/Data-driven-Methods-for-the-Reduction-of-Energy-Consumption-in-Warehouses-Use-Case

This is the repository that includes the code of the use case in the paper titled "Data-driven Methods for the Reduction of Energy Consumption in Warehouses: Use-Case Driven Analysis"

Primary LanguageJupyter Notebook

Data-driven-Methods-for-the-Reduction-of-Energy-Consumption-in-Warehouses-Use-Case

This is the repository that includes the code of the use case in the paper titled "Data-driven Methods for the Reduction of Energy Consumption in Warehouses: Use-Case Driven Analysis". All the code documentation and methods' definitions are similar to the content of the manuscript published in Elsevier Internet of Things: I. Shaer and A. Shami, Data-driven methods for the reduction of energy consumption in warehouses: Use-case driven analysis, Internet of Things (2023), doi: https://doi.org/10.1016/j.iot.2023.100882.

Before running the code, the following steps need to be followed:

  1. Create the datasets and results folders in the root directory (on the same levels as notebooks).
  2. Create the stats and figures directories in the results directory and the generated directory in the datasets directory.
  3. Download the dataset employed for this work, which can be found using this link: https://figshare.com/articles/dataset/CU-BEMS_Smart_Building_Electricity_Consumption_and_Indoor_Environmental_Sensor_Datasets/11726517. This link is included in the data availability subsection in the original manuscript. The csv files obtained should be extracted in the datasets folder. The path to any of these datasets should follow this convention: /datasets/{dataset_name}.

The explanation of each of the notebooks in the notebook directory is as follows:

  • preprocessing.ipynb: The code splits the single-floor data into its constituent zones. Additionally, the data points are defined in frames of 60 continuous minutes. From this point onwards, the analysis is conducted on zone 2 of the 6th floor.
  • 1D-CNN.ipynb: This notebook executes the code that applies the a1 method in the manuscript. It includes the feature construction process required to generate the inputs used by 1D-CNN models. The rest of the code applies the hyper-parameter optimization (HPO) process of 1D-CNN.
  • a2_method.ipynb: This notebook executes the features engineering and HPO for the a2 method, as per the process explained in the manuscript.
  • a3_method.ipynb: This notebook executes the features engineering and HPO for the a3 method, as per the process explained in the manuscript. The results for a1, a2, and a3 methods are generated in the results/stats directory. The results of these methods are reported in Table 5 in the original manuscript.
  • Best_methods_predictions.ipynb: This notebook includes the code that executes a1, a2, and a3 methods with the best hyper-parameters. The results depicted in Figures 4 and 5 in the paper are obtained using this notebook.
  • Results_Analysis.ipynb: This notebook includes the code that generates Figures 4 and 5 in the original manuscript.
  • Prediction_Analysis.ipynb: This notebook includes the code that generates the results in Table 6.

Contact-Info

Please feel free to contact me for any questions or research opportunities.