https://share.streamlit.io/marekwadinger/semestral_project/forecast_streamlit.py
This project is integration of loadforecast an automatic time series forecasting procedure of an electrical load based on prophet and pyod anomaly detection toolkit into user-friendly web app environment of streamlit.
It serves for automation of whole procedure of electrical load forecasting, making it fast and easy to use for both research workers and users with limited programming skills.
Quick way to get your hand on the load forecasting service is to use our Streamlit app to play around with example data.
As a quick example, we will run the online service use historical load data to fit new model and get prediction
- To start the service, open Streamlit app
- Import the historical load data
- The service silently fits your model
- Choose prediction period to get your prediction
- Visualize it pressing "Plot forecast"
- Download the prediction as JSON pressing "Download json file" or CSV pressing "Download csv file"
- Download the model pressing "Download model" for later use
Note: If the app is sleeping, you can build it by pressing the big red button on screen.
Now we'd like to make prediction while new historical data are available. We have our previously fitted model at our hands
- Import the extended historical load data
- Import the model
- The service silently checks whether the model is outdated and refits it when needed
- Get your prediction
- Download the updated model pressing "Download model"
The side panel allows you to customize your prediction model:
- Includes local holidays using coutry's ISO code
- Offers rich control over various time-series attributes
Feel free to contribute in any way you like, we're always open to new ideas and approaches.
- Feel welcome to open an issue if you think you've spotted a bug or a performance issue.