The workshop formed part of a series of activities with “Training and capacity building on load estimation and short-term residual load flow forecasting at LV and MV feeders along with approximate state estimation for Indian urban electricity distribution systems”. These trainings are undertaken under the development cooperation project “Energy transitions with DISCOMs” within the framework of IGEN-GIZ. The project is implemented by a consortium of companies led by Fraunhofer IEE. The focus of the project is for the geographical area of Delhi with the utilities BRPL, BYPL and Tata Power as the primary stakeholders. The project objective is to develop an approach for estimating the voltages, power flows, etc in the distribution network with a low number of real-time measurements based on advanced ANN / Machine learning techniques. The necessity of the activity drives by the fact that currently there is a lack of adequate online measurement devices while there is an increase in penetration of Solar rooftop PV, battery energy storage systems, etc in the Indian distribution grid. Therefore, monitoring of the distribution network becomes extremely crucial to maintain stability. The workshop had four different sessions. In the first session, the methodology of the exercise and some results with generic data sets was presented to all the three Delhi DISCOMs. Subsequently, there were three individual sessions with Delhi DISCOMs where the results pertaining to their specific datasets were discussed. The individual sessions also demonstrated the process to train the machine learning models for different scenarios from the historical data set provided by the respective utilities. Detailed deliberations were held about the possible next steps. 

This project was highly appreciated by the utilities as a key step towards the robust operation of the system in light of disruptive changes brought in by increased rooftop PV and e-mobility penetration. 

For further information, please contact Mr. Markus Wypior markus.wypior(at)giz.de.