Research on Status Assessment and Operation and Maintenance of Electric Vehicle DC Charging Stations Based on XGboost

Comments 1:  Mathematical model must be presented in a more rigorous form and format. The optimization models must be clarified in terms of inputs, outputs and optimization variables.  Each symbol must be clarified the first time they appear. The meaning of each equation must be better clarified.

Response 1: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

In this paper, the model includes a charging station status assessment model and an operation and maintenance model. The former assesses the operational status of a single station, while the latter focuses on the overall optimization of the operation and maintenance of multiple charging stations within the power grid, with the former serving as the foundation for the latter.

Comments 2: Several aspects have been neglected (or poorly considered) such as the uncertainties which typically affect this type of problem.

Response 2: Agree. I/We have, accordingly, done/revised/changed/modified…..to emphasize this point.

The operational characteristics of charging stations are closely related to users' charging behaviors and travel habits. As a special load within the power grid, they exhibit randomness and intermittency, and being exposed to outdoor conditions for extended periods makes them susceptible to various external environmental factors. This results in a high degree of uncertainty in the operational status of charging stations, leading to a relatively high failure rate. The proposed XGBoost method in this paper aims to achieve visualization of the uncertainty assessment of charging station states. Risk-based operation and maintenance of charging stations also fully consider these uncertainty characteristics, determining maintenance strategies based on the probabilistic assessment of their operational status.

Comments 3:  Literature on the topic must better studied and reported. Comparisons with the existing literature must clarify the need of this study and the value of the proposed approach.

Response 3: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

Current literature primarily studies the operation and maintenance methods of charging stations from a monitoring perspective. However, research on how to integrate the monitoring results of charging stations with electric vehicle charging behavior to develop reliability and risk maintenance strategies has not yet been conducted. DC charging stations consist of multiple components, including AC and DC devices, and operate at high power while being exposed to outdoor conditions, leading to a higher failure rate. Currently, the operation and maintenance work of charging stations draws on some methods from substation maintenance in the power grid, but there are issues in developing maintenance plans, such as overly simplistic proposals, excessive reliance on experience, and unclear targeted objectives. This paper proposes the application of the XGBoost method, which can achieve model evaluation visualization and clearly demonstrate the evaluation process of charging station operational status, enabling accurate assessment of operational status and optimization of maintenance from.

[update text in Section 2]

Comments 4:  The issue of the integration of charging station within the electrical systems is very important. However, it has been neglected in this paper. The Authors should discuss this topic with particular focus on the scheduling of grids connectinge multiple chaging stations. 

Response 4: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

As a special load within the power grid, charging stations have electrical characteristics closely related to users' charging behavior, exhibiting randomness and intermittency. The integration of charging stations into the power system is indeed very important. With the increasing number of electric vehicles, the scheduling and operation of the power grid must give special consideration to such loads. However, this article primarily focuses on the operation and maintenance of charging stations, which has a certain relationship with grid scheduling. In the examples presented, this paper also appropriately considers the issue of grid scheduling, analyzing the impact of different types of electric vehicle charging time characteristics on grid scheduling, followed by risk calculations and optimization of maintenance strategies.

[update text in Section 5.1]

Comments 5:  Aspects related to high power charging, such as ultra-fast charging stations, should be discussed.

Response 5: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

This article focuses on the operation and maintenance of DC charging stations. DC charging stations can quickly meet user travel needs due to their higher charging power and shorter charging times. However, they are also more prone to failures, and their maintenance work is more complex, especially in urban charging stations, where failures can lead to a backlog of electric vehicles, significantly impacting user travel. Therefore, accurately assessing the operational status of DC charging stations and conducting scientific maintenance is crucial and represents an important challenge currently faced.

Comments 6:  Results should be better discussed and reported. Results of numerical applications should demonstrate the significance and validity of the proposed approach.

Response 6: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

Based on the analysis of several basic operational status types of charging stations, XGBoost uses tree models as the base model to achieve model evaluation visualization. This clearly displays the evaluation process of the charging station's operational status. By continuously assessing through the tree model, the type with the highest probability is determined to be the closest to the actual state of the charging station. XGBoost makes the status assessment of charging stations more accurate, allowing corresponding operational maintenance tasks to be more targeted. By accurately evaluating the status of multiple charging stations, the order of operations and maintenance for stations in different states can be determined, allowing for a reasonable allocation of human and material resources to optimize maintenance work.

[update text in Section 5.2]

Comments 1: I think the integrated charging station state assessment and risk assessment may help optimize the maintenance schedule. However, the integrated framework or model is not clearly introduced. Actually, the XGBoost model is a part of the framework. It is suggested to rewrite the manuscript and make the whole framework and motivation clearer.

Response 1: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

The integrated charging station state assessment and risk assessment may help optimize the maintenance schedule.

This proposed model includes a charging station status assessment model and an operation and maintenance model. The former model utilizes the XGBoost method to evaluate the operational status of a single station. Based on accurate status assessments at each charging station, the latter model employs risk methods to optimize the operation and maintenance of multiple charging stations within the power grid. The XGBoost model is part of the overall operation and maintenance framework for charging stations, using tree models as the base model to visualize the assessment of charging station statuses. By continuously optimizing and evaluating through the tree model, the type with the highest probability is determined to be the closest to the actual operating state of the charging station. The Section 2 introduces the principles of the XGBoost model, while Section 3 describes the application model of XGBoost in assessing charging station status, including probability parameters, model parameters, loss functions, cross-validation, and the evaluation process of model trees.

[update text in Section 1]

Comments 2: The manuscript consider the current driving and charging behavior, however, the impact of behavior evolution in longer time on EV station operation and reliability should not be neglected. 

Response 2: Agree. Therefore, I have revised it as follow.

The operational characteristics of charging stations are closely related to users' charging behaviors and driving habits. As a special load within the power grid, they exhibit significant randomness and intermittency in the short term. This paper primarily addresses the issues arising from changes in the operational status of charging stations over short timeframes. In the long term, the operational characteristics of charging stations have certain regularities. By adjusting the duration and frequency of maintenance, the methods presented in this paper can guide long-term maintenance strategies. Coupled with the operation and maintenance methods of substations within the grid, appropriate planned maintenance can also be implemented for the long-term operation of charging stations.

Comments 3: It is suggested that a more detailed explanation of the chosen XGBoost parameters and how they were optimized are required. Please provides a sensitivity analysis or a discussion on the impact of parameter selection on model performance would strengthen the methodology. It would be insightful to include a comparative analysis with other machine learning algorithms.

Response 3: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

In this paper, the parameters of the XGBoost model primarily include learning rate, max depth (maximum depth of the tree), n_estimators (number of trees for each class), eval_metric (multi-class error rate), and min_child_weight (minimum weight of leaf node samples). The K-fold cross-validation method is used to divide the training data into K equal parts. Of these, K-1 parts are used for training, and 1 part is used for validation. The average classification error rate serves as the performance index to evaluate the model's classification effectiveness. The model parameters are adjusted, and the optimal parameter selection is determined based on the cross-validation results.

The training data for the model consists of historical fault data, maintenance data, monitoring data, and external environmental data. Through continuous training and calibration using multiple data sets, especially dynamic monitoring data, parameters that align with actual conditions are obtained. Compared to other machine learning algorithms, XGBoost first performs preliminary classification of the data using four types of basis functions, which avoids inconsistencies in parameter training caused by large variations in the data.

[update text in Section 5]

Comments 4:  It is important to acknowledge and discuss any limitations of the XGBoost model within the context of electric vehicle charging station maintenance.

Response 4: Thank you for pointing this out. I agree with this comment. Therefore, I have revised it as follow.

XGBoost has high requirements for the quality of input data, and noisy data and missing values can significantly affect the model's predictive performance. XGBoost relies on good feature values; the model itself does not automatically select or generate features, so the quality of feature selection and transformation directly impacts model performance. XGBoost has multiple hyperparameters that need tuning, and the configuration of these parameters has a substantial effect on model performance. This paper employs various methods to mitigate these limitations. Historical fault data is preprocessed, and before each maintenance operation, training is conducted using maintenance data, monitoring data, and external environmental data to obtain parameters that align with current conditions, continuously updating the parameters in the model. The model considers the impact of user and environmental randomness on the parameters, and during each training session, the average classification error rate under multiple parameters is used as a performance metric to evaluate the model's classification effectiveness.

[update text in Section 5]

Comments 5:  The reference format is wrong. Please make revision according to the required format.

Response 5: Thank you for pointing this out. I agree with this comment.

I have revised it in manuscript.

Comments 6:  There are lots of typos, Such like two “Error! Reference source not found.” In Section 4.1, The number of the author's organization is incorrect, too.

Response 6: Thank you for pointing this out. I agree with this comment.

I have revised it in manuscript.