Deploying Advanced Solar Forecasting may better predict Generation and support Grid Reliability

Exclusive to EnergyTech By the Electric Power Research Institute and New York Power Authority

As the U.S. penetration of solar and other renewables increase to support the clean energy transition, it remains essential to ensure the energy system remains stable and reliable. Uncertainty in renewables output could challenge system operations.

The New York Power Authority (NYPA), Electric Power Research Institute (EPRI), and other research partners have completed a multi-year study focused on improving and understanding the use of solar forecasting for New York state. The findings could be used by other states and countries as they work towards meeting clean energy targets.

The uncertainty in the output of solar resources could make balancing supply and demand challenging. If solar penetration is high enough to impact operations, a robust forecasting tool can help manage solar uncertainty for energy companies. It would allow grid operators to partially reduce risk and support successful integration of solar power. Solar power forecasting plays a critical role for system operators and may become even more important in the future.

Most system operators across the globe use solar power forecasts to support their decision making. However, the methods are still relatively early in their development. In high penetration solar regions, forecasts must be made, beginning several days out and through real time, for both transmission and distribution connected resources, whether rooftop or large-scale solar power plants. Of particular interest to grid operators are methods that can better forecast large changes in solar intensity due to the formation and movement of clouds.

Multi-Phase Project

The team conducted the research in three phases from 2017 to 2022. The work built on the foundation of a prior 2012-2015 DOE-funded forecasting project that resulted in a suite of models that could forecast solar output across a range of time frames.

The recent study considered a comprehensive approach to solar forecasting, including:

·        Deployment of innovative sky imager technologies and methods at suitable locations, tied to the use of New York State MesoNet data

·         Improvement and demonstration of physical and statistical modeling approaches for solar forecasting, targeted to the entire state

·         Consideration on how these forecasts would integrate into system operations

·         Analysis of how the forecasts performed compared to currently available methods.

Phase 1 focused on sky imager networks for short-term forecasting. As part of follow-up to the previous DOE project, the team deployed and demonstrated a network of ground-mounted cameras developed by Brookhaven National Labs (BNL). NYPA selected sites in Albany and Long Island as demonstration locations, and the team installed instrumentation networks at the sites. The installation consisted of small, high-definition digital cameras pointing upwards to provide a 360-degree view of the sky and equipment to transmit the data via the internet.

The researchers observed cloud conditions in real-time and anticipated changes in the output of solar generation for surrounding areas using the BNL sky imager-based system. At the end of the various tests and deployments, the team obtained more than a year of data for each region. The resulting forecasts demonstrated the ability to outperform existing forecasting methods for short-term time horizons.

In the second phase, numerical weather models and statistical approaches, previously developed by the National Center for Atmospheric Research, were leveraged to improve hours and day-ahead forecasts. Researchers saw improvements by combining multiple models and sources of data using machine-learning methods. This methodology demonstrated the potential benefits for “nowcasting,” while also improving the numerical weather models.

The final phase focused on the demonstration and integration of these forecasting models into transmission system operations. This included analysis of forecast performance, as well as understanding forecast needs for various aspects of system operations. The team also focused on ensuring the methods lined up with power system operations in New York State, particularly with the New York Independent System Operator. The researchers set up data streams and recorded lessons learned on data delivery. They also developed a New York State Solar Forecasting Roadmap, which focuses on when and how parts of the overall project can be deployed to improve solar forecasting.

Outcomes

The study helped develop an underlying platform for solar and related weather forecasting, based on improved solar forecasts.

The research identified ways the forecast models could be deployed and the feasibility of each: incorporation of research by a forecast provider, public deployment, or deployment by a private entity. The new models could also form the basis of improved commercial tools, particularly for day-ahead forecasting to inform system operations based on weather modeling.

The project demonstrated how more extensive data and advanced solar-focused models may increase the accuracy and granularity that may be needed to support grid operations for the clean energy transition. Improved forecasts could be applied to individual solar plants and to predict distributed solar across a large region for generation, transmission, distribution companies, and private developers.

The research was funded by NYPA, the New York State Energy Research and Development Authority, and the U.S. Department of Energy Solar Energy Technologies Office, and co-managed by EPRI. Other partners included Brookhaven National Lab, National Center for Atmospheric Research, and the State University of New York at Albany. Advisors included the New York Independent System Operator and Central Hudson Gas & Electric.