Modelling Extreme Daily Peak Electricity Demand Across Indian States Using Non-stationary Generalised Pareto Distribution Models

An unparalleled rise in peak electricity demand across the tropics over recent decades signals the need for conscious planning of investment in power infrastructure and a boosting of demand-side management measures. This paper estimates the extremes in daily peak electricity demand across eight Indian states from 2010 to 2018 by using the extreme value mixture models and the generalised Pareto distribution (GPD) method. A combination of different mixture models is used to obtain the optimum threshold level, and the exceedances above it are declustered and fitted with a non-stationary GPD using the daily maximum temperature and trend terms. To our knowledge, this is the first attempt to apply non-stationary GPD models in Indian context for analysing the extreme peak electricity demand in the country. Results show that the shape parameter of extreme peak demand appears to be time-variant for the different values of maximum temperature and shows a linear trend for Punjab, Madhya Pradesh, Maharashtra, Gujarat, and Haryana. However, the scale parameter is found to be time-variant for all the states. Most states experienced the highest monthly frequency of peak electricity demand during July–October, and the largest yearly frequency during 2015, 2016, and 2018. Additionally, the estimated return values highlight a higher increase in daily peak demand for Rajasthan, Delhi, Madhya Pradesh, and Maharashtra compared to other states during the next 25 years. These findings will be pertinent to the decision-makers in planning for peak reserve capacity in various Indian states.