In this project, I have conducted research as well as implemented data analytical approach for the evaluation of electricity consumption of every household in UK based on its historical data. The annual statistical data of domestic electricity consumption of all the countries provided by government of UK is used in this project (provided in the files section).
The dataset is applied to exploratory data analysis including following steps: data collection, data cleaning, descriptive analysis and inferential statistical analysis. Also, a comparison between different machine learning models implementation for prediction analysis is performed. Furthermore, electricity profiles of houses depending on their low, medium and high electricity usage are also predicted using five different classification models including decision tree, SVM and KNN etc.
I also discussed estimation of the carbon emission from energy consumption using Defra standard conversion rate which further applied to deep neural network for CO2 emissions prediction.
The outcome of experiments revealed that Random Forest regression model predicted the average household electricity consumption with the highest accuracy of 95% as compared to other models, while for electricity profile classification Decision Tree classifier and Support Vector Machine (SVM) has shown an equal accuracy rate of 98%. The prediction structure of this project can be implemented for analysis of energy consumption of commercial sector as well. The methodology can be implemented on different datasets of other countries.
-
The provided codes works in all python versions above 3.7.0.
-
The dataset used in this project was quite large, So Google Colab Pro is used. Otherwise any Python IDEs can be used.
There are few python packages and libraries that need to be installed which are provided in the python codes.
-
- Google Colab Pro
- Python language