UK-Electricity-Price-Forecasting

UK day-ahead electricity price prediction using weather-driven renewables forecasting + ML (XGBoost/SVR), including spike detection + battery arbitrage evaluation.

UK Electricity Price Forecasting (Weather → Renewables → Price)

This repository contains the final runnable pipeline (Sections 4–6) of our group project on forecasting UK day-ahead electricity prices using machine learning, benchmarking against time-series baselines, and assessing decision value via a battery arbitrage simulation.

Note: The full Stage 1 renewable generation modelling (weather → wind/PV training) is not included in this public repo.
This repo includes the processed datasets + predicted wind/PV inputs required to run the price modelling end-to-end from Section 4 onward.

---

Key results (test set)

• Price forecasting (XGBoost + SVR): R² ≈ 0.88, RMSE ≈ 12.49 EUR/MWh, MAE ≈ 8.73 EUR/MWh
• Extreme price spike detection (95th percentile): F1 ≈ 0.59
• Battery arbitrage backtest (100 MW / 200 MWh): ~€4.016M, capturing ~90.7% of perfect-foresight upper bound

Model Performance	Forecast & Backtest

---

Repository contents

Main notebook

• UK_Electricity_Price_Prediction_Complete.ipynb
  ✅ Runnable from: SECTION 4: ELECTRICITY PRICE PREDICTION onward.

Included datasets (Sections 4–6)

Located in: data/step2_prices/

• UK_WholesalePrices_Hourly.csv
• demanddata_2021-2025.csv
• SAP.xlsx (gas prices)
• Carbon Emissions Futures Historical Data UK.csv
• Final_Price_Predictions_Ensemble.csv
• wind_100Locs_XGBoost_WalkForward_Clean_PREDICTED-weather-data.csv (predicted wind input)
• xgboost_cv_predictions.csv (predicted solar/PV input)

---

What the notebook does (Sections 4–6)

• Builds a supervised learning dataset combining demand + fuel/carbon drivers + predicted renewables (wind/PV)
• Trains and evaluates XGBoost and SVR models for day-ahead price prediction
• Benchmarks performance against an ARIMA baseline model
• Simulates battery arbitrage decisions using predicted prices to quantify decision value

---