Thermodynamics-informed prefiltering framework for industrial heat pump integration: carnot-based targeting and adaptive genetic algorithm optimization

Industrial heat-pump integration requires exploring competing temperature levels, heat allocations and utility choices, while detailed superstructure models make scenario exploration computationally demanding. This study develops a thermoeconomic prefiltering framework to identify integration targets before detailed design. The framework couples an adaptive genetic algorithm (AGA) with heat allocation along residual grand composite curves to select source and sink temperature levels minimizing total annual cost (TAC). Heat-pump performance uses the Carnot coefficient of performance scaled by a second-law efficiency, nominally 0.50 and varied from 0.30 to 0.80. Boiler competition permits partial electrification, while rejected heat from fixed refrigeration services provides an additional source. Three Swiss industrial profiles, covering paper, meat and cheese production, are assessed through five-seed AGA sensitivity analyses of energy prices, operating hours, efficiency and segmentation costs. Across the investigated cases, relative energy prices and heat-pump second-law efficiency strongly influence heating coverage, whereas higher absolute prices can favor additional temperature levels. In meat processing, higher absolute prices result in six instead of three condenser levels at a fixed electricity-to-gas price ratio, while the heat-pump share remains about 96.2%. Near transitions to higher HP coverage, the influence of second-law efficiency depends on the assumed segmentation cost, which penalizes additional temperature levels. For the paper-mill reference, the AGA achieves TAC within 0.0026% of the best deterministic result under the same screening model, using 21,132 versus 2.87 million evaluations. Comparisons with detailed superstructure models identify promising thermal regions without establishing equipment-level feasibility or economic accuracy. The framework enables rapid prioritization before detailed modeling.