ABSTRACT
This paper addresses a design optimization of a gas turbine (GT) for marine applications. A gain-scheduling method incorporating a meta-heuristic optimization is proposed to optimize a thermodynamics-based model of a small GT engine. A comprehensive control system consisting of a proportional integral (PI) controller with additional proportional gains, gain scheduling, and a min-max controller is developed. The modeling of gains as a function of plant variables is presented. Meta-heuristic optimizations, namely a genetic algorithm (GA) and a whale optimization algorithm (WOA), are applied to optimize the designed control system. The results show that the WOA has better performance than that of the GA, where the WOA exhibits the minimum fitness value. Compared to the unoptimized gain, the time to reach the target of the power lever angle is significantly reduced. Optimal gain scheduling shows a stable response compared with a fixed gain, which can have oscillation effects as a controller responds. An effect of using bioethanol as a fuel has been observed. It shows that for the same input parameters of the GT dynamics model, the fuel flow increases significantly, as compared with diesel fuel, because of its low bioethanol heating value. Thus, a significant increase occurs only at the gain that depends on the fuel flow.
ABSTRACT
In the energy management of district cooling plants, the thermal energy storage tank is critical. As a result, it is essential to keep track of TES results. The performance of the TES has been measured using a variety of methodologies, both numerical and analytical. In this study, the performance of the TES tank in terms of thermocline thickness is predicted using an artificial neural network, support vector machine, and k-nearest neighbor, which has remained unexplored. One year of data was collected from a district cooling plant. Fourteen sensors were used to measure the temperature at different points. With engineering judgement, 263 rows of data were selected and used to develop the prediction models. A total of 70% of the data were used for training, whereas 30% were used for testing. K-fold cross-validation were used. Sensor temperature data was used as the model input, whereas thermocline thickness was used as the model output. The data were normalized, and in addition to this, moving average filter and median filter data smoothing techniques were applied while developing KNN and SVM prediction models to carry out a comparison. The hyperparameters for the three machine learning models were chosen at optimal condition, and the trial-and-error method was used to select the best hyperparameter value: based on this, the optimum architecture of ANN was 14-10-1, which gives the maximum R-Squared value, i.e., 0.9, and minimum mean square error. Finally, the prediction accuracy of three different techniques and results were compared, and the accuracy of ANN is 0.92%, SVM is 89%, and KNN is 96.3%, concluding that KNN has better performance than others.
