/forecasting-energy-consumption-LSTM

Development of a machine learning application for IoT platform to predict electric energy consumption in smart building environment in real time.

Primary LanguagePythonMIT LicenseMIT

forecasting-energy-consumption-LSTM

Development of a machine learning application for IoT platform to predict energy consumption in smart building environment in real time.

Development Platform

The project was built with google colab, which uses python jupyter notebook. The model and the performance scripts were in the same project. After, the training of the model we can use the performance.py code block for evaluation, because the trained variables and the model were saved internal in the environment.

Data Acquisition

The dataset, that was used for the development of the machine learning models, was taken from: https://www.kaggle.com/uciml/electric-power-consumption-data-set

Data Preprocessing

  • Handling missing values.
  • Data Smoothing (exponential smoothing).
  • Handling outliers (we detected them using standard deviation).
  • Data normalization (scaling the values between [0,1]).
  • Data resampling ().

Splitting the Dataset.

  • Training set.
  • Validation set.
  • Test set.

First Approach (LSTM).

We made use of Long Short-Term Memory (LSTM) cells to create a sequential model using the Keras API.

Second Approach (Seq2Seq).

We implemented a Sequence-to-Sequence model utilizing the Keras' functional API.

  • Results of prediction for the next day (1h to 24h).

    1. Case alt text
    2. Case alt text
    3. Case alt text
    4. Case alt text
    5. Case alt text

We took random prediction cases from the whole test set to examine the performance of our model visually.

License

Copyright © 2019 Christos Chousiadas

This repository is under the MIT License.