Load Forecasting in Nordic Residential Buildings

Abstract

Renewable power production technologies, such as solar photovoltaics, are becoming increasingly popular in the residential sector. Due to the volatility of renewable power production, and mismatch between peak production and consumption hours, it can be beneficial to augment local power production with energy storage systems, such as batteries. Connection to mains grid also facilitates selling overproduction, especially when the price of electricity is high. However, such a system requires automated decision making, which would highly benefit from knowing the state of electrical load, price and solar power production in advance. This study focuses specifically on the task of load forecasting, applied to Nordic multifamily residential buildings, which are mainly heated by district heating. While it is common to utilize weather data for load forecasting, this work shows by properly choosing and optimizing the forecast model, omitting the weather data can in this case improve the accuracy of the forecasts.