Predicting heat demand of residential buildings with lag and time variables
Open Access
- 1 January 2024
- conference paper
- Published by Centre for Evaluation in Education and Science (CEON/CEES)
- No. Automatics,p. 257-263
Abstract
Building heat demand depends on various parameters: the properties of the materials, geometry, climate conditions, occupancy patterns, usage habits, etc. Short-term forecasts of the heat demand can be based on different subsets of these parameters, depending on the application and available data. Recently, black-box models that apply machine learning methods became widely used to predict the heat demand. This paper is a part of a broader research effort and investigates the time-series prediction properties of a simple model based on the Random Forest regression that uses only lag and time-related variables as inputs. The time resolution is one hour. The lag variables are the heat demand one hour before, 24 hours before, and 25 hours before the time of prediction. The time-related variables are the hour of day, day of week, and month. The model has the coefficient of determination of over 0.99, root-mean-square error 38.2 kWh, and mean absolute error 19.1 kWh. The most important predictor is the 24-hours lag. Large errors occur mainly early in the morning and late in the evening, when the heat demand changes have high values and the setpoint temperatures are being modified.Keywords
This publication has 14 references indexed in Scilit:
- Exploiting building information modeling and machine learning for optimizing rooftop photovoltaic systemsEnergy and Buildings, 2024
- Surrogate modeling for long-term and high-resolution prediction of building thermal load with a metric-optimized KNN algorithmEnergy and Built Environment, 2023
- Modeling energy-efficient building loads using machine-learning algorithms for the design phaseEnergy and Buildings, 2023
- A novel method of creating machine learning-based time series meta-models for building energy analysisEnergy and Buildings, 2022
- Similarity-based grouping method for evaluation and optimization of dataset structure in machine-learning based short-term building cooling load prediction without measurable occupancy informationApplied Energy, 2022
- Machine learning-based cooling load prediction and optimal control for mechanical ventilative cooling in high-rise buildingsEnergy and Buildings, 2021
- A novel optimization method combining metaheuristics and machine learning for daily optimal operations in building energy and storage systemsApplied Energy, 2021
- Machine learning applications in urban building energy performance forecasting: A systematic reviewRenewable and Sustainable Energy Reviews, 2020
- Time series analytics using sliding window metaheuristic optimization-based machine learning system for identifying building energy consumption patternsApplied Energy, 2016
- Random ForestsMachine Learning, 2001