Impact on Distribution Network Hosting Capacity using Machine Learning Based Reactive Power Support from PV Smart Inverters

Abstract

Nowadays, due to growth of advanced technology and its impact on people lives, there is an increasing need for aligning and optimizing infrastructure with people needs. Harnessing renewable energy from different sources, store and eventually transmit and distributed it is an all-time challenge which nowadays by implementing and proper maintenance we lead to having higher efficiency to the point where we are able to maximize usage of these resources. There is a concept called Hosting capacity to measure proper amount of integration of distribution generators into the grid without causing any malfunctioning in the system. The goal is to find a way to enhance the HC. Predicting reactive power is a good solution to halt voltage violation. Techniques such as optimal power flow has been recommended for that however, machine learning algorithms is new approach for reactive power predication due to its better performance and ability to consider dominant variables affected on the data set. Thus, the literature review has been conducted to choose a ML algorithm for time series prediction of reactive power on a chosen Network. After, the methodology has been illustrated. Then, a dataset has been generated by doing power flow analysis and used in the ML algorithm in order to compare the results. In this case, for some parts the results were not fulfilling compared to the generated data where the affected factors have been discussed and future works proposed.