind solar power generation systems are connected by electric power lines. From the results of numerical simulations, the control method with the recurrent neural networks exhibits good performance even if more realistic conditions are installed.