pyaf/load_forecasting

Forecasting electric power load of Delhi using ARIMA, RNN, LSTM, and GRU models

Electric Load Forecasting

Under graduate project on short term electric load forecasting. Data was taken from State Load Despatch Center, Delhi website and multiple time series algorithms were implemented during the course of the project.

Models implemented:

models folder contains all the algorithms/models implemented during the course of the project:

• Feed forward Neural Network FFNN.ipynb
• Simple Moving Average SMA.ipynb
• Weighted Moving Average WMA.ipynb
• Simple Exponential Smoothing SES.ipynb
• Holts Winters HW.ipynb
• Autoregressive Integrated Moving Average ARIMA.ipynb
• Recurrent Neural Networks RNN.ipynb
• Long Short Term Memory cells LSTM.ipynb
• Gated Recurrent Unit cells GRU.ipynb

scripts:

• aws_arima.py fits ARIMA model on last one month's data and forecasts load for each day.
• aws_rnn.py fits RNN, LSTM, GRU on last 2 month's data and forecasts load for each day.
• aws_smoothing.py fits SES, SMA, WMA on last one month's data and forecasts load for each day.
• aws.py a scheduler to run all above three scripts everyday 00:30 IST.
• pdq_search.py for grid search of hyperparameters of ARIMA model on last one month's data.
• load_scrap.py scraps day wise load data of Delhi from SLDC site and stores it in csv format.
• wheather_scrap.py scraps day wise whether data of Delhi from wunderground site and stores it in csv format.

server folder contains django webserver code, developed to show the implemented algorithms and compare their performance. All the implemented algorithms are being used to forecast today's Delhi electricity load here [now deprecated]. Project report can be found in Report folder.

Team Members:

• Ayush Kumar Goyal
• Boragapu Sunil Kumar
• Srimukha Paturi
• Rishabh Agrahari

load_forecasting's Issues

Data not provided

Kindly provide the dataset used for this work. (.csv) files used in this work are not present here.

I can't open the website. And can you share your data with me?

The web site "http://forecast.energyandsystems.com" could not be loaded.
Besides, I don't know how to get the data from "http://www.delhisldc.org". Could you please share it with if it's ok for you.
Thanks in advance!

data-provide

Where is the original data set obtained

data not found

A guess for your training and testing dataset

Hello, thanks for your sharing. Your code is really helpful but I can't fully understand because I can't see your dataset.

(https://github.com/pyaf/load_forecasting/tree/master/models)
In "Training" part:
I don't understand the 4-th line code:
model.add(SimpleRNN(1, batch_input_shape=(batch_size, train_x.shape[1], train_x.shape[2])))

Could you please tell me how is your training and testing dataset like?

I guess, may be like this:
train_x : (number of total samples, 1, number of feature of each sample)
train_y : (number of total samples, 1, number of label of each sample)
Is that true?

I'd appreciate it if you can help me. Thank you very much.

whether_data page not found

It shows 'Error 404 Page Not Found' when I request weather data with whether_scrap.py