AI-based Peak Power Demand Forecasting Model focusing Economic and Climate features Provisionally Accepted

Abdul Aziz¹ Danish Mahmood¹ Muhammad Shuaib Qureshi^2* Muhammad Bilal Qureshi³ Kyungsup Kim²

• ¹Shaheed Zulfikar Ali Bhutto Institute of Science and Technology, Pakistan
• ²Chungnam National University, Republic of Korea
• ³University of Lakki Marwat, Pakistan

Economy of a country is directly proportional to power sector of that country. Unmanaged power sector causes instability in county. Pakistan is also facing this phenomenon due to uncontrolled power outage and circular debt. After literature review and discussion with domain experts, it is found that inaccurate power demand forecast is one of the main reasons of power crisis in Pakistan. Previously traditional statistical methods were used for power demand forecasting. Multiple linear regression model is being used since 2018 (IGCEP) using features like previous year load, demographic and economic variables for long term peak power demand forecasting till 2030. Moreover, even yearly peak power demand is not absolutely linear in nature, hence need to apply AI-based techniques that can handle non-linearity effectively. Not using system generated data, not using most appropriate features, not using appropriate forecasting time horizon and not using appropriate forecasting model are main reasons of inaccurate peak power demand forecasting. The issue can be resolved by forecasting monthly peak power demand for next five years by using NPCC's system generated data. Accurate monthly peak load forecasting leads to accurate yearly peak power demand. More accurate monthly peak power demand forecasting can be achieved by apply non-linear AI models on a comprehensive dataset comprises of new engineered features, climate features and number of consumers. All these features are mostly system generated and cannot be manipulated. This is why accuracy is improved and results are more reliable than existing models. The new features can be engineered from recent monthly peak load data generated by system operator (NPCC). Climate features are collected from Meteorological department of Pakistan through sensors or database connectivity. Number of electricity consumers can be extracted from NEPRA's Stateof-Industry report. All three datasets are combined on a common key (month-year) to a comprehensive dataset which is pass through different AL models. In experimental setup it is found that SVR produces most accurate results with R-Square of 99%, RMSE of 28 and MAPE of 0.1355 which are best results as compared to literature reviewed.