Deep Learning for Smart Grid Applications

In an effort to prevent climate change and make clean energy accessible to all, the global community is shifting away from electricity generation from fossil fuels and toward renewable energy sources. However, the electricity generated from renewable sources is stochastic in nature, exhibiting significant intermittency and variabilities, necessitating a smart power management system from the point of generation all the way through to the point of consumption. To make the power grid smarter, in this research, two individual projects are carried out with the aim of smart energy management from the generation end to the consumer end. With the help of deep learning integration, firstly, we modeled and analyzed a convolutional neural network (ConvNets)-based forecasting method for regional renewable energy (wind power) ramp. Secondly, we developed a reinforcement learning-based control scheme in droop-virtual inertia control for resilient community microgrids. Wind generation is highly intermittent with enumerable ramp events, making power management challenging. A smart forecasting method can improve system management and security to address this issue. However, spatial dynamics hinder the accurate forecasting of physical system data. To account for these spatial dynamics of physical systems, herein, we create a unique model for ramp forecasting in wind generation using ConvNets. First, we suggest a dependency and multi-resolution image processing technique that may enhance the geographical dependency of physical system data to circumvent the current constraints on the use of deep learning in this field. Because this multi-resolution enhancement maintains the high spatial dependency along spatial data, the kernel of ConvNets becomes very effective during feature extraction, and translationally variant characteristic has lifted. As a result of these two enhancements, the ConvNets may now utilize data from the physical system, allowing for more accurate forecasting compared to the state-of-the-art benchmark approaches. The community microgrid concept is becoming more attractive due to the large penetration of renewables and the adoption of battery systems of electric vehicles (EVs) as energy storage. However, the stochastic nature of these loads injects instability due to voltage and frequency fluctuation. Herein, we propose a novel microgrid control strategy incorporating reinforcement learning to address this issue in the microgrid system. A resilient community microgrid model, which is equipped with solar PV generation and electric vehicles (EVs) and an improved inverter control system, is considered. To fully exploit the capability of the community microgrid to operate in either grid-connected mode or islanded mode, as well as to achieve improved stability of the microgrid system, universal droop control, virtual inertia control, and a reinforcement learning-based control mechanism are combined in a cohesive manner, in which adaptive control parameters are determined online to tune the influence of the controllers. The microgrid model and control mechanisms are implemented in MATLAB/Simulink and set up in real-time simulation to test the feasibility and effectiveness of the proposed model. Experiment results reveal the effectiveness of regulating the controller's frequency and voltage for various operating conditions and scenarios of a microgrid.