Predict the grid. Power what's next.
An end-to-end platform for electricity load forecasting. Upload historical demand, train LSTM, gradient-boosted, and seasonal models, and benchmark them side-by-side — all in the browser.
Operations-grade forecasting, without the ops.
Three models, one click
LSTM (Elman RNN with BPTT), gradient-boosted regression on engineered lags, and Holt-Winters seasonal — train and benchmark all three.
Bring your own data
Drag-and-drop CSV with datetime + load (and optional temperature). Missing values are interpolated automatically.
Walk-forward validation
Time-respecting train/test split with MAE, RMSE, and MAPE — the metrics utilities actually trust.
24-hour & 7-day horizons
Point forecasts with 95% confidence intervals, residual diagnostics, and seasonal pattern decomposition.
Runs in your browser
All training and inference happen client-side. Your operational load data never leaves the device.
Export anywhere
Download forecasts as CSV for downstream dispatch, balancing, or reporting workflows.
Start with the sample dataset.
60 days of synthetic hourly load with realistic daily, weekly, and seasonal cycles — ready to train against.