• Renewable energy generation
• Solar energy power generation forecasting
• Clean energy generation including new nuclear reactor designs
• Smart and resilient electric grid with distributed power generation
• Micro-grid requirements and robustness

The renewable energy research program being developed at Cal Poly by Dr. Belanger has three long term goals. The first one is to develop a model using sky imaging analysis and commercially available solar power generation software to forecast up to an hour in advance the actual power generation of a typical utility scale solar field. The new Cal Poly solar field is being used as the primary test site for this investigation. Some of this work is done in collaboration with First Solar, one of the largest solar companies in the country, who has developed a new power generation software now available on the market.

The second goal of this research program is to collaborate with REC Solar (REC), the Cal Poly solar field managing company, Dr. Dale Dolan (electrical engineering) and Dr. Andrew Davol (mechanical engineering) to find ways to better optimize the power generation of the Cal Poly solar field, called Gold Tree. The overall power generation of the Goal Tree field at this time is significantly lower than the capacity originally projected by their model. The lessons learned in this research program will also be applicable to other solar fields elsewhere in the country. To investigate this problem, a programmable dual axis tracker built by Cal Poly students is being used to test the accuracy of the forecasting model and validate some of the modeling assumptions imbedded in the power generation software. The programmable dual axis tracker (see Fig. 1) was recently completed by two students with grant support from Summer Undergraduate Research Program (SURP) and Research, Scholarly and Creative Activity (RSCA).

The tracker is also be involved in the third goal of this research program, which consists of developing a micro-grid laboratory with Dr. Andrew Davol. The laboratory will facilitate testing and experimenting with different components of these systems including the dual axis tracker. Solar and wind energy, as well as energy storage and overall system control, will be tested for both component basic research and as an integrated system. The micro-grid laboratory will also be an excellent educational tool for the department where students will be able to learn about and experiment with different components currently used in the renewable energy world.

During the 2018-2019 academic year, this research program mostly focused on testing and validating different tools, including the solar power generation software and the dual axis tracker. The goal then was to build some confidence in the accuracy of the software used to evaluate potential power generation of different solar fields. For the 2019-2020 academic year, the plan is to advance at least one-step further and start building the tools necessary for predicting weather conditions, especially cloud coverage and movement, combined with a more reliable power generation model to start forecasting future availability of solar energy. For the 2020-2021 academic year, the goal is to have a reliable tool capable of forecasting the actual power generation of a typical utility scale solar field up to an hour in advance. To help in this forecasting research project, I am actively recruiting a graduate student with a strong computer science background.

Reaching that one hour forecasting goal is becoming an important issue in California because more than 12% of the electricity now comes from solar energy, and this percentage is projected to double in the next 10 years. Because of the significance of the solar electricity generation in the state, the Grid System Operator (GSO) has to be able to compensate for large fluctuations in solar electricity production that often involve imminent cloud coverage over the field. The GSO typically responds to the fluctuation by rapidly getting natural gas peak power plants on line when these events occur. These plants typically can take 30 minutes to get ready before they can produce the required electricity to compensate for the loss in solar electricity generation.

Different imaging tools can be used to try to achieve this forecasting goal. Satellite weather images can be analyzed to give a big picture of the sky around a solar field but the images are low resolution, low frequency and the difficulty of extracting quantitative data about cloud coverage and cloud intensity make them a very uncertain forecasting tool when used alone. For a more detailed map of the sky’s cloud coverage, ground-mounted equipment is proficient for providing the necessary resolution around the field. One of the imaging instruments used is a Total Sky Imager (TSI), like one installed in the Cal Poly solar field. An image of the sky taken by that camera on May 22, 2019 is shown in Figure 2. The main issue with these cameras is that they project a 2-D representation of the sky, when in reality the cloud coverage is more often multi-layered, located at different altitudes and moving at varying speeds and in different directions.

The first step towards achieving our primary goal of predicting cloud coverage will be to use the data from the TSI to determine how good the data from our sole sky imager can be at forecasting power generation. To get a head start on this work, I teamed up last winter/spring with a CSC senior student. In the context of his senior project, he analyzed pictures generated by the TSI and tested some algorithms to forecast cloud movement and anticipated power generation of the solar field. His work also included testing of Artificial Intelligence (AI) learning algorithms that could in the long term become a sophisticated forecasting tool.

Once a forecast of the cloud coverage is generated, a power production software is used to determine the projected power generation. That power generation forecast can then be compared to the actual power generation of the site to validate the accuracy of the forecast. To build some confidence in the projected power generation model, this year we worked at determining how effective these commercially available software are at estimating hourly, daily, monthly, and yearly power generation using weather data recorded at the site. The two software used in the investigation are PVSyst, which has been the standard software used in the industry for many years, and PlantPredict developed by our partner First Solar, which is now a free software available on the market. The plan is to use both these software to determine how they compare and make recommendations to First Solar for making PlantPredict an even better tool than it is today.

The second goal of this research project is to work with REC, the Cal Poly solar field managing company, to investigate ways to optimize the power generation of the field and other similar fields. Discussions with employees at REC are currently ongoing to identify and understand the power generation issues they are facing with the solar field. Recent interactions with REC included multiple visits to the site to better visualize some of the challenges they are experiencing with topography, panel soiling and moisture issues. The plan is to use our power generation software to first identify possible panel tracking algorithm improvement, particularly in the morning and the late afternoon when the discrepancies are the most significant.

The next step will be to generate an actual 3-D topography of the site that would include the single axis solar trackers and panels to better understand the shading issues they are experiencing due to the uneven ground at the site. Significant morning dew and panel soiling are also issues that will need to be investigated to determine their overall effect on power generation. The programmable dual axis tracker will be used in this part of the investigation to help better understand and possibly reproduce some of those discrepancies in power generation of the Cal Poly solar field. We will be able to physically test how partially shaded panels, indirect/diffused light, panel soiling and a low angle of incident light can effect power generation and compare that to the assumptions made by the two different software. The result should be a more reliable power generation model and a better optimized power generation output for the Cal Poly solar field.

The ultimate goal of this research program is to combine these different efforts on an accurate forecasting model and on a reliable solar power generation software. The result would be a model capable of accurately forecasting the actual power generation of a typical utility scale solar field up to an hour in advance.

As for the third and last goal of the research program, we are actively working to build a micro-grid in the department solar lab. To accomplish this goal, we have initiated a Senior Design Project group of four students to build a versatile fixed mount for solar panels during the 2019-2020 academic year. Using an independent study, we are also including a student interested in instrumentation design. Recently we contacted the Wind Energy Club at Cal Poly to help design a data acquisition system for the wind energy component of the new micro-grid.