Development of regression model for electrical load forecasting during the COVID-19 pandemic lockdown

Abstract

Yeni koronavirüs (COVID-19) salgını, dünya çapında kamu hizmetleri ve şebeke operatörleri için benzeri görülmemiş zorluklar yarattı. Bu tezde, yük tahmini problemine odaklanıyoruz. Katı sosyal mesafe kısıtlamaları nedeniyle, dünya çapındaki güç tüketimi profilleri hem büyüklük hem de günlük kalıplar açısından değişti. Bu değişiklikler, kısa vadeli yük tahmininde önemli zorluklara neden olmuştur. Algoritmalar tipik olarak hava durumunu, zamanlama bilgilerini ve önceki tüketim seviyelerini girdi değişkenleri olarak kullanır, ancak pandemi sırasında sosyoekonomik davranıştaki büyük ve ani değişiklikleri yakalayamazlar. Bu tezde, tahmin algoritmalarının mevcut yapı taşlarını tamamlamak için ekonomik faaliyetlerin bir ölçüsü olarak bir regresyon modeli sunuyoruz. Böyle bir veri kümesiyle ilgili en büyük zorluk, son pandemi ile yalnızca sınırlı hareketlilik kayıtlarının ilişkili olmasıdır.

The coronavirus (COVID-19) outbreak has presented unprecedented challenges to utilities and grid administrators worldwide. One area of particular concern is load forecasting, as the pandemic has caused significant changes in the magnitude and daily power consumption patterns due to strict social distancing measures. These changes have introduced complexities in accurately predicting short-term electricity demand. Traditionally, load forecasting algorithms rely on weather conditions, timing information, and historical consumption levels to make predictions. However, these algorithms struggle to capture the large and abrupt shifts in socioeconomic behavior during the pandemic. This limitation is due to the lack of available data on mobility, a crucial factor in understanding economic activity during this period. To address this challenge, this dissertation proposes introducing a regression model as an additional measure of economic activity. By incorporating this model into existing forecasting algorithms, we aim to enhance their accuracy and robustness in capturing the effects of the pandemic on electricity demand. However, one of the main obstacles in utilizing such a dataset is the scarcity of mobility records specifically associated with the recent pandemic. By exploring alternative data sources and developing innovative approaches, this dissertation seeks to overcome the limitations posed by the lack of mobility records. The goal is to improve load management forecasting capabilities during significant socioeconomic shifts, such as the ongoing pandemic. This research aims to contribute to advancing load forecasting methodologies and provide valuable insights for utilities and grid administrators grappling with the challenges posed by the COVID-19 pandemic.