Predicting Smart Grid Stability with Optimized Deep Models

In a smart grid, consumer demand information is collected, centrally evaluated against current supply conditions and the resulting proposed price information is sent back to customers for them to decide about usage. As the whole process is time dependent, dynamically estimating grid stability becomes not only a concern but a major requirement. Decentral Smart Grid Control (DSGC) systems monitor one particular property of the grid—its frequency. So, it ties the electricity price to the grid frequency so that it is available to all participants, i.e., all energy consumers and producers. DSGC has some assumptions to infer the behavior of participants. DSGC system is described with differential equations. In this paper, we study on optimized deep learning (DL) models to solve fixed inputs (variables of the equations) and equality issues in DSGC system. Therefore, measuring the grid frequency at the premise of each customer would suffice to provide the network administrator with all required information about the current network power balance, so that it can price its energy offering—and inform consumers—accordingly. To predict smart grid stability, we use different optimized DL models to analyze the DSGC system for many diverse input values, removing those restrictive assumptions on input values. In our tests, DL model accuracy has reached up to 99.62%. We demonstrate that DL models indeed give way to new insights into the simulated system. We have learned that fast adaptation generally improves system stability.