Published:
A conceptual architecture and a working prototype for an LLM-based agent that advises on battery dispatch: day-ahead price forecasting with LightGBM, BESS sizing and scheduling in Pyomo, and a provider-agnostic ReAct tool-use agent on top.
Published:
A conceptual architecture and a working prototype for an LLM-based agent that advises on battery dispatch: day-ahead price forecasting with LightGBM, BESS sizing and scheduling in Pyomo, and a provider-agnostic ReAct tool-use agent on top.
Published:
Study notes on market structures, pricing mechanisms, day-ahead clearing, intraday trading, balancing markets, and offering strategies.
Published:
Building a custom Gym environment for smart heat pump control with thermal storage using Stockholm temperature data and ENTSO-E day-ahead electricity prices.
Published:
An introduction to heat pump technologies, thermodynamic cycles, coefficient of performance (COP), and building physical-informed digital twins.
Published:
A conceptual architecture and a working prototype for an LLM-based agent that advises on battery dispatch: day-ahead price forecasting with LightGBM, BESS sizing and scheduling in Pyomo, and a provider-agnostic ReAct tool-use agent on top.
Published:
Introduction to electricity market design, demand/supply curves, marginal utility, equilibrium pricing, and social surplus.
Published:
Study notes on market structures, pricing mechanisms, day-ahead clearing, intraday trading, balancing markets, and offering strategies.
Published:
A working two-stage stochastic battery dispatch model, solved in Pyomo, with scenarios built from real quantile forecasts instead of hand-picked historical days. Compared against a naive single-forecast schedule and a perfect-foresight upper bound on a real DK1 test day.
Published:
From deterministic to stochastic decision-making: two-stage and multi-stage stochastic programs, common solution approaches, and why battery storage scheduling needs to account for uncertain prices, demand, and generation.
Published:
Bootstrapped residuals for DK1 day-ahead prices: in-sample vs out-of-sample residuals, binning by predicted value, and multiple interval levels, adapted from skforecast’s bootstrapped-residuals guide and evaluated on real DK1 data.
Published:
Implementing Quantile Regression and Quantile Regression Forest for DK1 day-ahead prices, evaluated with walk-forward cross-validation: real code, real results, and what changes once a single train/test split becomes four folds spread across different seasons.
Published:
An introduction to probabilistic electricity price forecasting: the European and Danish power markets, quantiles and prediction intervals, Quantile Regression and Quantile Regression Forest, and evaluation metrics such as Pinball Loss, CRPS, and the PIT histogram.
Published:
A conceptual architecture and a working prototype for an LLM-based agent that advises on battery dispatch: day-ahead price forecasting with LightGBM, BESS sizing and scheduling in Pyomo, and a provider-agnostic ReAct tool-use agent on top.
Published:
Transitioning from point forecasts to probabilistic predictions. Comparing Quantile Regression (QR), Quantile Regression Forest (QRF), and Bootstrapped Residuals to estimate uncertainty in wave height and mean wave period.
Published:
Stacking Ridge Regression, Random Forest, and LightGBM with a ridge meta-model to forecast significant wave height and mean wave period, and checking whether it recovers the long-lead skill that pure LightGBM lost in Part 1.
Published:
Forecasting significant wave height and mean wave period 1 to 48 hours ahead with LightGBM on ERA5 ocean reanalysis data off northern Portugal, then combining both into a wave power flux estimate.
Published:
Bootstrapped residuals for DK1 day-ahead prices: in-sample vs out-of-sample residuals, binning by predicted value, and multiple interval levels, adapted from skforecast’s bootstrapped-residuals guide and evaluated on real DK1 data.
Published:
Implementing Quantile Regression and Quantile Regression Forest for DK1 day-ahead prices, evaluated with walk-forward cross-validation: real code, real results, and what changes once a single train/test split becomes four folds spread across different seasons.
Published:
An introduction to probabilistic electricity price forecasting: the European and Danish power markets, quantiles and prediction intervals, Quantile Regression and Quantile Regression Forest, and evaluation metrics such as Pinball Loss, CRPS, and the PIT histogram.
Published:
A working two-stage stochastic battery dispatch model, solved in Pyomo, with scenarios built from real quantile forecasts instead of hand-picked historical days. Compared against a naive single-forecast schedule and a perfect-foresight upper bound on a real DK1 test day.
Published:
From deterministic to stochastic decision-making: two-stage and multi-stage stochastic programs, common solution approaches, and why battery storage scheduling needs to account for uncertain prices, demand, and generation.
Published:
Building a custom Gym environment for smart heat pump control with thermal storage using Stockholm temperature data and ENTSO-E day-ahead electricity prices.
Published:
An introduction to heat pump technologies, thermodynamic cycles, coefficient of performance (COP), and building physical-informed digital twins.
Published:
Transitioning from point forecasts to probabilistic predictions. Comparing Quantile Regression (QR), Quantile Regression Forest (QRF), and Bootstrapped Residuals to estimate uncertainty in wave height and mean wave period.
Published:
Bootstrapped residuals for DK1 day-ahead prices: in-sample vs out-of-sample residuals, binning by predicted value, and multiple interval levels, adapted from skforecast’s bootstrapped-residuals guide and evaluated on real DK1 data.
Published:
Implementing Quantile Regression and Quantile Regression Forest for DK1 day-ahead prices, evaluated with walk-forward cross-validation: real code, real results, and what changes once a single train/test split becomes four folds spread across different seasons.
Published:
An introduction to probabilistic electricity price forecasting: the European and Danish power markets, quantiles and prediction intervals, Quantile Regression and Quantile Regression Forest, and evaluation metrics such as Pinball Loss, CRPS, and the PIT histogram.
Published:
Building a custom Gym environment for smart heat pump control with thermal storage using Stockholm temperature data and ENTSO-E day-ahead electricity prices.
Published:
A working two-stage stochastic battery dispatch model, solved in Pyomo, with scenarios built from real quantile forecasts instead of hand-picked historical days. Compared against a naive single-forecast schedule and a perfect-foresight upper bound on a real DK1 test day.
Published:
From deterministic to stochastic decision-making: two-stage and multi-stage stochastic programs, common solution approaches, and why battery storage scheduling needs to account for uncertain prices, demand, and generation.
Published:
Transitioning from point forecasts to probabilistic predictions. Comparing Quantile Regression (QR), Quantile Regression Forest (QRF), and Bootstrapped Residuals to estimate uncertainty in wave height and mean wave period.
Published:
Stacking Ridge Regression, Random Forest, and LightGBM with a ridge meta-model to forecast significant wave height and mean wave period, and checking whether it recovers the long-lead skill that pure LightGBM lost in Part 1.
Published:
Forecasting significant wave height and mean wave period 1 to 48 hours ahead with LightGBM on ERA5 ocean reanalysis data off northern Portugal, then combining both into a wave power flux estimate.
Published:
Why ocean waves are a large, steady, low-emission energy resource, the physics that governs them, the eight main types of wave energy converter, and how wave conditions are actually measured.