An Open, Validated Design and Performance-Prediction Methodology for Ammonia Water Kalina-Cycle Recovery of Sub-150 °C Industrial Waste Heat

Abstract

This study develops an open, verification-centered design and performance-prediction methodology for recovering sub-150 °C industrial waste heat with ammonia–water Kalina-cycle systems. The methodology is deliberately separated into source characterization, ammonia–water property verification, component-level mass/energy/exergy modeling, constrained cycle design, multi-variable optimization, surrogate performance prediction, uncertainty propagation, and reproducibility controls. The framework is motivated by the large but thermodynamically difficult inventory of low-temperature industrial heat and by the temperature-glide advantage of non-azeotropic ammonia–water working fluids. A reference implementation is defined around a configurable KCS-11-like architecture and an open workflow in which a process simulator such as DWSIM may host the flowsheet while Python-based scripts execute validation tests, optimization, surrogate fitting, and audit-ready reporting. A literature-anchored transparent reduced-order model is used only to demonstrate the workflow; it is not represented as a replacement for a validated ammonia–water equation of state. A 2,400-point synthetic design space spanning 80–150 °C source temperature, 0.55–0.95 ammonia mass fraction, 1.5–5.0 MPa high-side pressure, 3–10 K pinch, component efficiencies, sink temperature, recuperator effectiveness, and source duty is evaluated. The demonstration reproduces three published low-temperature benchmark efficiencies with a mean relative discrepancy of 2.73%, and a random-forest surrogate predicts the transparent model’s net power with R² = 0.9825 and RMSE = 15.3 kW. Optimization shows that the best ammonia concentration and high-side pressure move with source temperature rather than remaining universal constants. Monte Carlo propagation further shows why design output should be reported as an uncertainty distribution rather than a single deterministic value. The principal contribution is therefore not a new proprietary Kalina correlation, but a transparent validation hierarchy and reproducible design protocol that can be implemented with qualified thermodynamic property routines and then extended to industrial project data.