|
| Types of plant | Plants’ decision parameters | AI-based modelling tools | Responses | Remarks on deficiency | Findings | References |
|
| Combined cycle power plant | Ambient temperature, vacuum, ambient pressure, relative humidity | Machine learning approaches (MLAs) | CCPP hourly electric output power prediction | Various MLAs such as the K-nearest neighbours (KNN), gradient-boosted regression rate (GBRT), linear regression (LR), artificial neural networks (ANN), and deep neural networks (DNN), estimate the electric power output with significant outcomes | Results show that the state-of-the-art outperformance of GBRT in terms of optimum electric output power solutions is depicted | Siddiqui et al. [15] |
| Combined cycle power plant | Ambient temperature, vacuum, ambient pressure, relative humidity | Machine learning approaches (MLAs) | Power plant’s electrical power output | MLPRM was not accosted in predicting the power output of the plant | Results show that the bagging REP tree method is the most suitable among the MLAs | Tüfekci [20] |
| Combined cycle power plant | Ambient temperature, vacuum, ambient pressure, relative humidity | Hybrid machine learning approaches | Power plant’s electrical power output with less waste | Butterfly optimisation algorithm (BOA) combined with a Phasmatodea population evolution (PPE) algorithm (BOAPPE), jointly with the support vector machine (SVM) predicted accurately the power output of CCCP | Results show that the BOAPPE strategy improved the convergence speed and avoided trapping into local optimum outcomes | Wang et al. [51] |
| Combined cycle gas turbine power plant | Ambient temperature, vacuum, ambient pressure, relative humidity | ANN-electrostatic discharge algorithm (EDA): ANN-EDA | Power plant’s electrical output power prediction | The multilayer perceptron regression method (MLPRM) was not adopted in the modelling of their plant | The reliability of prediction of power electricity by SDA-ANN established | Zhao and Kok Foong [54] |
| Combined cycle gas turbine power plant | Inlet flue temperature, absorber column operating pressure, amount of exhaust gas recycled, and amine concentration | Taguchi’s design of the experiment | Optimisation of postcombustion CO2 capture | Monoethanolamine solvent, employed via Taguchi’s design experimental method, mitigated the energy demands of the system | The statistical optimisation concept of the postcombustion capturing of CO2 is demonstrated | Alexanda Petrovic and Masoudi Soltani [56] |
| Natural gas-fired combined cycle power plant | Flue gas emission dataset between 2011 and 2015 | Hybrid machine learning method | Power plant’s NOx emission prediction | ANFISGA model predicted accurately the NOx emissions at the minimum error | Results show that the impact of the genetic algorithm (GA) on ANFIS performance towards optimum solutions | Dirik [57] |
| Gas turbine combined cycle power plant | Dynamic optimal set point for the regularisation level | Hybrid machine learning approaches | Power plant’s performance prediction | A fuzzy logic model (FL) coupled with a genetic algorithm (GA) predictive supervisory controller predicted accurately the performance of the power plant | Results show that the hybrid fuzzy GA predictive controller captures the system’s nonlinearities with optimum performance solutions | Saez et al. [58] |
| Combined cycle power plant boiler | Input data are selected via a sensitivity analysis approach | Machine learning approaches (MLAs) | CCPP boiler performance | A cluster number of optimum Taguchi–Sugeno fuzzy logic (FL) models, derived accurately the performance of the CCPP | Results show the economic optimisation of the plant’s performance with the nonlinear FL model and the superheated steam pressure via the linear model FL | Sáez and Zuñiga [59] |
|
