Indoor air quality is a key concern in building design, as occupant exposure is strongly governed by pollutant transport and removal through ventilation flows. Computational fluid dynamics (CFD) provides detailed indoor contaminant concentration fields, but its computational cost limits repeated evaluations over many pollutant source locations. In this work, we develop a Fourier Neural Operator (FNO) framework for rapid prediction and inverse localization of indoor contaminant transport, in which a soft rectangular source mask provides both geometric fidelity and differentiability. The FNO achieves a mean relative L2 field error of 0.032, predicts a full field in approximately 5.5 ms, about 15,000 times faster than the corresponding CFD simulation, and outperforms FCN, U-Net, and DeepONet baselines. Zero-shot super-resolution tests show that models trained on moderately coarse grids can be evaluated directly on finer grids with little loss of accuracy: the model trained at 192 × 64 increases the mean root-mean-square error by only 14.0% on the 384 × 128 reference grid. For inverse localization, the differentiable surrogate is combined with coarse global screening and continuous gradient-based refinement to estimate source coordinates from eight sensor observations, achieving mean localization errors of 0.103, 0.165, and 0.372 m at relative noise levels of 0, 0.02, and 0.05, the lowest among the three evaluated inverse strategies. These results demonstrate that differentiable neural-operator surrogates can support rapid field prediction and reliable inverse source localization for indoor contaminant transport
Industrial gas turbines play a vital role in modern power generation systems, producing flexible and efficient electricity that supports grid stability, especially in systems with increasing share of renewable energy. Accurate and interpretable prediction of gas turbine power output is thus of critical importance for performance optimization, operational reliability, and economical energy management. This study proposes a datadriven methodology for predicting gas turbine power output using longterm operating measurements of an industrial gas turbine in a combinedcycle power plant. Four regression-based machine learning models, Linear Regression, Random Forest, XGBoost and an Artificial Neural Network are trained and evaluated on two datasets, the original mixedregime dataset and a preprocessed dataset where lowload and shutdown regimes are removed to isolate steadystate operation. The coefficient of determination (R²), mean squared error (MSE), root mean square error (RMSE) and mean absolute error (MAE) were used to measure the model performance. The original dataset allows XGBoost to achieve an R² value of 0.996, an RMSE of 3.59 MW and an MAE of 1.28 MW. The preprocessing improves these metrics, reaching an R² value of 0.9964, an RMSE of 1.46 MW and an MAE of 1.11 MW, which corresponds to an RMSE decrease of around 59% and an MAE decrease of approximately 13%. Feature importance and SHAP analyses reveal that temperaturerelated variables dominate under mixed operating conditions, whereas Compressor Exit Pressure becomes the primary predictor under steady state conditions. These results show that the combination of thermodynamic interpretation, machine learning and SHAP analysis may successfully support the prediction of gas turbine power output in diverse operating conditions.