Multiple Regression Model Based on Weather Factors for Predicting The Heat Load of A District Heating System in Dalian, China—A Case Study

Abstract

The managers need to have a reasonable guide on the operation and management in district heating system (DHS), so it's very necessary to predicting the heat load for DHS. In this paper, the relationships between the heat load and weather conditions have been researched in order to determine the inputs variables and output variable of the future heat load prediction model. Using the given data from the obtained database, the multiple regression modelling and analysis method was carried out so as to establish the corresponding heat load prediction models for the DHS. The results shown that the square correlation coefficient between the heat load's measured value and estimated value are all greater than 0.9000, and the mean absolute percentage error (MAPE) between the heat load's measured value and estimated value are all less than 4.00%. Moreover, the corresponding maximum absoult relative errors between the heat load's measured value and estimated value are all less than 8%. The results also indicated that the heat load prediction model's accuracy is relatively high. Furthermore, these 5 heat load prediction models can be applied in the real DHS and this multiple regression method can be promoted into the other engineering field.This is an open access article licensed under the terms of the (https://creativecommons.org/licenses/by/4.0/legalcode), which permits unrestricted, noncommercial use, distribution and reproduction in any medium, provided the work is properly cited.

Development of Rice Yield Forecasting Model Using Linear Regression for Imphal West District, Manipur, India

“…"Multiple linear regression (MLR) is the standard and simplest approach for the development of calibration models. Feature selection in the form of stepwise MLR (SMLR) gives good results over large datasets [7,8,9]. A stepwise regression procedure was adopted for the selection of the best regression variable among many independent variables and found that models were able to explain 51 to 79% variability for rice yield.…”

Geographically Weighted Machine Learning and Downscaling for High-Resolution Spatiotemporal Estimations of Wind Speed

“…At a high temporal resolution (e.g., 3 h), the reanalysis data are still at a coarse spatial resolution [e.g., 0.25 • (latitude) × 0.3125 • (longitude)], and thus cannot provide local variations at a fine spatial scale for wind speed, which is affected by prevailing pressure, air temperature and local site characteristics with a volatile nature. Additionally, such finely resolved variability of wind speed also might not be captured well (low accuracy) by traditional approaches, including multiple linear regression [12], nonlinear regression [13] and spatial interpolation [14,15] (e.g., inverse distance weighting or kriging) due to the sparse spatial distribution of wind speed monitoring stations, and the limited generalization of these methods compared to advanced machine learning methods, such as XGBoost and deep learning.…”