Research devises new Software Application for Power Grid Simulations

Fundamentally, most modern cars and phones are programmed to learn from their environment – facial features, sounds, and even common driving routes. Due to the patterns of recognition, these systems can precisely predict and suggest favored options at the blink of the eye.

On these lines, if a system could be designed for the precision and responsiveness of critical national challenges, such as weather forecasting, disease diagnoses, and power grid reliability.

Smart Power Grid Simulator – a new software application – uses neural networks for efficiently solving power grid simulations. Such simulations are crucial for planning and optimizing electricity delivery. In fact, the initial test results of Smart Power Grid Simulator showed it solved power flow computation approximately three times faster than the traditional numerical model, that too without loss of precision.

Application first ever AI-based for Power Grid Simulations

Meanwhile, Smart Power Grid Simulator uses a new neural network technique called multi-task learning modeling, according to developers of the software application. And, the application is the first ever of AI for power grid.

“In the last several years, advancements in AI and high-performance computing has allowed to explore the method,” said one of the researchers behind the application.

The virtual presentation of the research is scheduled in November during Supercomputing 2020. The event is the world’s largest yearly gathering of professionals in the fields of networking, high-performance computing, storage, and analysis.

In terms of capability, Smart Power Grid Simulator is nearly three times speedier with fewer iterations and nearly the same accuracy, when compared with current power flow simulation instruments.

Currently, for the flow of power at optimum levels, power grid operators use offline computer models that examines data based on a host of scenarios.

AI finds use in new way to Increase Power Grid Resiliency

A new artificial neural network model can handle static as well as dynamic features of a power system with a considerably high degree of accuracy, according to a team of scientists at Argonne behind this.

In the real sense, America’s power grid system is dynamic and large, which makes it challenging to manage. From a practical perspective, human operators have the required skill to maintain systems when conditions are static. However, in the event, conditions change abruptly, for example due to sudden faults, operators tend to falter. They lack understanding of how the system should best adapt to satisfy system security and safety requirements.

To address this, a research team at the U.S. Department of Energy, Argonne National Laboratory has developed a novel approach for operators to understand how to act. The approach enables operators to better control power systems employing artificial intelligence. And, could help operators control power systems more effectively. This could improve the resiliency of power grid, says a recent article published in IEEE Transactions on Power Systems.

New approach enables to handle Static and dynamic Issues

Employing the new approach, involving a sole decision-making model with improved accuracy, it enables operators to make decisions taking into account static as well dynamic features of a power system.

“The decision to switch on or switch off a generator and learn its power output is an example of static decision. On the other hand, electrical frequency, which is associated with the speed of a generator is an example of dynamic feature. This is because electrical frequency fluctuates over time in the event of a disruption or an operation,” stated the costa-author of the study. Whilst, if static and dynamic formulations are included in the same model, it is practically impossible to solve.