I-Corps: Deep Learning for Solar Energy Systems to Optimize Whole System Performance

点击翻译按钮获取中文摘要

The broader impact/commercial potential of this I-Corps project is to provide secure and high-performance solar energy to reduce pollution and energy cost, and help establish self-sustainable micogrids. As the proposed new deep learning-based controller taps into both hardware and software sub-segments for microgrids, the commercial opportunity is considerable. The proposed technology has the potential provide a large-scale microgrid system with affordable off-grid capability and establish a reliable microgrid system with cyberprotection. It offers a cost-effective energy solution to homeowners, schools, business owners, hospitals and clinics, military facilities and other organizations with significant power demands.This I-Corps project is based on the development of a new microgrid controller integrated with a deep neural network algorithm to optimize the system performance with both output and security. The model will help redesign photovoltaic (PV) filters, boost converters and micro-inverters by setting different parameters. Comparing with a simulation-based model, the deep neural network algorithm has a 15% margin on prediction accuracy and speeds up the response time from minute-level to millisecond-level. As a result, 4700 kWh of electricity can be saved annually for a 2000 sq ft house in a state with 80% sunny days. Due to the millions of parameters used in the algorithm, the new controller is able to support hundreds of users in a single grid. The core of the proposed project includes two deep neural networks (DNNs) using internet of things (IoT)-based sensors for the microgrid data processing. By leveraging the deep learning model and IoT sensors, the proposed system can monitor and detect abnormal usage pattern for the entire system in real time.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.