Machine Learning-based Model to Estimate the Dynamic Hosting Capacity in Distribution Network

The efficient and reliable operation of electrical distribution networks (EDNs) relies on enhancing their hosting capacities (HCs). This paper introduces a machine learning (ML) model to estimate the dynamic hosting capacity of an EDN, focusing on the seamless integration of air source heat pumps (ASHPs) and their interaction with demand response programs. A predictive model has been developed using optimization techniques coupled with ML algorithms, including the Multilayer Perceptron. This model can dynamically adapt to parameter variations, such as substation capacity, outdoor temperatures, customer engagement, and load flexibility. The methodology has been implemented on the IEEE 33-bus test EDN, demonstrating the model’s proficiency in precisely evaluating the impact of uncertain parameters on the HC. The results prove the high accuracy of the designed model to enhance the planning and operation of modern power systems, facilitating more effective integration of ASHPs and optimizing grid performance under diverse scenarios.