# XGBoost energy consumption prediction based on multi-system data HVAC

**Authors:** Yunlong Li, Yiming Peng, Dengzheng Zhang, Yingan Mai, Zhengrong Ruan  
**arXiv:** [2105.09945](https://arxiv.org/abs/2105.09945)  
**Date:** 2021-05-20  
**Summary page:** https://www.summarizepaper.com/en/paper/2105.09945v1/

## Key points

- The paper focuses on energy consumption prediction of HVAC systems in public buildings
- Implementing an efficient energy consumption prediction model can contribute to effective energy-saving transformations
- The authors propose a novel approach that utilizes XGBoost and LightGBM models to predict energy consumption based on multi-system data
- This approach extracts features from large datasets using XGBoost, unlike traditional models that use small datasets
- Multiple models are trained separately and their predictions are fused with LightGBM's independent prediction results using Mean Absolute Error (MAE)
- The approach is successfully applied to a self-developed IoT platform in a university's air-conditioning system in southern China
- Results demonstrate high fitting and accuracy compared to support vector machine (SVM) models
- Accurate prediction of energy consumption can assist engineers in carrying out effective energy-saving transformations, improving device testing, and ensuring the continued health of the HVAC system
- Engineers can make informed decisions regarding energy-saving measures and optimize overall efficiency of public building systems
- The study presents an innovative approach leveraging XGBoost and LightGBM models for predicting HVAC system energy consumption
- The successful application on an IoT platform demonstrates its potential for real-world implementation and contribution to sustainable building practices.

## Summary

This paper focuses on the energy consumption prediction of HVAC systems in public buildings. The HVAC system is responsible for a significant portion of the overall energy consumption in these buildings, and implementing an efficient energy consumption prediction model can greatly contribute to effective energy-saving transformations. In this study, the authors propose a novel approach that utilizes XGBoost and LightGBM models to predict energy consumption based on multi-system data. Unlike traditional models which extract features from small datasets, this approach extracts features from large datasets using XGBoost. These features are then trained separately to obtain multiple models. The authors then fuse the predictions from these models with LightGBM's independent prediction results using Mean Absolute Error (MAE). This fusion process helps infer energy consumption-related variables more accurately. To validate their approach, the authors successfully apply their model to a self-developed Internet of Things (IoT) platform in a university's air-conditioning system in southern China. The results obtained using this approach demonstrate a high degree of fitting and accuracy compared to support vector machine (SVM) models. The accurate prediction of energy consumption provided by this model can assist engineers in carrying out effective energy-saving transformations, improving device testing, and ensuring the continued health of the HVAC system. By accurately predicting energy consumption, engineers can make informed decisions regarding energy-saving measures and optimize the overall efficiency of public building systems. Overall, this study presents an innovative approach to predicting HVAC system energy consumption by leveraging XGBoost and LightGBM models. The successful application of this model on an IoT platform demonstrates its potential for real-world implementation and its ability to contribute to sustainable building practices.