Tech Insights

‘Brain’ Power: Smart Management of Smart Grid Renewables—Part 3

May 29, 2024 by Liam Critchley

Artificial intelligence can address management challenges caused by integrating renewables into the grid.

Integrating renewable energy into the grid is predominantly facilitated through microgrids. Microgrids utilize various renewable energy sources and components, including solar panels, wind turbines, battery storage, and backup generators. These comprise distributed energy resources (DER). 

However, adding these renewables to the grid can cause compatibility issues, inertial changes, and management problems. Artificial intelligence (AI) can help smooth the process by monitoring grid operations and processes with more data insights. 

 

AI and renewable energy.

AI and renewable energy. Image adapted from Wikimedia Commons

 

AI in Distributed Energy Resource Management

AI can help improve overall energy management capabilities by using its inference capabilities to cope with missing or unstructured data. This could support multiple management areas, from optimizing electric vehicle charging and mobile energy loads to scheduling resources when most needed. These processes, especially load optimization, are challenges for many energy management systems. However, machine learning algorithms can learn the resource patterns over time to determine the optimal loads for given scenarios.

These advantages make AI an ideal solution for integrating renewable energy sources and optimizing microgrids and DER. Improving DER can also act as a defensive line when backup power and peak demand responses are required.

 

AI-Enabled DERM Solutions

Distributed energy sources are more complex to integrate into the grid and can often compromise grid reliability. AI can help optimize DER assets in different locations and weather conditions. Several platforms exist for managing DER using AI. 

Machine-learning-based platforms manage DER interactions and track real-time flexibility. Machine learning algorithms can improve reliability and optimize grid operations because the AI can track DER assets and account for their different interactions. These platforms can create virtual power plants to assess the needs of various DERs around a region and offer energy distribution services on demand. Without integrating AI, combining and tracking so many variables in real time would not be possible.

Machine learning algorithms work by detecting any trade-offs or synergies between resources and how they can be optimized to fulfill grid objectives. Understanding each resource’s level of flexibility (historical patterns, behavioral impacts, and weather data) enables the AI algorithm to optimize which DER resources should be distributed at different times. To do this, the algorithms factor in the natural fluctuations that solar and wind energy generation systems undergo and the grid’s dynamic nature (including peak demand) to adapt in real time to changing demand. The machine learning algorithms can also detect and adapt to any anomalies to improve grid reliability and stability.

 

NREL’s Colorado DERM Project

In 2021, the National Renewable Energy Laboratory (NREL) implemented a distributed energy resource management (DERM) platform, Reorg, in Colorado. The project aimed to increase system grid resilience in community microgrids. The project optimized the community grid to make it more stable, resilient, and scalable for integrating multiple solar installations and mobile battery energy storage systems

Reorg used multi-agent deep reinforcement learning to facilitate a two-level control strategy for managing DER in the microgrid. The aim was to manage normal grid operations and resilient grid operations. The project’s system used a cell control agent to control the DER and cell clustering agents to reconfigure the networks.

 

Machine learning to integrate renewable energy and other DER.

Machine learning to integrate renewable energy and other DER. Image used courtesy of NREL

 

The project also utilized an area where solar generators and services could cause critical loads on the microgrid. NREL chose the site because the region has 760 residential and 379 commercial photovoltaic (PV) customers, with 4.5 MW peak demands and 400 kW distributed PVs. On the energy generation side, a 5 MW PV farm had just been installed when the project started, with plans for >10 MWh energy storage systems. Several sites require peak energy at different times, including a waste-water treatment facility, Roaring Fork Transportation Authority (RFTA), and Aspen Pitkin Airport.

Reorg has shown that machine learning systems can make microgrid organizations more stable and resilient. Combining AI algorithms with IoT platforms can realize a distributed and adaptable cell management system. Solar PVs and other DERs can adapt to varying system conditions during local wildfires. The developed approach may be a scalable method for implementation in community microgrids to improve grid resilience nationwide.

 

AI Improves Renewable Energy Modeling

With many renewable energy sources acting as DERs on the grid, understanding how they interact with the local grid and affect performance ensures renewable energy sources can be integrated successfully. One example is ChatGrid, a generative AI tool from Pacific Northwest National Laboratory (PNNL) that visualizes power grids. ChatGrid uses data from another PNNL software tool called Exascale Grid Optimization, which simulates the power grid in real time to detect disruptions. ChatGrid runs on a public large language model and operates like predictive text on phones and computer programs.

ChatGrid was developed as a question-based interface where human operators can inquire about DERs around the grid. For example, they can ask about the wind power generators’ capacity status in a certain region. ChatGrid provides a visual representation of the information the operator is looking for, such as power flow, voltage, or generation capacity.

 

Video used courtesy of PNNL

 

A collaboration between MIT’s Laboratory for Information and Decision System and Tennessee Tech University is taking place over the next three years in Appalachia. This project will study the effects of integrating different renewable energy generation technologies in rural communities to understand the local grid’s load when renewable energy harvesters are connected to the grid at various locations. The project uses a generative AI model alongside a microgrid simulation platform.

 

Intelligent Renewable Integration

Using AI in the renewable energy sector primarily focuses on optimizing distributed energy resources to enable better energy resource management. AI assists grid operators in determining how these energy sources can be used as a backup to changing energy demands. Interest in using AI to enhance modeling approaches and impact assessments for renewable energy integration in microgrids is also growing.