Saving Energy can help human live, using energy more efficiently can be the fastest, most cost-effective ways to save money, reduce green house gas emissions, create jobs, and meet growing energy demand [1]. And saving energy consumption can help us to reduces air and water pollution and conserves natural resources, which can help us to get healthier living environment [2].

       Therefore in this case I make a prediction using time series algorithm in deep learning to predict energy consumption to help saving the energy consumption and get healthier live.

       The dataset can be download on this link kaggle - Energy Consumption

1. Check Dataset

       In this dataset there are few missing values, I found that there are 948 missing values in column Site ID and 948 missing values in column Energy Consumption (kwh).

       Therefore I remove all misisng values and got 8316 rows and 3 columns.

2. Plot timestamp and value

1. Normalize value

2. Create window dataset

       For the model I used Bidirectional LSTM and last dense has only 1 num of filters.

       I used SGD optimizers with learning_rate = 1e-4 and moemntum = 0.9. Loss function that I used is Huber() and metrics=['mae'].

       Then I got loss: 0.0067 - mae: 0.0765

[1] https://www.epa.gov/statelocalenergy/local-energy-efficiency-benefits-and-opportunities

[2] https://www.aceee.org/sites/default/files/ee-improves-environment.pdf