The uneven velocity distribution formed at the lateral inlet/outlet poses a significant risk of damaging the trash racks. Reasonable design of the inlet/outlet structures requires the consideration of two major aspects: the average velocity (Vm) and the coefficient of unevenness (Uc). This paper developed an optimization framework that combines an interpretable Gradient Boosting Decision Tree (SOBOL-GBDT) with a Non-dominated Sorting Genetic Algorithm (NSGA-II). 125 conditions are simulated by performing CFD simulations to generate the dataset, followed by GBDT implemented to establish a nonlinear mapping between the input parameters including vertical (α) and horizontal (β) diffusion angles, diffusion segment length (LD), channel area (CA), and the objectives Uc and Vm. The SOBOL analysis reveals that in Uc prediction, CA and α play more significant roles in the model development compared to β and LD. Besides, GBDT is observed to better capture interactive effects of the input parameters compared with other machine learning models. Subsequently, a multi-objective optimization framework using GBDT-NSGA-II is developed. The framework calculates the optimal Pareto front and determines the best solution using a pseudo-weight method. The results demonstrate that this framework leads to significant improvements in flow separation reduction in the diffusion segment and the normalized velocity distribution. The SOBOL-GBDT-NSGA-II framework facilitates a rational and effective design of the inlet/outlet.

  • Sobol indices can effectively explain the effect of individual and their interactions on the predictive model.

  • Within multi-objective optimization problems, the GBDT-NSGA-II framework can effectively identify optimal solutions.

  • In the optimal shape, lateral inlet/outlet reduces the flow separation areas in the channel, simultaneously diminishing the likelihood of flow-induced vibrations in the trash racks.

To achieve the goals of peak carbon emissions and carbon neutrality (Wei et al. 2022; Yang et al. 2022), China is actively and systematically advancing the development of renewable and clean energy utilization. Among these, pumped-storage hydropower technology plays a vital role in renewable energy systems. As an effective method of energy storage, it is crucial for electricity regulation due to its high load-balancing efficiency, reliability as a backup power source, long service life, and large storage capacity. Lateral inlet/outlet is one of the key components in pumped-storage power stations (PSPS), designed for bidirectional flow to meet the varied power demand. During inflow conditions, water is contracted and flows from the reservoir through the inlet/outlet into the tunnel; during outflow conditions, water is expanded and flows from the tunnel back into the reservoir. The bidirectional flow creates complex fluid dynamics within the inlet/outlet, which can damage the trash rack, affecting the efficiency of the power generation and, in severe cases, even damaging the turbines (Tsikata et al. 2009; Raynal et al. 2013; Xue et al. 2023). Damage to the trash rack is mainly influenced by the following two factors: (1) large turbulence fluctuations in the water flow cause long-term fatigue damage to the trash rack (Nguyen & Naudascher 1991), and (2) high local flow speeds can lead to resonance damage when the frequency of the Karman vortex street matches the natural frequency of the trash rack (Naudascher & Wang 1993). In response to these challenges, the design of lateral inlets/outlets should strictly control Vm and Uc at the trash rack section to prevent damage (Gao et al. 2018; NDRC 2019; Zhu et al. 2023).

The design of the lateral inlet/outlet primarily focuses on studying the flow velocity distribution at the trash rack. Sun et al. (2007) measured the flow velocity distribution at the trash rack under different diversion pier arrangements in lateral inlets/outlets and proposed methods to deal with bending and negative flow velocities. Ye & Gao (2011) used the realizable k-ɛ model for numerical simulation of lateral inlets/outlets, analyzed flow distribution, head loss, and inlet vortices, and compared these with experimental results. Terrier et al. (2014) conducted hydraulic model experiments to study surge wave issues in the tailrace tunnels of PSPS and analyzed head loss during sudden starts and stops. Zhu et al. (2023) performed CFD simulations to derive the flow velocity distribution and turbulence intensity at the trash rack and compared these with results from a Laser Doppler Velocimeter (LDV), finally analyzing the causes of fatigue damage to the trash rack. Gao et al. (2018) used Response Surface Methodology (RSM) to construct mappings between influencing factors in the inlet/outlet and head loss and unevenness coefficient. They performed multi-objective optimization with NSGA-II. However, RSM uses quadratic polynomial fitting that is suitable for merely simple data sets, whereas machine learning (ML) models, with their ability to handle more complex, nonlinear data, are applicable in a wider range of scenarios.

ML models are rapidly advancing in the field of water science, for instance, employing SVR (Roushangar et al. 2018; Mia & Dhar 2019; Yu et al. 2023) for dam safety monitoring, discharge coefficient prediction, and optimization of hard turning of steel; using RF (Hong et al. 2012) for predicting scour depth; utilizing GBDT (Qiu et al. 2021) to forecast ground vibration; and employing XGBOOST (Sen et al. 2023) for predicting global temperature anomalies. However, ML models are often viewed as ‘black boxes,’ with their internal decision-making processes not fully understood. This has led to a growing interest in interpretability analysis to demystify their internal mechanisms and results credibility. Initially, the SHapley Additive exPlanations (SHAP) method is widely adopted. Mo et al. (2023) analyzed the importance of different features in cavity water and length models using the XGBOOST-SHAP model. Karakaş et al. (2023) improved the accuracy and interpretability of ML model predictions for the bearing capacity of closed and open piles using SHAP values. While SHAP focuses on the impact of variables acting independently and emphasizes the contribution of individual sample points, interactive effects of more than two variables can be derived from SOBOL analysis. Zouhri et al. (2022) enhanced the robustness of SVMs against feature uncertainty by introducing the SOBOL analysis method. Xu et al. (2023) combined ML models with the SOBOL analysis to propose a global sensitivity analysis approach for assessing the safety and effectiveness of dam structural parameters. Therefore, with the ability to reveal the dependence of the target variables on the input variables, interpretable analysis offers not only valuable insights during the design phase but also the stage for multi-objective optimization analysis.

Currently, the NSGA-II is widely used in solving multi-objective optimization problems. For example, Komasi & Goudarzi (2021) used NSGA-II to design a water network consisting of 12 observation points and significantly reduced the number of monitoring sites. Dey et al. (2019) employed numerical simulation to study flow and heat transfer around square and circular shapes and used NSGA-II for optimization successfully finding the optimal roundness of the cylinder corner. Zhao et al. (2023) combined RSM with NSGA-II for optimizing lithium batteries, finally effectively reducing the maximum temperature difference and pressure difference while enhancing the PCM liquid phase fraction. In this study, considering preventing the damage to the trash rack requires the constraint of the lateral inlets and outlets to reach the desired Vm and Uc, multi-objective optimization analysis is carried out.

To achieve the multi-objective optimization of lateral inlet/outlet, this paper is organized as follows: initially, CFD simulations are performed for 125 different scenarios. Subsequently, the SOBOL-GBDT method is used to establish a nonlinear mapping between influencing factors and the objective functions Vm and Uc. The importance of each factor is analyzed through SOBOL first-order and second-order indices, considering the individual and interactive effects vary at different CA. The model's predictive accuracy and interpretability analysis are then compared with the results from SVR, RF, and XGBOOST models, confirming the reliability of the GBDT model. Finally, the SOBOL-GBDT-NSGA-II multi-objective optimization framework is applied to obtain the Pareto front solution set. Using a pseudo-weight method, the optimal solutions for two design scenarios, S1 and S2, are identified. The paper also analyzed the convergence of the algorithm and the distribution characteristics of the variable space based on these solutions.

Model parameters of the inlet/outlet

The lateral inlet/outlet of the pumped-storage power station shown in Figure 1(a) features a three-pier, four-channel arrangement with horizontal (α) and vertical (β) diffusion angles of 33.774° and 6.12°, respectively. The lateral inlet/outlet connects to the reservoir and the water conveyance system via an open channel and a tunnel, respectively. Water flows into the inlet/outlet for power generation and flows out for water pumping. The height (H) and width (B) of the channel are 13.5 m and 7.5 m, respectively, with a channel area (CA = B × H) of 101.25 m2. The diameter (D) of the tunnel is 9 m, and the length of the diffusion section (LD) is 42 m.
Figure 1

Structure of the lateral inlet/outlet. (a) Plan and section drawings; (b) model experiment; (c) velocity monitors in trash rack section; and (d) three-dimensional model.

Figure 1

Structure of the lateral inlet/outlet. (a) Plan and section drawings; (b) model experiment; (c) velocity monitors in trash rack section; and (d) three-dimensional model.

Close modal

The experiment, as shown in Figure 1(b), uses a scale model ratio of 1:40. The inlet/outlet is made of acrylic glass. Under the condition of full discharge in pumping mode, the flow rate is 262 m3/s, with an average velocity of 4.11 m/s. Flow velocities in each channel are measured using an Acoustic Doppler Velocimeter (ADV), which has a measurement accuracy of 1 × 10−3 m/s. In the inflow condition, water enters from a water tower into a triangular weir, which controls the flow rate with high precision. The water then flows into the reservoir and finally enters the inlet/outlet and downstream tunnel. The experiment mainly measures the flow velocity across the trash rack cross-section, as depicted in Figure 1(c). Time-averaged velocities at five points along the vertical direction of each channel are measured for validation of the numerical simulation.

Numerical setup and validation

The computational domain, depicted in Figure 2, primarily consists of the reservoir, lateral inlet/outlet, and the water conveyance tunnel. For a pumped-storage power station, the inlet/outlet plays a crucial role and often requires optimized design. To accomplish this, the inlet/outlet is divided into four tunnels by three separation piers. Each tunnel spans a length of 17 times its diameter (17D). The inlet is designed as a velocity inlet with a turbulence intensity of 5%. The computational length of the reservoir is 20 times its diameter (20D), with its boundary set as a pressure outlet. The free surface is treated as a pressure inlet, while all walls are assigned a no-slip boundary condition. Inflow/Outflow conditions are shown in Table 1.
Table 1

Inflow/outflow conditions

Flow directionFlow regionInlet flow rate (m3/s)Inflow velocity (m/s)Outlet water level (m)
Inflow Contracted flow 261 4.11 740 
Outflow Diffusive flow 261 4.11 740 
Flow directionFlow regionInlet flow rate (m3/s)Inflow velocity (m/s)Outlet water level (m)
Inflow Contracted flow 261 4.11 740 
Outflow Diffusive flow 261 4.11 740 
Figure 2

Boundaries conditions and computational domain.

Figure 2

Boundaries conditions and computational domain.

Close modal
The numerical simulation of the lateral inlet/outlet is conducted using the realizable k-ε turbulence model (Shaheed et al. 2019). Figure 3 illustrates the global and local mesh structures for these components. To improve computational efficiency and accuracy, a structured meshing technique is employed using ICEM. The boundary layer is resolved by 10 layers of grids with a resolution of the first layer next to the wall set as 0.01 m. The mass, momentum, and k-ɛ transport equations are solved using the finite volume method. To deal with the coupling between pressure and velocity in the Navier–Stokes equations, the PISO algorithm is applied. Spatial discretization is achieved through a second-order upwind scheme, while time discretization adopts a first-order implicit approach. The Volume of Fluid (VOF) method is utilized to track the free surface. A time step of 0.001s is employed, converging to 10−5 for each step. Averaging of the flow starts as the flow rates at the lateral inlet/outlet are stabilized.
Figure 3

Mesh topology. (a) The whole computational mesh; (b) detailed mesh on the section of the diffusion segment; and (c) detailed mesh on middle pier and channels.

Figure 3

Mesh topology. (a) The whole computational mesh; (b) detailed mesh on the section of the diffusion segment; and (c) detailed mesh on middle pier and channels.

Close modal
Three different meshes, with cell amounts of 300W, 220W, and 120W are adopted in the simulations and a comparison of their derived flow velocity distribution in the trash rack section of flow channel 1 under outflow and inflow conditions with the experiments is plotted in Figure 4(a) and 4(b). While large error exists in the coarse mesh results near the bottom surface where a large velocity gradient exists, results from both fine and medium meshes agree reasonably well with the experiments. Considering the aspects of both accuracy and efficiency, the medium mesh is selected to perform the simulations.
Figure 4

Velocity distribution in channel1 along vertical direction at the trash rack section derived from three mesh sizes: (a) outflow and (b) inflow.

Figure 4

Velocity distribution in channel1 along vertical direction at the trash rack section derived from three mesh sizes: (a) outflow and (b) inflow.

Close modal

Gradient boosting decision tree

GBDT (Friedman 2001) is a renowned ML algorithm that uses decision trees as base learners. By continuously iterating to build regression trees, each tree further optimizes the model in the direction of reducing the residual error of the previous tree. The initial model h0(x), which forms the foundation of the ensemble, is defined by minimizing the loss function L(y,h(x)). The ensemble model C is represented as:
formula
(1)

Here, f(x) denotes the final model obtained by summing the outputs of all the decision trees. The term Cm represents the contribution of the m-th tree, while I(xRm) is an indicator function that determines if the input x falls within the region Rm covered by the m-th tree.

Within each iteration, the algorithm fits a regression tree to the negative gradient:
formula
(2)
This negative gradient symbolizes the direction for the most significant reduction in the loss function, based on the current predictions. The model then updates with each tree, factoring in the optimal values Cmj found for specific regions Qmj within the tree:
formula
(3)

Here, hm(x) is the updated model after adding the m-th tree, hm−1(x) is the model from the previous iteration, Cmj is the optimal value that minimizes the loss for the j-th region Qmj, and I(xQmj) is an indicator function assigning Cmj to the inputs falling within the region Qmj.

Sobol interpretability analysis

Sobol analysis (Sobol 2001), proposed by Sobol in 2001, is a significant method used for sensitivity analysis. It is primarily employed to assess the impact of different input parameters on the output results of a model. This analysis is particularly suitable for dealing with complex models, especially in cases where there are interdependence or nonlinear relationships between the model's input parameters. In Sobol analysis, it is assumed that the model can be represented as y = f(x), where x = (x1,x2,…,xn) are the model's input parameters, and y is the output result. The core of Sobol analysis is based on the method of variance decomposition, which breaks down the total variance of the model output into contributions from different input variables. This method is primarily implemented through Sobol indices, which can be categorized into two types:

First-order index Si, which represents the contribution of a single variable xi to the output variance V(y), defined as Si = Vi/V(y), where Vi is the variance of y due to xi alone. Second-order index Sij, which accounts for the effect of interactions between two variables xi and xj, defined as Sij = Vij/V(y), where Vij represents the variance of y due to the interactive effects of xi and xj.

The method developed by Sobol was further enhanced by Saltelli et al. (2010), who suggested algorithmic extensions for calculating the sensitivity indices. The Saltelli extension incorporates new strategies for generating sample matrices and is applicable to models with a large number of input variables. Taking a model with m input variables as an example, a sample matrix of size n(2m + 2) is generated, where n is a base sample size to optimize computational efficiency. For example, with a base sample size of n = 1,024 and m = 4 input variables, the sample matrix would consist of 10,240 samples.

NSGA-II and Pareto-font solution set

NSGA-II (Deb et al. 2002) (Non-dominated Sorting Genetic Algorithm II), developed by Kalyanmoy Deb and others in 2002, is an advanced algorithm designed for multi-objective optimization problems. It improves upon NSGA by effectively addressing complex issues with multiple conflicting objectives. The algorithm is distinguished by its non-dominated sorting process, which classifies solutions based on their dominance levels. This is critical for identifying solutions that are closest to the ideal outcome in scenarios with competing objectives. Additionally, NSGA-II integrates a crowding distance mechanism, ensuring a diverse range of solutions and preventing the over-concentration of solutions in a particular region of the solution space.

The Pareto front (Poulos et al. 2001), in the context of multi-objective optimization, is a key concept intimately linked with NSGA-II's objectives. It represents a set of non-dominated solutions, meaning that no other solution in the set is superior in all objectives (Safari et al. 2018). The Pareto front forms a conceptual boundary that maps the distribution of these optimal solutions, showcasing the optimal trade-offs among different objectives. This boundary is crucial for decision-makers as it guides them in identifying the most effective solutions that strike a balance between competing objectives. NSGA-II aims to approximate this Pareto front as closely as possible (Babajamali et al. 2022), providing a comprehensive set of balanced solutions for complex optimization problems.

In multi-objective optimization, the main components include input variables and objective functions. This study is dedicated to finding the optimal shape for lateral inlet/outlet. Therefore, vertical diffusion angle, horizontal diffusion angle, length of diffusion segment, and flow channel area are selected as variables. The objective function is adjusted to reduce average flow velocity and achieve as uniform a velocity distribution as possible across the trash rack section. To address the multi-objective optimization problem effectively, the Pymoo package in Python was utilized. The algorithm was configured with a population size of 100 and iterated across 120 generations.

Objective function and dataset

Due to the potential risk of the rack section fatigue damage caused by the flow separation and intense turbulence that results from overly high average flow velocity (v) at the trash rack section (Zhu et al. 2023), it is necessary to properly design the inlet/outlet therefore control v within a reasonable range (Administration 2019), which is usually less than 1.2 m/s. During the design, the maximum v that corresponds to the most hazardous condition is chosen as the target function Vm.
formula
(4)
where i indicates the channel number.
Under the outflow condition, the inlet/outlet experiences diffusion flow, resulting in an uneven distribution of velocity across the trash rack section. The velocity near the bottom of the channel is particularly high, with local velocities reaching 2 to 4 times the average cross-sectional velocity and causing strong disturbances and fluctuations in the flow. Existing literature (wS Hamilton 1982; Nguyen & Naudascher 1991) attributes many cases of trash rack damage to significant deformations in the middle and bottom parts caused by flow-induced vibration. These issues primarily arise from the resonance between the vortex shedding and the trash rack. Therefore, the level of local velocity unevenness at the trash rack cross-section Uc defined as:
formula
(5)
where vi represents the average flow velocity, and vi,max denotes the maximum local velocity at the trash rack section for channel i. In the design of lateral inlet/outlet, it is necessary to meet the requirements for both Vm and Uc simultaneously.
The extensive experimental work conducted by previous researchers implies that the flow rate (Qi) under specific inflow conditions is governed by four variables: α, β, LD, and LS. As illustrated in Figure 1, these variables represent the vertical and horizontal diffusion angles, diffusion segment length, and spacing between diversion piers, respectively. Hence, the average velocity (vi) is determined as the ratio of the flow rate (Qi) and the channel cross-sectional area (CA), expressed as vi = Qi/CA. When LS remains constant, the simplified equation becomes
formula
(6)
Similarly, the velocity unevenness coefficient Uc is influenced by α, β, LD and CA, described by the relationship:
formula
(7)

This paper aims to provide a rational design for the inlet/outlet, as well as enrich the database for research on ML and multi-objective optimization. Specifically, five values have been selected for the vertical diffusion angle (α): 3, 4, 5, 6.12, and 7; and five values for the horizontal diffusion angle (β): 27, 30, 33.77, 37, and 41. Additionally, the diffusion segment length (LD) includes values of 32, 37, 42, 47, and 52. Ranges of the input parameters are listed in Table 2 and CFD simulations are performed for these 125 cases.

Table 2

Parameter setting in CFD

α (°)β (°)LD (m)CA (m2)UCVm (m/s)Samplers
Data 3.00,4.00,5.00,6.12,7.00 27.00,30.00, 33.77,37.00, 41.00 32,37,42,47,52 52.99–166.74 1.28–5.35 0.50–1.54 125 
α (°)β (°)LD (m)CA (m2)UCVm (m/s)Samplers
Data 3.00,4.00,5.00,6.12,7.00 27.00,30.00, 33.77,37.00, 41.00 32,37,42,47,52 52.99–166.74 1.28–5.35 0.50–1.54 125 

The technical routine of this paper, depicted in Figure 5, can be divided into three primary parts: (1) Establishing an ML model, (2) Conducting a SOBOL interpretability analysis, and (3) Performing multi-objective optimization of the lateral inlet/outlet.
Figure 5

Flowchart of optimization of lateral inlet/outlet using SOBOL-GBDT-NSGA-II frame.

Figure 5

Flowchart of optimization of lateral inlet/outlet using SOBOL-GBDT-NSGA-II frame.

Close modal

Establishment of interpretable ML models

Optimization of ML models

Evaluations of the performance of the traditional ML models including SVR (Vapnik 1999), RF (Breiman 2001), and XGBOOST (Chen & Guestrin 2016) together with GBDT are carried out. With 80% of the data allocated as the training set and the remaining 20% as the test set, Bayesian Optimization (BO) (Bergstra et al. 2011) was employed to fine-tune the hyperparameters of each ML model and the results are listed in Table 3. The predictive performance was indicated using common metrics: R2, RMSE, and MAE (Mia & Dhar 2019; Mo et al. 2023), and Table 4 provides a comparison of model performance before and after optimization on the test set. The results indicate that the optimized models have better performance, hence the analysis proceeded with these optimized models.

Table 3

Optimization results of hyperparameters

ModelHyperparameters
Uc SVR-BO C = 98.56 γ = scale ε = 0.04 
RF-BO n_estimators = 948 max_depth = 40 min_samples_leaf = 1 
GBDT-BO n_estimators = 660 max_depth = 10 Learning_rate = 0.17 
XGboost-BO n_estimators = 718 max_depth = 5 Learning_rate = 0.11 
Vm SVR-BO C = 69.51 γ = scale ε = 0.01 
RF-BO n_estimators = 279 max_depth = 30 min_samples_leaf = 1 
GBDT-BO n_estimators = 327 max_depth = 8 Learning_rate = 0.29 
XGboost-BO n_estimators = 952 max_depth = 4 Learning_rate = 0.13 
ModelHyperparameters
Uc SVR-BO C = 98.56 γ = scale ε = 0.04 
RF-BO n_estimators = 948 max_depth = 40 min_samples_leaf = 1 
GBDT-BO n_estimators = 660 max_depth = 10 Learning_rate = 0.17 
XGboost-BO n_estimators = 718 max_depth = 5 Learning_rate = 0.11 
Vm SVR-BO C = 69.51 γ = scale ε = 0.01 
RF-BO n_estimators = 279 max_depth = 30 min_samples_leaf = 1 
GBDT-BO n_estimators = 327 max_depth = 8 Learning_rate = 0.29 
XGboost-BO n_estimators = 952 max_depth = 4 Learning_rate = 0.13 
Table 4

Performance evaluation between ML models with and without BO optimization

ModelUc test
Vm test
R2RMSEMAER2RMSEMAE
SVR 0.939 0.204 0.161 0.924 0.058 0.046 
SVR-BO 0.945 0.193 0.154 0.995 0.015 0.010 
RF 0.940 0.202 0.162 0.995 0.015 0.011 
RF-BO 0.947 0.190 0.146 0.995 0.015 0.010 
GBDT 0.972 0.137 0.101 0.997 0.012 0.009 
GBDT-BO 0.987 0.093 0.077 0.997 0.012 0.009 
XGBOOST 0.959 0.168 0.127 0.987 0.024 0.020 
XGBOOST-BO 0.974 0.133 0.101 0.993 0.017 0.013 
ModelUc test
Vm test
R2RMSEMAER2RMSEMAE
SVR 0.939 0.204 0.161 0.924 0.058 0.046 
SVR-BO 0.945 0.193 0.154 0.995 0.015 0.010 
RF 0.940 0.202 0.162 0.995 0.015 0.011 
RF-BO 0.947 0.190 0.146 0.995 0.015 0.010 
GBDT 0.972 0.137 0.101 0.997 0.012 0.009 
GBDT-BO 0.987 0.093 0.077 0.997 0.012 0.009 
XGBOOST 0.959 0.168 0.127 0.987 0.024 0.020 
XGBOOST-BO 0.974 0.133 0.101 0.993 0.017 0.013 

In Uc ​ prediction, the GBDT-BO showed the best performance, with R2, RMSE, and MAE values of 0.987, 0.093, and 0.077, respectively, followed by the XGBOOST-BO model with R2, RMSE, and MAE values of 0.974, 0.133, and 0.101. The SVR-BO model had the poorest performance that presenting 4.2% lower while 54.4 and 100% higher magnitudes of R2, RMSE and MAE, respectively compared to the GBDT-BO model.

In Vm prediction, the R2 values for the SVR-BO, RF-BO, GBDT-BO, and XGBOOST-BO were 0.995, 0.995, 0.997, and 0.993, respectively, indicating strong predictivity against the test data. GBDT-BO presents the least error, with RMSE and MAE of 0.012 and 0.009, whereas XGBOOST-BO had the largest error, with its RMSE and MAE of 41.6 and 44.4% higher than GBDT-BO, respectively.

Sobol interpretability analysis

With the established ML model, the interactions and significance of various input variables in relation to the model are further investigated. The Sobol method is applied to calculate first- and second-order Sobol indices, revealing the significance of individual variables and the interactive effect of two variables during the prediction process. Considering larger Vm has a relatively milder effect on the trash rack and requires long-term accumulation of fatigue that finally leads to the trash rack failure, whereas the increase of Uc leads to a significant increase of flow-induced vibration and causes large deformation of the lower and middle parts of the trash rack (Nguyen & Naudascher 1991). Therefore, targeted on the output Uc, relative significances of the influencing factors including α, β, LD, and CA are investigated to guide the optimization for the design of lateral inlet/outlet.

Tables 5 and 6 present first-order (Si) and second-order (Sij) Sobol indices in the ML models that correspond to the individual and interactive effects of the input variables, respectively. Focusing on the Si listed in Table 5 firstly, CA presents predominant significance in the SVR and RF models with SCA of 0.97, whereas XGBOOST presents a relatively lower magnitude of SCA and higher Si for α, β and LD, indicating increased importance of these variables in the model development. In GBDT-BO, a further decrease of SCA accompanied by the increase of Sα Sβ and SLD can be observed.

Table 5

First-order Sobol's indices of ML models

SαSβSLDSCA
SVR 0.000 0.000 0.025 0.972 
RF 0.012 0.000 0.001 0.978 
GBDT 0.244 0.156 0.049 0.511 
XGBOOST 0.113 0.121 0.057 0.709 
SαSβSLDSCA
SVR 0.000 0.000 0.025 0.972 
RF 0.012 0.000 0.001 0.978 
GBDT 0.244 0.156 0.049 0.511 
XGBOOST 0.113 0.121 0.057 0.709 
Table 6

Second-order Sobol's indices of ML models

α,βα,LDα,CAβ,LDβ,CALD,CA
SVR 0.000 0.006 0.091 0.033 0.000 0.000 
RF 0.000 0.000 0.079 0.013 0.001 0.000 
GBDT 0.010 0.000 0.217 0.020 0.140 0.040 
XGBOOST 0.020 0.000 0.132 0.030 0.114 0.000 
α,βα,LDα,CAβ,LDβ,CALD,CA
SVR 0.000 0.006 0.091 0.033 0.000 0.000 
RF 0.000 0.000 0.079 0.013 0.001 0.000 
GBDT 0.010 0.000 0.217 0.020 0.140 0.040 
XGBOOST 0.020 0.000 0.132 0.030 0.114 0.000 

As for the interactive effects of dual variables listed in Table 6, noticeable Sij only exists between α and CA in SVR and RF with magnitudes of less than 0.1, whereas both GBDT and XGBOOST have Sα,CA and Sβ,CA of greater than 0.1, indicating the possible interactive effect between α and CA and β and CA that are captured by these models.

Considering the designs of the inlet/outlet are usually based on a given flow velocity that is determined by CA, a range for CA is selected between 52.99 and 166.74, within which 200 CA are chosen for analysis. By sequentially studying these 200 CA values, the Sobol method is employed in each iteration to calculate the relative significances of α, β, and LD in the ML model development at specific CA values. This approach provides a rational reference for designing the inlet/outlet structure for selected specific CA values.

Figures 6(a)–6(d) show the dependency of different ML models on α, β, and LD with respect to changes in CA. Results in Figure 6(a) indicate that in the SVR prediction, LD is the least important input, while both α and β present a stronger effect. The relative importance of these inputs remains invariant with the change of CA. The higher significance of α compared with LD can also be observed in RF, GBDT, and XGBOOST models as shown in Figures 6(b)–6(d), respectively, whereas the significance of β lies in between them. Overall, these results indicate LD plays the most significant role whereas β is the least important parameter in the Uc prediction.
Figure 6

Variation of Si values of different ML models with varying CA: (a) SVR, (b) RF, (c) GBDT, (d) XGBOOST, and (e) variation of Uc with respect to CA.

Figure 6

Variation of Si values of different ML models with varying CA: (a) SVR, (b) RF, (c) GBDT, (d) XGBOOST, and (e) variation of Uc with respect to CA.

Close modal

To reflect the generalization performance of the models, samples were taken during the iterative process mentioned above. The ML models then predicted and averaged the outcomes for the 10,240 samples drawn in that iteration. Variation of UC derived from each ML model with respect to CA were obtained and compared with the original dataset, as shown in Figure 6(e). The figure indicates SVR had the poorest predictive performance, fitting well only when CA was between 72 and 137; followed by RF and XGBOOST, showing errors near the maximum and minimum CA. Lastly, results from GBDT present the best agreement with the experimental data.

Multi-objective optimization analysis of inlet/outlet

Results in Figure 6(e) indicate that the increase of CA leads to an increase of Uc, which increases damage risk due to flow-induced vibrations. However, the increase of CA also causes Vm decreases, suggesting that lower velocities with less turbulent energy reduce fatigue damage risk. Considering the multiple critical output factors including Uc and Vm and input parameters including α, β, LD, and CA in the lateral inlets/outlets design, the necessity of a multi-objective optimization approach is necessary to achieve an optimal design. In this work, the GBDT model is adopted to establish the relationship between input variables α, β, LD, CA, and objectives Uc, Vm. During the model development, two constraints, S1 and S2 are applied: S1 requires Vm below 1.2 m/s whereas S2 keeps CA constant and performs optimization by adjusting α, β, and LD.

Algorithm convergence analysis

The NSGA-II genetic algorithm was utilized for multi-objective optimization with a configuration of a population size of 100 and 120 generations. The convergence process in S1, as depicted in Figure 7, highlights several key metrics. The green curve represents population diversity, which initially increases and reaches a peak of 0.96 after 330 evaluations, indicating a thorough exploration of the solution space. Then, it decreases to a minimum of 0.63 at 970 evaluations, suggesting a narrowing down of the search range and approaching the optimal solutions. The blue curve represents the hypervolume metric with an initial magnitude of 0.57 and increases to 0.80 after 800 evaluations, denoting improved solution quality and algorithmic convergence. The red curve represents the average constraint violation of the population, which exhibits a rapid drop in the first 300 iterations and reaches zero after 460 evaluations, confirming that all individuals in the population adhere to the set constraints, ensuring compliance within the solution set.
Figure 7

Algorithm convergence analysis.

Figure 7

Algorithm convergence analysis.

Close modal

Analysis of patterns in objective and variable space

Figure 8 shows the Pareto front for S1 and S2, illustrating the trade-off between two objective functions, Uc and Vm. Any improvement in one solution leads to the deterioration of the other solution, known as a non-dominated solution. It is observed that the range of the solution set for S1 is significantly larger compared to S2, indicating that S2 has stricter constraints in the model development. The selected point (marked as a red cross) represents the optimal solution achievable in the design phase obtained using pseudo-weights (Kalyanmoy 2001). In this process, the weights for Uc and Vm are 0.7 and 0.3, respectively, indicating a greater emphasis on mitigating the adverse effects caused by Uc during the selection process. The objective function values for S1, S2 and the original design are presented in Table 7. It is evident that Uc decreases sharply in S1 with a magnitude of 33% compared to the original design. However, an increase of Vm with 16.7% compared with the original design exists. S1 aims to increase Vm to some extent while reducing Uc to prevent excessive Uc-induced flow-induced vibration on the debris barrier. In S2, Vm remains essentially unchanged at 0.804 m/s, while Uc decreases by 18.5% compared to the original design.
Table 7

Comparison of design variables for different cases

α (°)β (°)LD (m)CA (m2)Vm (m/s)Uc
Original 6.12 33.77 42.00 101.25 0.803 3.18 
S1 3.29 28.80 49.60 88.76 0.937 2.13 
S2 3.67 32.10 50.14 101.78 0.804 2.59 
α (°)β (°)LD (m)CA (m2)Vm (m/s)Uc
Original 6.12 33.77 42.00 101.25 0.803 3.18 
S1 3.29 28.80 49.60 88.76 0.937 2.13 
S2 3.67 32.10 50.14 101.78 0.804 2.59 
Figure 8

Plot of the pareto optimal frontier: (a) S1 and (b) S2.

Figure 8

Plot of the pareto optimal frontier: (a) S1 and (b) S2.

Close modal
Figure 9 shows the variable density and scatter distributions from the NSGA-II optimization. Blue and red denote variables for S1 and S2, respectively. The cross markers indicate optimal combinations. In S1, α are predominantly distributed between 3° and 4°, and in S2, the range becomes 3° and 4.5°. Significant reduction exists compared to its original magnitude of 6.12°. Optimal β in S1 and S2 are reduced by 14.7 and 4.9%, respectively, from the original 33.77°. The reductions in α and β aim to prevent flow separation, thus reducing Uc. The range of CA is more varied in S1 but concentrated between 95 m2–110 m2 in S2, with optimal values at 88.79 m2 for S1 and 101.78 m2 for S2. LD is focused around 45–50 m in S1 and 49–51 m in S2, compared to the original 42 m, with optimal values at 49.6 and 50.14 m. The optimization strategies, reducing α, β, and CA while increasing LD, are evident in Figures 9(i), 9(j), 9(g), and 9(l), effectively lowering Uc while controlling Vm.
Figure 9

Sample distribution in the variable space.

Figure 9

Sample distribution in the variable space.

Close modal

Analysis of optimization results

CFD simulations were conducted using the optimized input variable combinations from S1 and S2, respectively and comparisons with the original scheme were carried out. Focus on the flow patterns and fluid velocity distribution unevenness that are critical influencing factors of trash rack failure. Slices of the x-directional fluid velocity distribution in Channel 1 are shown in Figure 10. Under outflow conditions, the original scheme experienced flow separation after the diffusion section, concentrating the main flow velocity near the middle and bottom. Compared to the original case, S1 and S2 significantly reduced the area of flow separation therefore resulting in a more uniform velocity distribution across the trash rack section. During inflow conditions, characterized by contracting flow, the differences in flow patterns were relatively minor.
Figure 10

Average velocity contour distribution in the vertical cross-section: (a) outflow and (b) inflow.

Figure 10

Average velocity contour distribution in the vertical cross-section: (a) outflow and (b) inflow.

Close modal
Figure 11 shows the velocity distributions normalized by the averaged flow velocity for Channels 1 and 2. In outflow conditions (Figure 11(a) and 11(b)), Channel 2 exhibits a larger variation of flow velocity than Channel 1, indicating higher fluid velocity and more flow separation exists near the bottom surface and has a greater risk of vibration damage to the trash rack. The maximum normalized velocity in Channel 2 was reduced from 3.10 in the original case to 2.16 in the optimized S1 scheme, while S2 showed less reduction to 2.47, which may result from the constraint of constant CA. In Channel 1, the maximum normalized flow velocity at S1 is 1.57, which is a 29.2% reduction compared to the original case, while the maximum normalized flow velocity at S2 is 2.16, only decreasing by 2.7%. In inflow conditions, the velocity distribution across the trash rack section is relatively similar for different schemes, hence these results are not presented.
Figure 11

Normalized velocity distribution in (a) channel 1 and (b) channel 2 under outflow conditions.

Figure 11

Normalized velocity distribution in (a) channel 1 and (b) channel 2 under outflow conditions.

Close modal

Disccusion

This study carried out interpretability analysis for the ML models, providing crucial insights for capturing and understanding the impact of input features on model predictions (Altmann et al. 2010; Mi et al. 2020). Most researchers (Karakaş et al. 2023; Mo et al. 2023; Zhao et al. 2024) conduct interpretability analysis of ML models using the Shapley value method, which calculates the Shap values of input variables to analyze feature importance. However, these values only represent the individual effect of features on the model. This paper employs the SOBOL method to calculate first- and second-order SOBOL indices, revealing that the results are influenced not only by the individual action of features but also by the interactions among features, especially in GBDT and XGBOOST models, as shown in Table 6. The SOBOL method can reveal the impact of variables or variable combinations on model output, serving as a powerful tool for analyzing high-dimensional data and nonlinear models. From an engineering perspective, the quantified relative significances of input variables under different CA provide an intuitional reference for the design of the inlet/outlet.

However, limitations also exist in the current investigation. Due to the extensive time required for CFD simulations, the range of input variables is restricted. Future work should aim to extend the investigated parameter range to derive the ML model with better general applicability. Moreover, different multiple optimization algorithms exist in the literature which makes it worth carrying out an investigation for evaluating their relative performances when applied to the optimized lateral inlet/outlet design.

In the PSPS, there is a presence of complex bidirectional flow in the lateral inlet/outlet. It is essential for the design process to effectively reduce the excessive turbulence intensity in the flow channels and prevent resonance damage caused by aligning the water flow frequency with the trash rack frequency. Given the multiple factors that impact the overall design, it is crucial to employ a robust optimization method to simultaneously optimize the Vm and Uc at the trash rack. To address these concerns, this paper proposes the implementation of the SOBOL-GBDT-NSGA-II framework for optimizing the structures of both the inlet/outlet. The principal findings of this study are as follows:

The GBDT model accurately predicts the nonlinear mapping relationship between input variables, such as the vertical diffusion angle (α), horizontal diffusion angle (β), diffusion segment length (LD), and channel area (CA), with the outputs Uc and Vm. Additionally, the precision of GBDT's results is enhanced through BO. For instance, when considering Uc, after optimization, GBDT achieves an R2 of 0.987, RMSE of 0.093, and MAE of 0.077. These performance metrics showcase the high level of accuracy and reliability of the GBDT model, surpassing the results obtained by SVR, RF, and XGBOOST models.

In the GBDT model, SCA is significantly larger than SαSβ and SLD, indicating that CA has the greatest impact on the Uc prediction model. Meanwhile, non-negligible magnitudes for Sα and Sβ are observed, denoting a discernible impact of α and β on the GBDT model. While in other ML models, notably SVR and RF, trivial magnitudes of Sα and Sβ indicate their much lower influence on the model's efficacy. Moreover, the GBDT model exhibits significant values for both the Sα,CA and Sβ, CA, indicative of the pronounced effect of these pair interactions on the model's performance, whereas no interactive effect between the input variables is captured in SVR and RF. This indicates that GBDT is better at extracting interactive information from variables, facilitating the capture of complex, nonlinear relationships between the input variables and the output. As CA increases, α has the greatest influence on the model, followed by β, while LD has the least. This pattern reflects how changes in the CA can significantly affect the relative importance of different design parameters in predicting the Uc, which is crucial for optimizing the design of lateral inlet/outlet in hydraulic structures.

The SOBOL-GBDT-NSGA-II framework is used to identify the optimal Pareto front solutions for two different design scenarios, S1 and S2, aiming to minimize both Vm and Uc. The optimal condition in both S1 and S2 signified lower α, β, and CA with higher LD compared with the original case. This optimization approach can provide guidance for similar projects and help determine appropriate parameter ranges. The multi-objective optimization takes into account the mutual influences of both objectives. Through this framework, flow separation in the lateral inlet/outlet is significantly reduced. In S1, Uc is reduced by 33% while Vm increases by 16.7%, and in S2, Uc decreases by 18.5% with Vm remaining essentially unchanged. This effectively lowers the likelihood of resonance damage to the trash rack.

This research is supported by the National Nature Science Foundation of China (Grant Nos 52179060, 51909024, and 52209081).

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.

Administration
N. E.
2019
NB/T 10072—2018 Code for Design of Pumped Storage Power Stations
.
China Water and Power Press
,
Beijing
.
Altmann
A.
,
Tolosi
L.
,
Sander
O.
&
Lengauer
T.
2010
Permutation importance: A corrected feature importance measure
.
Bioinformatics
26
(
10
),
1340
1347
.
Available from: https://www.ncbi.nlm.nih.gov/pubmed/20385727. http://doi.org/10.1093/bioinformatics/btq134
Babajamali
Z.
,
Aghadavoudi
F.
,
Farhatnia
F.
,
Eftekhari
S. A.
&
Toghraie
D.
2022
Pareto multi-objective optimization of tandem cold rolling settings for reductions and inter stand tensions using NSGA-II
.
ISA Transactions
130
,
399
408
.
http://doi.org/10.1016/j.isatra.2022.04.002
.
Bergstra
J.
,
Bardenet
R.
,
Bengio
Y.
&
Kégl
B.
2011
Algorithms for hyper-parameter optimization
. In:
Paper Presented at the Advances in Neural Information Processing Systems
.
Breiman
L.
2001
Random forests
.
Machine Learning
45
(
1
),
5
32
.
http://doi.org/10.1023/A:1010933404324
.
Chen
T.
&
Guestrin
C.
2016
XGBoost
. In:
Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
, pp.
785
794
.
http://doi.org/10.1145/2939672.2939785
Deb
K.
,
Pratap
A.
,
Agarwal
S.
&
Meyarivan
T.
2002
A fast and elitist multiobjective genetic algorithm: NSGA-II
.
IEEE Transactions on Evolutionary Computation
6
(
2
),
182
197
.
http://doi.org/10.1109/4235.996017
.
Dey
P.
,
Bose
S.
&
Das
A. K.
2019
Multi-objective optimization of the heat transmission and fluid forces around a rounded cornered square cylinder
.
Neural Computing and Applications
31
,
1195
1209
.
http://doi.org/10.1007/s00521-017-3094-5
.
Friedman
J. H.
2001
Greedy function approximation: A gradient boosting machine
.
Annals of Statistics
1189
1232
.
http://doi.org/10.1214/aos/1013203451
.
Hong
J.-H.
,
Goyal
M. K.
,
Chiew
Y.-M.
&
Chua
L. H. C.
2012
Predicting time-dependent pier scour depth with support vector regression
.
Journal of Hydrology
468–469
,
241
248
.
http://doi.org/10.1016/j.jhydrol.2012.08.038
.
Kalyanmoy
D.
2001
Multi-objective Optimization Using Evolutionary Algorithms
.
John Wiley & Sons, New York
.
Karakaş
S.
,
Taşkın
G.
&
Ülker
M.
2023
Re-evaluation of machine learning models for predicting ultimate bearing capacity of piles through SHAP and Joint Shapley methods
.
Neural Computing and Applications
1
19
.
http://doi.org/10.1007/s00521-023-09053-3
.
Mi
J.-X.
,
Li
A.-D.
&
Zhou
L.-F.
2020
Review study of interpretation methods for future interpretable machine learning
.
IEEE Access
8
,
191969
191985
.
http://doi.org/10.1109/access.2020.3032756
.
Mia
M.
&
Dhar
N. R.
2019
Prediction and optimization by using SVR, RSM and GA in hard turning of tempered AISI 1060 steel under effective cooling condition
.
Neural Computing and Applications
31
,
2349
2370
.
http://doi.org/10.1007/s00521-017-3192-4
.
Mo
T.
,
Li
S.
&
Li
G.
2023
An interpretable machine learning model for predicting cavity water depth and cavity length based on XGBoost–SHAP
.
Journal of Hydroinformatics
.
http://doi.org/10.2166/hydro.2023.050
Naudascher
E.
&
Wang
Y.
1993
Flow-induced vibrations of prismatic bodies and grids of prisms
.
Journal of Fluids and Structures
7
(
4
),
341
373
.
http://doi.org/10.1006/jfls.1993.1021
.
NDRC
2019
Design Guide for Pumped Storage Power Station
.
China Electric Power Press
,
Beijing
.
Nguyen
T. D.
&
Naudascher
E.
1991
Vibration of beams and trashracks in parallel and inclined flows
.
Journal of Hydraulic Engineering
117
(
8
),
1056
1076
.
http://doi.org/10.1061/(ASCE)0733-9429(1991)117:8(1056)
.
Poulos
P.
,
Rigatos
G.
,
Tzafestas
S.
&
Koukos
A.
2001
A Pareto-optimal genetic algorithm for warehouse multi-objective optimization
.
Engineering Applications of Artificial Intelligence
14
(
6
),
737
749
.
http://doi.org/10.1016/S0952-1976(01)00036-7
.
Qiu
Y.
,
Zhou
J.
,
Khandelwal
M.
,
Yang
H.
,
Yang
P.
&
Li
C.
2021
Performance evaluation of hybrid WOA-XGBoost, GWO-XGBoost and BO-XGBoost models to predict blast-induced ground vibration
.
Engineering with Computers
.
http://doi.org/10.1007/s00366-021-01393-9
Raynal
S.
,
Chatellier
L.
,
Courret
D.
,
Larinier
M.
&
David
L.
2013
An experimental study on fish-friendly trashracks–Part 2. Angled trashracks
.
Journal of Hydraulic Research
51
(
1
),
67
75
.
http://doi.org/10.1080/00221686.2012.753647
.
Roushangar
K.
,
Akhgar
S.
&
Salmasi
F.
2018
Estimating discharge coefficient of stepped spillways under nappe and skimming flow regime using data driven approaches
.
Flow Measurement and Instrumentation
59
,
79
87
.
http://doi.org/10.1016/j.flowmeasinst.2017.12.006
.
Safari
M. A. M.
,
Masseran
N.
&
Ibrahim
K.
2018
Optimal threshold for Pareto tail modelling in the presence of outliers
.
Physica A: Statistical Mechanics and its Applications
509
,
169
180
.
http://doi.org/10.1016/j.physa.2018.06.007
.
Saltelli
A.
,
Annoni
P.
,
Azzini
I.
,
Campolongo
F.
,
Ratto
M.
&
Tarantola
S.
2010
Variance based sensitivity analysis of model output. Design and estimator for the total sensitivity index
.
Computer Physics Communications
181
(
2
),
259
270
.
http://doi.org/10.1016/j.cpc.2009.09.018
.
Sen
D.
,
Huseyinoglu
M. F.
&
Günay
M. E.
2023
Prediction of global temperature anomaly by machine learning based techniques
.
Neural Computing and Applications
1
14
.
http://doi.org/10.1007/s00521-023-08580-3
.
Shaheed
R.
,
Mohammadian
A.
&
Kheirkhah Gildeh
H.
2019
A comparison of standard k–ε and realizable k–ε turbulence models in curved and confluent channels
.
Environmental Fluid Mechanics
19
,
543
568
.
http://doi.org/10.1007/s10652-018-9637-1
.
Sobol
I. M.
2001
Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates
.
Mathematics and Computers in Simulation
55
(
1–3
),
271
280
.
http://doi.org/10.1016/S0378-4754(00)00270-6
.
Sun
S.
,
Liu
H.
,
Li
Z.
&
Sun
C.
2007
Study on velocity distribution behind the trashrack in lateral intake/outlet of pumped storage power station
.
Journal of Hydraulic Engineering
38
(
11
),
1329
1335
.
Terrier
S.
,
Bieri
M.
,
Cesare
G. D.
&
Schleiss
A. J.
2014
Surge wave propagation in a common tailrace channel for two large pumped-storage plants
.
Journal of Hydraulic Engineering
140
(
2
),
218
225
.
http://doi.org/10.1061/(ASCE)HY.1943-7900.0000809
.
Tsikata
J. M.
,
Tachie
M. F.
&
Katopodis
C.
2009
Particle image velocimetry study of flow near trashrack models
.
Journal of Hydraulic Engineering
135
(
8
),
671
684
.
http://doi.org/10.1061/(ASCE)HY.1943-7900.0000070
.
Vapnik
V.
1999
The Nature of Statistical Learning Theory
.
Springer Science & Business media, Berlin
.
Wei
Y.-M.
,
Chen
K.
,
Kang
J.-N.
,
Chen
W.
,
Wang
X.-Y.
&
Zhang
X.
2022
Policy and management of carbon peaking and carbon neutrality: A literature review
.
Engineering
14
,
52
63
.
http://doi.org/10.1016/j.eng.2021.12.018
.
wS Hamilton
J.
1982
Tulis Some design factors for the Both county trashracks. Water Power& Dam Construction August
.
Xu
B.
,
Wang
S.
,
Xia
H.
,
Zhu
Z.
&
Chen
X.
2023
A global sensitivity analysis method for safety influencing factors of RCC dams based on ISSA-ELM-Sobol
.
Structures
51
,
288
302
.
http://doi.org/10.1016/j.istruc.2023.03.027
.
Xue
B.
,
Mao
Y.
,
Zhang
C.
,
Zhang
H.
&
Qi
Z.
2023
Nonlinear fluid-induced vibro-impact analysis on the fatigue failure pattern of a large-scale trashrack with a reduced-order model
.
Structures
49
,
467
478
.
http://doi.org/10.1016/j.istruc.2023.01.091
.
Yang
P.
,
Peng
S.
,
Benani
N.
,
Dong
L.
,
Li
X.
,
Liu
R.
&
Mao
G.
2022
An integrated evaluation on China's provincial carbon peak and carbon neutrality
.
Journal of Cleaner Production
377
,
134497
.
http://doi.org/10.1016/j.jclepro.2022.134497
.
Ye
F.
&
Gao
X.-p.
2011
Numerical simulations of the hydraulic characteristics of side inlet/outlets
.
Journal of Hydrodynamics
23
(
1
),
48
54
.
http://doi.org/10.1016/s1001-6058(10)60087-x
.
Yu
X.
,
Li
J.
&
Kang
F.
2023
A hybrid model of bald eagle search and relevance vector machine for dam safety monitoring using long-term temperature
.
Advanced Engineering Informatics
55
,
101863
.
http://doi.org/10.1016/j.aei.2022.101863
.
Zhao
D.
,
Chen
M.
,
Lv
J.
,
Lei
Z.
&
Song
W.
2023
Multi-objective optimization of battery thermal management system combining response surface analysis and NSGA-II algorithm
.
Energy Conversion and Management
292
,
117374
.
http://doi.org/10.1016/j.enconman.2023.117374
.
Zhao
J.
,
Zhang
C.
,
Wang
J.
,
Abbas
Z.
&
Zhao
Y.
2024
Machine learning and SHAP-based susceptibility assessment of storm flood in rapidly urbanizing areas: A case study of Shenzhen, China
.
Geomatics, Natural Hazards and Risk
15
(
1
),
2311889
.
http://doi.org/10.1080/19475705.2024.2311889
.
Zhu
H.
,
Gao
X.
,
Liu
Y.
&
Liu
S.
2023
Numerical and experimental assessment of the water discharge segment in a pumped-storage power station
.
Energy
265
,
126375
.
http://doi.org/10.1016/j.energy.2022.126375
.
Zouhri
W.
,
Homri
L.
&
Dantan
J.-Y.
2022
Handling the impact of feature uncertainties on SVM: A robust approach based on Sobol sensitivity analysis
.
Expert Systems with Applications
189
,
115691
.
http://doi.org/10.1016/j.eswa.2021.115691
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).