This is the original source code used for all experiments in the paper A hybrid RF-LSTM based on CEEMDAN for improving the accuracy of building energy consumption prediction

Access this Binder by clicking the blue badge above

If you use the code in this paper, please cite the paper as below:

Karijadi, I., & Chou, S. Y. (2022). A hybrid RF-LSTM based on CEEMDAN for improving the accuracy of building energy consumption prediction. Energy and Buildings, 259, 111908

Irene Karijadi, Shuo-Yan Chou

corresponding author: [email protected] (Irene Karijadi)

An accurate method for building energy consumption prediction is important for building energy management system. However, building energy consumption data often exhibits nonlinear and nonstationary patterns, which makes prediction more difficult. In this study, we proposed a hybrid method of Random Forest (RF), and Long Short-Term Memory (LSTM) based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to predict building energy consumption.

This is the framework of the proposed method

The proposed method is coded in Python 3.7.6 and the experiments were performed on Intel Core i3-8130U CPU, 2.20GHz, with a memory size of 4.00 GB. The python version is specified in runtime.txt. In order to run the experiments, a number of packages need to be installed. Here are the list of the package version that we used to run all the experiments

• EMD-signal==0.2.10
• pandas==0.25.3
• keras==2.4.3
• tensorflow>=2.0.0
• sklearn==0.22.1
• numpy==1.18.1
• matplotlib
• dataframe_image

The complete list of packages can be found in requirements.txt.

In order to run the model, the packages need to be installed first using this line of code:

pip install -r requirements.txt()

Non-python file

• Dataset - This folder includes all dataset used in this study
• Figures - This folder includes all generated figures to be used in reporting
• README.md - The README file for using this code
• License - The License file
• requirement.txt - This file contains list of packages used in this study
• runtime.txt - This file contains python version used in this study

Python Files:

• 1.Experiments for University Dormitory Building.ipynb - This notebook is the main file used to conduct the experiments for university dormitory building using parameter settings specified in Setting.py
• 2.Experiments for University Laboratory Building.ipynb - This notebook is the main file used to conduct the experiments for university laboratory building using parameter settings specified in Setting.py
• 3.Experiments for University Classroom Building.ipynb - This notebook is the main file used to conduct the experiments for university classroom building using parameter settings specified in Setting.py
• 4.Experiments for Office Building.ipynb - This notebook is the main file used to conduct the experiments for office building using parameter settings specified in Setting.py
• 5.Experiments for Primary Classroom Building.ipynb- This is the main file used to conduct the experiments for primary classroom building using parameter settings specified in Setting.py
• Plot CEEMDAN result.ipynb - This notebook contains the script to plot the Decomposition results
• Plot Prediction results using proposed hybrid RF-LSTM based CEEMDAN method.ipynb - This notebook contains the script to plot the prediction results from proposed method
• Plot Dataset.ipynb - This notebook contains the script to visualize the data
• myfunctions.py - This python script includes all functions required for building proposed method and other benchmark methods that are used in the experiments
• Setting.py - This is python script includes a number of parameters that are used throughout the functions in the directory

We used a public dataset from the Building Data Genome Project

Five different buildings are used in this study and you can download it here: Office Building, University Classroom Building, University Dormitory Building, University Laboratory Building, Primary Classroom Building

To visualize hourly energy consumption from five buildings, please running the Plot Dataset.ipynb

A number of parameters (e.g. LSTM learning rate, RF feature number, etc) are defined in the Setting.py

The code that generated results presented in table 3 and 4 Section 4.4 in the paper can be executed from these notebooks:

1. Experiments for University Dormitory Building.ipynb

2. Experiments for University Dormitory Building.ipynb

3. Experiments for University Dormitory Building.ipynb

4. Experiments for University Dormitory Building.ipynb

5. Experiments for University Dormitory Building.ipynb

notes: The number indicate the order of the files need to be run

1. Run the 1. Experiments for University Dormitory Building.ipynb By running all the cell in this notebook, it will:

   • Train and test the proposed method and other benchmark methods on University Dormitory dataset.
   • Generate Table which summarize the performance results of the proposed method and other benchmark methods on University Dormitory dataset. This output is used as part of Table 3 Section 4.4 in the paper
   • Generate Table which calculate the percentage of improvement of the proposed method and other benchmark methods on University Dormitory dataset. This output is used as part of Table 4 Section 4.4 in the paper

2. Run the 2. Experiments for University Laboratory Building.ipynb By running all the cell in this notebook, it will:

   • Train and test the proposed method and other benchmark methods on University Laboratory dataset.
   • Generate Table which summarize the performance results of the proposed method and other benchmark methods on University Laboratory dataset. This output is used as part of Table 3 Section 4.4 in the paper
   • Generate Table which calculate the percentage of improvement of the proposed method and other benchmark methods on University Laboratory dataset. This output is used as part of Table 4 Section 4.4 in the paper

3. Run the 3. Experiments for University Classroom Building.ipynb By running all the cell in this notebook, it will:

   • Train and test the proposed method and other benchmark methods on University Classroom dataset.
   • Generate Table which summarize the performance results of the proposed method and other benchmark methods on University Classroom dataset. This output is used as part of Table 3 Section 4.4 in the paper
   • Generate Table which calculate the percentage of improvement of the proposed method and other benchmark methods on University Classroom dataset. This output is used as part of Table 4 Section 4.4 in the paper

4. Run the 4. Experiments for Office Building.ipynb By running all the cell in this notebook, it will:

   • Train and test the proposed method and other benchmark methods on Office dataset.
   • Generate Table which summarize the performance results of the proposed method and other benchmark methods on Office dataset. This output is used as part of Table 3 Section 4.4 in the paper
   • Generate Table which calculate the percentage of improvement of the proposed method and other benchmark methods on Office dataset. This output is used as part of Table 4 Section 4.4 in the paper

5. Run the 5. Experiments for Primary Classroom Building.ipynb By running all the cell in this notebook, it will:

   • Train and test the proposed method and other benchmark methods on Primary Classroom dataset.
   • Generate Table which summarize the performance results of the proposed method and other benchmark methods on Primary Classroom dataset. This output is used as part of Table 3 Section 4.4 in the paper
   • Generate Table which calculate the percentage of improvement of the proposed method and other benchmark methods on Primary Classroom dataset. This output is used as part of Table 4 Section 4.4 in the paper

1. To visualize the decomposition results Plot CEEMDAN result.ipynb must be executed. The output generated from this notebook is used in the Figure 5 Section 4.4 in the paper
2. To obtain and visualize prediction results using proposed hybrid RF-LSTM based CEEMDAN method,Plot Prediction results using proposed hybrid RF-LSTM based CEEMDAN method.ipynb must be executed. The output generated from this notebook is used in the Figure 6 Section 4.4 in the paper