Real live demonstration of MPC for a power-to-gas plant

This paper presents the results and the way towards them for a field trial of a 120 kWei PEM electrolyser located within the city boundaries of Freiburg, Germany. The plant is equipped with on-site hydrogen storage and connected to the local gas and electricity network. Restrictions in gas feed-in and allowed peak electricity demand apply, and limit operation of the unit. Model predictive controls have been identified in simulations as a promising control strategy for such a case. However to move from simulation into practice few things, presented in this work, should be considered. A linear model predictive controller is successfully used in a field trial to optimise the operation of the power to-gas unit in the presence of network restrictions and time varying electric prices. The impact of imperfect forecasts as well as deviations between the optimisation model and the real unit on controller performance are discussed. In the final part of the paper the lessons learned, improvements and potential solutions for integrating power-to-gas units into urban energy systems are presented. Additionally a modular controller design framework is introduced, which allows a rapid control development by interchanging historic time-series data with realtime values and simulation models with real components. In a separate testing sequence the power-to-gas unit is characterised regarding its dynamic properties, showing its potential for fast response, but also limitations in ramp-rates, tracking accuracy and losses due to warm stand-by.