IRES Track I: Sensors and Machine Learning for Solar Power Monitoring and Control

Arizona State University, Scottsdale AZ

Investigators

Abstract

This program promotes international multidisciplinary research opportunities for U.S. students at the overlap of sustainability, power systems and signal processing with the aim of improving efficiency in PV power generation. Algorithms for shading prediction and fault optimization will advance the state of the art in remote solar array management. Training students in machine learning, vision and data processing for energy systems is unique and requires an integrative approach. IRES participants will be immersed in producing and understanding solar analytics and creating algorithms and software to control solar arrays. The IRES program will engage faculty researchers from the Arizona State University SenSIP center and from the University of Cyprus KIOS Center in solar energy research. Programs and workshops will be established so that IRES participants are trained in machine learning for energy systems and present their research results in international settings. Weekly presentations at the international site and guidance by international mentors will enrich the cohort research experience. Embedding students in the KIOS center research labs funded by large European Union (EU) grants will provide knowledge on EU and international research practices, energy standards and policies. Students will spend six summer weeks at the University of Cyprus KIOS center to improve their research skills and elevate their cultural competencies. This international research endeavor will energize students to innovate and disseminate results globally. Solar energy or photovoltaic (PV) arrays encounter loss of efficiency under conditions of shading, panel faults and temperature variations. In fact, shading, weather patterns, soiling, and temperature reduce power output considerably. For example, a malfunction of one panel will cause an entire PV string to fail. To minimize inefficiencies, individual panel current-voltage (I-V) measurements, weather information, and imaging data are essential. Controlling the power output is possible through solar panel matrix switching and optimization (i.e., changing certain array connections from series to parallel using actuators). Matrix switching using programmable relays allows for different interconnection options. The research goal is to optimize PV array systems by: a) exploiting the measured I-V patterns to detect faults using machine learning, b) employing advanced imaging and vision techniques to predict shading, c) using temperature, irradiance and weather data to elevate PV efficiency, and d) include smart grid interfaces. This collaborative IRES project between Arizona State University and University of Cyprus will engage students in the following research problems: a) How do we use imaging to detect cloud movement, predict shading and elevate efficiency? b) How can the array connections be reconfigured based on imaging, weather, and I-V data to elevate efficiency? c) How can we detect and classify panel faults in real time using machine learning and other algorithms? d) How do we extend these solar monitoring and control concepts from utility-scale solar farms to house rooftop systems? 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.