Data-Driven Power Flow Solutions for Distribution Networks with Solar Energy Resources

This article explores the efficacy of artificial neural networks (ANNs) to approximate the AC-power flow solutions for distribution networks (DNs) with solar energy resources. Generally, the power flow equations are non-linear and non-convex, making them difficult to solve directly. Traditionally, different classical linear algebra and heuristic search-based methods are developed to solve power flow problems, but those are difficult to implement for bulk radial distribution networks due to their complex and challenging mathematical formulations. Recent advancements in data-driven techniques, particularly neural networks, open the opportunity to tackle such complexities. In this regard, this article developed two ANN structures for determining the bus voltages, network loss, and substation power for radial DNs. Initially, a power flow dataset is synthesized by generating the load and solar power scenarios using Monte Carlo sampling followed by power flow calculation in OpenDSS. Later, the prepared dataset is implemented on the two developed ANN structures and the performance comparisons of them are provided for the IEEE 33 bus test system.