The stability of power grids supplying energy to our homes and businesses is emerging as a critical success factor for maintaining both physical and economic security. Regardless of the electric grid size—a microgrid with several hundred users or a regional grid with millions of users—grid complexity is on the rise. Trends such as the sharp increase in Internet of Things (IoT) devices and digital solutionsthat tax large and small grid capacities, a growing installed base of renewable energy generation, and the proliferation of distributed energy resources (DER) behind the meter at the grid’s edge, are all contributing to this complexity.
Unfortunately, the electric design of existing grids has not been optimized to accommodate these new behaviors. Local and national grids are experiencing significant changes in energy flows. In the past, the energy consumption pattern was easy to interpret. At around 6:00 p.m. rates of energy consumption were high and not much energy was consumed during the night. Now, businesses, hospitals, and homeowners, for example, may be producing energy at midday with solar panels. An increasing quantity of electric cars are also consuming energy in the middle of the night to recharge their batteries. Energy prices that sometimes change minute-by-minute are also altering where and when enterprises choose to consume the most energy. This makes the flow of energy within grids more volatile. Instead of traditional top down energy distribution, grids must deal with bidirectional energy flows which change throughout the day and night.
To accommodate the changing nature of the grid, stakeholders are investing in both technology and manpower to better forecast short-term evolution of electricity in their operation. Without accurate consumption forecasting, nodes within the grid can become saturated with energy demand that is too high which, in turn, can lead to grid failures and blackouts.
Why AI Digital Solutions Can Improve Forecasting Accuracy
This is where we at Predictive Layer, a Schneider Electric Technology Partner, offer support. Electrical utilities and microgrid managers need to forecast, with high accuracy and speed, short-term grid energy consumption across the geography of their grids. We have developed an artificial intelligence (AI) engineto address the new layers of grid network complexity.
Today most large grids deal with the challenge of energy consumption forecasting by assigning teams of mathematicians to automate and scale the forecast. Should the parameters of energy consumption change—an extreme weather event, or a societal event where many people are forced to work from home—those complex mathematical models have to be rapidly reworked in order to align with the new reality. When thousands of grid network nodes are involved, the task becomes too complex to execute in a rapid fashion—even for the best mathematicians. The task is also challenging for microgrids and smaller local grids who typically have few human resources to dedicate to forecasting.
An AI-driven forecasting engine learns from both past energy consumption patterns and from real time live situations. For example, the changes that are taking place within a grid (like the addition of new transformers in a location) and the intricate nuances in energy consumption demand automatically are fed into the model on an ongoing basis. This includes data from transformers, information about the behavior of subscribers connected to the network, and information surrounding weather-related and calendar date-based patterns of consumption.
Benefits Include Less Downtime and Lower Operational Costs
The AI engine learns how to better forecast what is happening every day, everywhere within the confines of the grid, and continues to get smarter as it processes more data. At each point of the day or night it can generate, on behalf of grid operators, local schedules of how much energy should be delivered to designated locations and expected volumes of energy coming in or out. The AI forecasting tool also automatically reconfigures itself as changes occur across the grid. If a new enterprise suddenly turns the lights on in a new facility within a geography, the engine can recalculate that the energy consumption peak will now occur at 4:00 p.m. rather than at 6:00 p.m. in that area.
Instead of a team of people generating 10 forecasts of what is happening in each of their regions, the software can generate thousands of such forecasts in minutes. These forecasts can account for every locality within the grid. (In New York state, for example, there may be 10,000 transformers in the entire network that distribute energy to users.)
The benefits of having access to such an automated system include less network downtime, better operational management and cost control, and higher efficiency of electricity distribution. The tool also makes it easier to manage a crisis where rapid changes in consumption require immediate actions. The goal is to sustain the level of customer expected quality and uptime regardless of the changes that are taking place across the grid.
Originally posted onSE Blog & Authored by Olivier Cognet
About the author:
Oliver Cognet, CEO, Predictive Layer, is a business leader, corporate developer, and entrepreneur. He has 25+ years of experience in High Tech, Software Development, and Telecom markets focusing on enterprise and consumer markets. Mr. Cognet knows that technology and innovation can help businesses better understand the behavior and needs of consumers, thus helping businesses to grow. Predictive Layer brings the full power of machine learning artificial intelligence (AI) to businesses so they can simply and intuitively model and establish forecasts and predictive analytics of their key target indicators.