Addressing Microgrid Uncertainties With Predictive Forecasting Models
A predictive model from South Korean researchers shows uncertainties in microgrid operations, considering intermittency, load, and day-ahead price factors.
Researchers from Chung-Ang University in South Korea recently developed a novel deep learning-based forecasting model to predict the optimal conditions for operating profitable microgrids.
This graphic from Siemens lays out the different types of microgrids. Image used courtesy of Siemens
The paper was recently published in Applied Energy’s special “Microgrids 2023” issue. According to Chung-Ang University’s press release, the model generated a lower forecasting error than incumbent prediction tools. It also yielded superior correlation values for predicting renewable availability in solar panels. In contrast, existing prediction methods consider future scenarios and probabilities but aren't always the most accurate.
The forecasting model addresses critical gaps in the planning of microgrids, which serve as networks of distributed energy resources and connected loads. Managing uncertainty has long been challenging in the cleantech market, as operators and investors contend with risks such as intermittent solar generation, supply/demand imbalances, and uncertain pricing. The same applies to microgrids, which typically integrate renewables, energy storage systems, and traditional energy resources.
In their study, researchers from Chung-Ang University’s School of Energy Systems Engineering focused on microgrids, a growing market supporting the global transition from fossil fuels-based resources.
The abstract explains that the study involved developing a “long short-term memory network” covering the intrinsic uncertainties of microgrids. To ensure the model was accurate, the researchers optimized its hyperparameters using a genetic algorithm-adaptive weight particle swarm optimization (GA-AWPSO) algorithm. They also added a global attention mechanism to mine input datasets for key features and improve the model’s overall performance.
Another part of the study addresses day-ahead price factors, which allow energy companies to lock in prices ahead of time. On the demand side, the researchers created a data mining and incentive-based demand response (or DM-CDIR) program to provide ideal incentive rates for potential customers. The tool uses clustering and classification functions to match customers’ bid/offer data with the ideal incentives.
They then simulated historical PJM Interconnection datasets, proving that the model performed well in managing uncertainties. PJM Interconnection data would be critical for this task as one of seven regional transmission organizations in the U.S. The Pennsylvania-headquartered firm serves electricity to 65 million people across 13 states and the District of Columbia, with 186 gigawatts of installed generation capacity as of mid-2022.
The researchers said the model could help accelerate the integration of renewables into power supply networks while allowing microgrid operators to manage day-ahead challenges. This will improve the reliability of regional grids and lower the cost of clean energy.
Microgrid Market Snapshot
With more utilities investing in solar and energy storage, microgrids are increasingly attractive for managing volatile energy costs, meeting the demand for continuous service, and offsetting outages during extreme weather events and other emergencies.
California’s Blue Lake Rancheria microgrid integrates a solar array, battery storage, and control systems. Image used courtesy of Schatz Energy Research Center
That last point is of particular value to states like California, whose grid resilience is threatened by wildfires and heat wave-induced demand response events. According to the U.S. Department of Energy’s Microgrid Installation Database, the state is home to nearly 100 microgrids, with most being solar/storage sites.
Beyond the west coast, microgrids are also popular across Texas, Alaska, and northeastern states such as New York. They’re also popping up in the Midwest—for instance, EE Power recently covered the first microgrid project to launch in Columbus, Ohio.
This map from the Department of Energy tracks renewable and non-renewable microgrid installations across the mainland U.S. Image used courtesy of the DOE’s Office of Energy Efficiency and Renewable Energy
The microgrid market is gaining significant traction. According to a recent analysis from Wood Mackenzie, U.S.-based microgrids reached 10 GW in the third quarter of 2022. Solar and solar-plus-storage installations show particular growth, with more than 175 projects scheduled to come online at the end of last year.