Temperature dependent surrogate modelling for MILP based dispatch and waste heat recovery of alkaline electrolysers: A comparative assessment

Alkaline water electrolysis is central to decarbonising hard-to-abate sectors, and its economics benefit from exploiting both dispatch flexibility and waste-heat recovery. Mixed-integer linear programming (MILP) is widely used for electrolyser dispatch optimisation, yet many formulations assume constant efficiency and neglect temperature-dependent part-load behaviour. This paper presents an MILP-compatible framework, derived from a semi-empirical electrochemical model, that captures the dependence of performance on current density and temperature. Three surrogates of increasing fidelity, a temperature-constant piecewise-linear (PWL) model, a tangent-plane (TAN) approximation, and a novel triangulated convex-combination (TRI) approach preserving the full current density-temperature surface, are compared in a generic case study maximising operational profit from hydrogen sales, electricity export, and heat utilisation under dynamic thermal constraints. Annual hydrogen production and operational profit are moderately affected by surrogate choice, whereas recoverable waste heat differs by up to 30% across formulations: simpler surrogates misrepresent thermal behaviour under part-load and dynamic conditions. PWL remains adequate for hydrogen-oriented dispatch, TAN offers a pragmatic compromise where temperature is required as a state variable, and TRI provides the most informative basis where waste-heat recovery and part-load consistency are the focus.