Tech Lead | VPP Systems Engineer
Bodil Energi
Responsible for the architecture and operation of systems enabling distributed energy flexibility and grid-support participation, including the core algorithmic control system coordinating distributed household devices in real time. Serving as the primary in-house engineer, responsible for the systems.
Designed and implemented core real-time coordination algorithms for distributed energy assets, orchestrating real-time coordination of distributed household devices using custom high-performance control algorithms for enabling ancillary services. Conducted code reviews and collaborated with the engineering team to ensure system robustness and maintainability.
Deployed the platform across households, controlling devices based on live frequency monitoring to participate in ancillary service markets. The control algorithm achieved 33% of target capacity within 500ms, 86% within 700ms, and full portfolio stability within 3 seconds. Developed a forecasting system for bid capacity availability in virtual power plant portfolios, and applied Monte Carlo Dropout methods to quantify uncertainty in neural network predictions for heat pump power consumption. Built multi-source telemetry infrastructure collecting device data across multiple brokers and APIs.
Projects
System for providing grid stability through creating virtual power plants
End-to-end platform for aggregating distributed heating assets into virtual power plant portfolios supporting participation in grid support services.
Bid forecasting for auxiliary services in virtual power plants
Built a system to forecast expected capacity availability for auxiliary services within virtual power plant portfolios.
Frequency Monitoring, Simulation & Energy Market Activation
Live monitoring of grid frequency measurements with automated portfolio state management supporting grid stabilization participation.
Monte Carlo Dropout for Uncertainty Estimation in Neural Networks
Developed a Monte Carlo Dropout method to predict heat pump power usage using neural networks with quantified uncertainty in neural network predictions, enhancing model robustness and interpretability.
Control Algorithm for Aggregated Energy Assets
Real-time system for monitoring and activating physical power units to reach capacity setpoints across portfolios — delivering 33% of target capacity within 500ms, 86% within 700ms, and achieving full stability within 3 seconds.
Live Telemetry Over Various Channels
This is a multi service setup, for collecting telemetry from various sources, such as multiple brokers and APIs.