Application of artificial neural networks for power system state estimation - Validation with a weighted least squares algorithm

Abstract:
The energy transition led to significant challenges for distribution grid operators. In addition to the increasing renewable energies, increasing mobility and heat supply electrification are expected. Volatile feed-in situations, variable and sectorcoupled loads complicate the state estimation in less monitored power distribution grids. This paper uses artificial intelligence as a data-based approach for state estimation in low-voltage grids. The method is applied to the example of an urban test grid from the SimBench project and a rural test grid. The results are validated using a classical weighted least squares (WLS) algorithm. The introduced procedures consider a high penetration of electric vehicles and photovoltaic systems. AI-based methods turn the computationally intensive training phase to offline time and operate fast in online mode. Such methods are especially well-suited for real-time applications.