A comprehensive Python library for downloading Midcontinent Independent System Operator (MISO) public reports into pandas dataframes.

Machine learning and regression models for day-ahead and balancing market trading of wind power using real-world data from Bornholm wind farms.

Forecasts of the power market prices before the day-ahead market closure time are prone to significant error. I predict whether the day ahead (PTF in Turkish) or balancing (standard marginal price (SMF in Turkish)) will be higher at a given hour next day. It creates decisive forecasts for reducing the balancing cost.

This program optimizes the operation and bidding strategy of renewable-based Virtual Power Plants (RVPPs) under different sources of uncertainty using MILP-based flexible robust optimization approaches. The model considers RVPP participation in the Day-Ahead, Secondary Reserve, and Intra-Day Iberian electricity markets.

Case study results for the paper "Moving from linear to conic markets for electricity"

Battery sizing under dispatch fidelity and imbalance-penalty regimes (DK1 + ERCOT, 2021-2023)

Co-optimization framework for Battery Energy Storage Systems between day-ahead energy arbitrage and balancing reserves. Builds a unified optimization model and evaluates revenue of multiple strategies using French market data (RTE/Ember).

ERCOT BESS arbitrage: LightGBM dispatch adds +$2.6M/yr (100 MW battery) over persistence on 2-year held-out test (77% of ceiling).

Day-ahead GB electricity price forecasting: weather → renewables → price pipeline with ML and deep learning (BENV0148 UCL)

SQL analytical toolkit for Turkish electricity market data. Demonstrates window functions, CTEs, rolling aggregates — DuckDB + Python + Jupyter.

Risk-adjusted 10-year TCO framework for hyperscale data center siting (5 U.S. markets) — combines AEO + Cambium forward prices, calibrated XGBoost stress events, Monte Carlo NPV with CVaR-95, and 6-case rank-robustness sensitivity.

End-to-end forecasting pipeline for Turkish Day-Ahead electricity prices. Demonstrates time-series feature engineering and proper rolling-window backtesting.

ML pipeline for predicting Romanian grid imbalance and capturing intraday/imbalance arbitrage spread

Downloading and storing the TenneT aFRR bid ladders at regular intervals

Interactive map dashboard visualizing redispatch data for Germany, 2024

Causal ML for Swiss aFRR price spike prediction — German wind forecast errors → unplanned loop flows → balancing cost spikes. XGBoost + SHAP + MLflow + Databricks + Airflow pipeline with PSI drift monitoring and champion/challenger retraining.

Real-time LMP price forecasting dashboard for Nine Mile Point nuclear node — PJM Data Miner API, SQLite, scikit-learn, Streamlit

Near-Orthogonal Regimes in ISO New England: Price Level and Temporal Persistence as Non-Redundant Dimensions