Fuel cell state estimation method, state evaluation equipment and storage medium

Abstract

The invention provides a fuel cell state estimation method, state estimation equipment and a storage medium. The method comprises the following steps: dividing working modes; establishing a comprehensive evaluation index system; operating the fuel cell at different stages, obtaining the numerical value of the accuracy index and the numerical value of the stability index in the comprehensive evaluation index system of the observer, and performing homogenization processing; determining an accuracy score and a stability score of each sub-index in each working mode; determining accuracy sub-index scores and stability sub-index scores of different observers in each working mode by adopting a combined empowerment-multiplication synthesis method; determining that different observers are used for comprehensive scoring in different stages; and establishing a control strategy of the fused observer based on the comprehensive score of each observer in different stages. The evaluation result is objective and reliable, and a direction can be provided for improvement of fuel cell state estimation.

Description

Technical Field

[0001] The present invention relates to the field of fuel cells, and particularly to a fuel cell state estimation method, a state evaluation device and a storage medium. Background Art

[0002] Since fuel cells need to frequently cope with complex working conditions such as vehicle dynamic loading, start-stop operations, continuous low load or idling operation, and high power output. In order to ensure the stable operation of fuel cells and improve their efficiency, it is necessary to monitor the state of fuel cells in real time.

[0003] Currently, the observers of fuel cell systems include Luenberger observer, Kalman filter observer (KF), sliding mode observer (SMO), etc. These observers have played an important role in estimating the state of PEMFC systems.

[0004] Although a single observer can provide a certain state estimation ability under certain conditions, its limitations gradually emerge in complex practical applications. When the working load, temperature or fuel quality of the fuel cell changes, a single observer may not be able to accurately capture these changes, resulting in state estimation deviation. Secondly, a single observer lacks redundancy. Once a failure or performance degradation occurs, it will directly affect the operation of the entire system, reducing the reliability and stability of the system. In addition, due to the lack of diverse observation data sources, the single observer has limited state estimation and fault detection capabilities, and it is difficult to detect and handle abnormal situations in a timely manner, which further increases the risk of system operation.

[0005] In contrast, the multi-observer fusion strategy has shown significant advantages in fuel cell state estimation. The multi-observer fusion strategy significantly improves the accuracy of state estimation by integrating the observation results from multiple observers. Different observers are based on different assumptions or modeling methods, and their complementary effects help to reduce the errors caused by the deviation of a single observer, thus providing more accurate state estimation values. Summary of the Invention

[0006] Aiming at the deficiencies in the prior art, the present invention provides a fuel cell state estimation method, a fuel cell state evaluation device and a storage medium, which can establish a comprehensive score according to the accuracy index and stability index of the observer, and is applicable to fuel cell state estimation under different working modes.

[0007] The present invention achieves the above technical objectives through the following technical means.

[0008] A fuel cell state estimation method includes the following steps: The working mode is divided into a starting stage, a steady state stage, and a variable load stage according to the change of the working current; An integrated evaluation index system is established. The integrated evaluation index system includes an observer accuracy integrated evaluation matrix G and a stability integrated evaluation matrix D; the observer accuracy integrated evaluation matrix G includes the weights of the accuracy indexes in different working stages, and the observer stability integrated evaluation matrix D includes the weights of the stability indexes in different working stages; The fuel cell is operated in different stages to obtain the values of the accuracy indexes and the values of the stability indexes in the integrated evaluation index system of the observer, and they are normalized; the values of the accuracy indexes include the root mean square error value, the mean absolute percentage error value, the mean absolute percentage error, and the coefficient of determination; the values of the stability indexes include the maximum convergence time, the average convergence time, the maximum chattering amplitude, and the average chattering amplitude; the normalized values of the sub-indexes of accuracy form an accuracy index normalization matrix J, and the normalized values of the sub-indexes of stability form a stability index normalization matrix H; The observer accuracy integrated evaluation matrix G and the stability integrated evaluation matrix D of the sub-indexes are combined with the obtained accuracy normalization matrix And the stability sub-index normalization matrix For comprehensive evaluation to determine the accuracy scores of each sub-index in each working mode And the stability scores ; The combination weighting-multiplication synthesis method is used to determine the accuracy sub-index scores and stability sub-index scores of the SMO observer in each working mode; the combination weighting-multiplication synthesis method is used to determine the accuracy sub-index scores and stability sub-index scores of the UKF observer in each working mode; Determine the comprehensive scores of using the SMO observer and the comprehensive scores of using the UKF observer in different stages; Based on the comprehensive scores of each observer in different stages, a control strategy for the fusion observer is established.

[0009] Furthermore, the observer accuracy integrated evaluation matrix G is expressed as: , represents The weight of the u-th accuracy index in the stage; , u = 1 represents the root mean square error; u = 2 represents the mean absolute percentage error; u = 3 represents the maximum error; u = 4 represents the coefficient of determination; , represents the starting stage, represents the steady state stage, represents the variable load stage; The observer stability integrated evaluation matrix D is expressed as: , Represents the weight of the z-th stability index in the stage; , where z = 1 represents the maximum convergence time; z = 2 represents the average convergence time; z = 3 represents the maximum chattering amplitude; z = 4 represents the average chattering amplitude.

[0010] Furthermore, run the fuel cell under different stages to obtain the numerical values of the accuracy indexes in the comprehensive evaluation index system of the observer , , where: A1 is the root mean square error value, A2 is the mean absolute percentage error value, A3 is the mean absolute percentage error; A4 is the coefficient of determination; ; ; ; ; where: n represents the number of samples collected, m represents the position of each observed value in the dataset, represents the actual value, represents the observed value using the unscented Kalman filter observer, represents the predicted value of the observed value using the unscented Kalman filter observer, represents the average value of the actual values. k represents the number of times of data collection; t k represents the time of the data collected at the k-th time; t k+1 represents the time of the data collected at the (k + 1)-th time; Run the fuel cell under different stages to obtain the numerical values of the stability indexes in the comprehensive evaluation index system of the observer , where: X1 is the maximum convergence time; X2 is the average convergence time; X3 is the maximum chattering amplitude; X4 is the average chattering amplitude; ; ; ; ; where, represents the time taken for data convergence in the d-th data collection, and e represents the total number of data collections; is the maximum value of the observed value using the unscented Kalman filter observer, is the minimum value of the observed value using the unscented Kalman filter observer; The values of the sub-indexes in different stages and Normalization is performed, and the normalized values of the sub-indicators of accuracy form the accuracy indicator normalization matrix and the normalized values of the sub-indicators of stability form the stability indicator normalization matrix , where: ; ; In the formula: represents the normalized value of the u-th accuracy indicator after using the observer in the stage; represents

[0011] the normalized value of the z-th stability indicator after using the observer in the , ; the accuracy scores of the sub-indicators under each working mode are expressed as , which represents the v-th column of the matrix; Determine the stability scores of the sub-indicators under each working mode , ; the accuracy scores of the sub-indicators under each working mode are expressed as , which represents the w-th column of the matrix.

[0012] Furthermore, the combined weighting-multiplication synthesis method is used to determine the accuracy sub-indicator scores and stability sub-indicator scores of the SMO observer under each working mode; the combined weighting-multiplication synthesis method is used to determine the accuracy sub-indicator scores and stability sub-indicator scores of the UKF observer under each working mode. Specifically: The accuracy scores of the sub-indicators under each working mode obtained by the SMO observer are denoted as ; the accuracy indicator normalization matrix under each working mode obtained by the SMO observer is denoted as ; the stability scores of the sub-indicators under each working mode obtained by the SMO observer are denoted as ; the stability indicator normalization matrix under each working mode obtained by the SMO observer is denoted as ; the accuracy score matrix of the sub-indicators under each working mode obtained by the SMO observer is calculated with the observer accuracy normalized value matrix using the combined weighting-multiplication synthesis method to obtain the accuracy score of the u-th sub-indicator in the stage in the SMO observer, denoted as ; The stability scoring matrix of each sub-index under each working mode obtained by the SMO observer and the observer stability normalization matrix are calculated using the combined weighting-multiplication synthesis method to obtain the stability score of the z-th sub-index in the stage of the SMO observer, denoted as ; The accuracy score of each sub-index under each working mode obtained by the UKF observer is denoted as , and the accuracy index normalization matrix under each working mode obtained by the UKF observer is denoted as ; The stability index normalization matrix under each working mode obtained by the UKF observer is denoted as ; The accuracy scoring matrix of each sub-index under each working mode obtained by the UKF observer and the observer accuracy normalization matrix are calculated using the combined weighting-multiplication synthesis method to obtain the accuracy score of the u-th sub-index in the stage of the UKF observer, denoted as ; The stability scoring matrix of each sub-index under each working mode obtained by the UKF observer and the observer stability normalization matrix are calculated using the combined weighting-multiplication synthesis method to obtain the stability score of the z-th sub-index in the stage of the UKF observer, denoted as .

[0013] Further, determine the comprehensive scores using the SMO observer and the UKF observer in different stages, specifically: According to the accuracy score of the u-th sub-index and the stability score of the z-th sub-index in the stage of the SMO observer, determine the comprehensive score of the SMO observer in the ; According to the accuracy score of the u-th sub-index and the stability score of the z-th sub-index in the stage of the UKF observer, determine the comprehensive score of the UKF observer in the ; Where: is the comprehensive score using the SMO observer in the is The comprehensive score is observed using the SMO observer in this stage.

[0014] Furthermore, a control strategy for the fusion observer is established based on the comprehensive scores of each observer in different stages, specifically as follows: In stage, compare the comprehensive score of the state estimation using the SMO observer with the comprehensive score of the state estimation using the UKF observer , If , and , then in stage, use the SMO observer as the observer for fuel cell state estimation to evaluate the performance; If , and , then use the fusion observation strategy; If , and , then in stage, use the UKF observer as the observer for fuel cell state estimation to evaluate the performance; If , and , then use the fusion observation strategy; a is a set value.

[0015] Furthermore, the specific fusion observation strategy is as follows: Calculate the accuracy score and stability score of using the SMO observer and the UKF observer respectively. The calculation formulas are as follows: ; ; In the formula: is the accuracy score of using the SMO observer in stage; is the stability score of using the SMO observer in stage; is the accuracy score of using the UKF observer in stage; is the stability score of using the UKF observer in stage; When > , and > , then use the SMO observer for performance evaluation; When < , and < , then use the UKF observer for performance evaluation; When > and < then the SMO observer is used for accuracy performance evaluation, and the UKF observer is used for stability performance evaluation; When < and > then the UKF observer is used for accuracy performance evaluation, and the SMO observer is used for stability performance evaluation.

[0016] A fuel cell state evaluation device includes a processor and a memory. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the fuel cell state estimation method as described above are run.

[0017] A storage medium stores a computer program. When the computer program is executed by a processor, the steps in the fuel cell state estimation method are run.

[0018] The beneficial effects of the present invention are as follows: 1. The fuel cell state estimation method of the present invention aims to comprehensively consider the accuracy and stability of fuel cell state evaluation in different working stages. By constructing a comprehensive evaluation matrix applicable to each working mode, this method can provide more comprehensive and accurate state monitoring and prediction, thereby ensuring the efficient, safe, and reliable operation of the fuel cell system.

[0019] 2. The fuel cell state estimation method of the present invention can obtain scores of evaluation indicators related to the state estimation of the fuel cell hydrogen supply system, and can provide a reference direction for the improvement of fuel cell state estimation.

[0020] 3. For the R & D personnel of fuel cell state estimation, the fuel cell state estimation method of the present invention can obtain the best evaluation effect of the observer in different working stages, so as to guide the improved design of the fuel cell hydrogen supply system state estimation components. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. The following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, it is obvious that other drawings can also be obtained based on these drawings.

[0022] Figure 1Schematic diagram for dividing the operating modes of the fuel cell according to the present invention.

[0023] Figure 2 Flowchart of the fusion observer control strategy according to the present invention. Detailed implementation manners

[0024] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0025] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "axial", "radial", "vertical", "horizontal", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality" means two or more unless otherwise specifically defined.

[0026] In the present invention, unless otherwise clearly specified and defined, the terms "mounted", "connected", "connected", "fixed", etc. should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0027] The fuel cell state estimation method according to the present invention includes the following steps: S01: Dividing the operating mode. According to the change of the operating current, the operating mode is divided into a starting stage, a steady state stage, and a variable load stage. As Figure 1 shown, the specific division of the operating mode is as follows: According to different loads, the operating current of the fuel cell can be divided into n stages, where the current in the i-th stage is set as I i , i ∈ (1, n); Startup phase: The load current I gradually increases from 0 to During this process, when the fuel cell operates, the load current I will experience a period of acceleration until it reaches a stable operating state. During this stage, the change in the current load is relatively large, and the performance of the circuit may not be stable enough.

[0028] Steady-state phase: For different loads, within at least time, the current remains unchanged, which is called the i-th steady-state phase. Within another at least time, the current remains unchanged, which is called the (i + 1)-th steady-state phase.

[0029] Load change phase: The process in which the current changes from the i-th steady-state phase to the (i + 1)-th steady-state phase. The change in the load current between adjacent stable phases is defined as the load change phase to flexibly adapt to the operating states under various load conditions.

[0030] S02: Establish a comprehensive evaluation index system, including constructing a comprehensive evaluation matrix G for the accuracy of the observer and constructing a comprehensive evaluation matrix D for the stability of the observer; The comprehensive evaluation matrix G for the accuracy of the observer includes the weights of the accuracy indexes in different working phases, and the comprehensive evaluation matrix D for the stability of the observer includes the weights of the stability indexes in different working phases; The accuracy indexes include root mean square error RMSE, mean absolute percentage error MAPE, maximum error MaxError, coefficient of determination R-squared. Use the letter to represent the weight of the accuracy index, and u takes values of 1, 2, 3, 4; represents the weight of the root mean square error RMSE in the phase; represents the weight of the mean absolute percentage error MAPE in the phase; represents the weight of the maximum error MaxError in the phase; phase; represents the weight of the root mean square error in the startup phase; represents the weight of the u-th accuracy index in the

[0031] The construction of the comprehensive evaluation matrix G for the accuracy of the observer can be expressed as: ; The stability indicators include the maximum convergence time MCT, the average convergence time ACT, the maximum chattering amplitude MJA, and the average chattering amplitude AJA. Similarly, use letters to represent the weights of the stability indicators, where z takes values of 1, 2, 3, 4; represents the weight of the maximum convergence time MCT in the represents the weight of the average convergence time ACT in the represents the weight of the maximum chattering amplitude MJA in the represents the weight of the average chattering amplitude AJA in the represents the weight of the maximum convergence time MCT in the startup stage; represents the weight of the z-th stability indicator in the

[0032] The construction of the comprehensive evaluation matrix D for the observer stability can be expressed as: ; In the embodiment, the specific weights are determined as follows: In the weights of the accuracy indicators: for the in the startup stage is 0.4, is 0.3, is 0.2, is 0.1.

[0033] The weight of the root mean square error in the startup stage is taken as 0.4. In the startup stage, the focus of the accuracy evaluation is usually to ensure that the error between the observer and the fuel cell system is minimized. RMSE is a commonly used evaluation index, which can represent the degree of difference between the observer output and the actual value. RMSE is given a higher weight because it penalizes the sum of squares of the results for the entire dataset, which is very important for detecting whether there are large errors in the startup stage.

[0034] The weight of the mean absolute percentage error in the startup stage is taken as 0.3. MAPE is a relative error metric, which can reflect the relative error degree between the observer output value and the actual value. In the startup stage, evaluating the percentage of the error helps to understand the accuracy of the observer, especially in the case of percentage errors. Giving MAPE a moderate weight of 0.3 is to balance the emphasis on percentage errors to measure the performance of the observer in the startup stage.

[0035] The weight of the maximum error in the startup phase is taken as 0.2. Max Error reflects the possible Max Error values that the observer may have in the startup phase. In this phase, the dynamic response of the system may cause instantaneous errors, and highly accurate measurement results are not necessarily required. Although Max Error may not have a significant impact on the overall performance, assigning a certain weight of 0.2 to Max Error helps to evaluate the accuracy and robustness of the observer.

[0036] The weight of the coefficient of determination in the startup phase is taken as 0.1. R-squared is an indicator that measures the proportion of the explainable part in the change of the observer output. In the startup phase, the goodness of fit of the observer can help determine whether it can accurately reflect the changes in the fuel cell system estimates. Although R-squared may be relatively low in the startup phase, a small weight of 0.1 is still assigned to R-squared to comprehensively consider the impact of the goodness of fit on the observer performance.

[0037] The weight assignment described in the evaluation of the observer performance in the startup phase is to balance factors such as accuracy, relative error, maximum error, and goodness of fit. By reasonably assigning weights, these factors can be comprehensively considered to determine whether the observer can provide accurate and robust current estimates in the startup phase.

[0038] The described steady-state phase is 0.3, is 0.4, is 0.2, is 0.1.

[0039] The weight of the root mean square error in the steady-state phase is taken as 0.3. RMSE is a useful indicator to measure the error between the estimate and the actual value. In the steady-state operation phase, it is desired that the observer can ensure the accuracy of the estimate. A lower RMSE value indicates that the observer can accurately track the changes in the fuel cell system. It is hoped that the observer can provide accurate estimates to ensure the stable operation of the system. Therefore, a higher weight is assigned to RMSE to emphasize the importance of accuracy.

[0040] The weight of the mean absolute percentage error in the steady-state phase is taken as 0.4. In the steady-state operation phase, the current value may have certain fluctuations, and MAPE can effectively measure the percentage error of the observer output relative to the actual value. A lower MAPE value indicates that the observer can provide more accurate estimates in the case of large current fluctuations. By assigning a higher weight to MAPE in the weight assignment, the importance of the observer providing accuracy during steady-state operation is emphasized.

[0041] The weight of the maximum error in the steady state stage is taken as 0.2. Monitoring Max Error can ensure that the observer provides sufficient safety protection during the steady state operation stage and promptly detects possible errors or abnormal conditions. Although RMSE and MAPE can provide a measure of the overall error situation, they ignore the anomalies at individual observation points. Assigning an appropriate weight to Max Error helps to ensure the sensitivity of the observer to abnormal conditions, thereby enhancing the safety of steady state operation.

[0042] The weight of the coefficient of determination in the steady state stage is taken as 0.1. R-squared is used to evaluate the degree of correlation between the observer output and the actual current. A high R-squared value indicates that the observer can better estimate the system state, enhancing the accuracy and reliability of the observer. Although R-squared may be relatively less important during the steady state operation stage, by allocating a small portion of the weight to R-squared, we can still consider the goodness of fit of the observer as part of the evaluation.

[0043] In the evaluation of the observer performance during the steady state operation stage, the reason for the weight division is to balance multiple indicators such as accuracy, safety, and goodness of fit. Such weight allocation can comprehensively consider various factors to ensure that the observer can provide accurate and safe predictions during the steady state operation and maintain relevance to the fuel cell system state.

[0044] In the variable load stage is 0.2, is 0.4, is 0.3, is 0.1.

[0045] The weight of the root mean square error in the variable load stage is taken as 0.2. In the variable load stage, the change in the current load may have a greater impact on the accuracy of the observer. Under variable load conditions, the observer needs to adapt to the change in the current load and estimate the state value of the fuel cell system as accurately as possible. As a commonly used error metric, RMSE can represent the degree of difference between the observer output and the actual value. Giving an appropriate weight of 0.2 to RMSE can reflect the attention to the adaptability and accuracy of the observer in the variable load stage.

[0046] The weight of the mean absolute percentage error in the variable load stage is taken as 0.4. In the variable load stage, relative error metrics such as MAPE are very important for the performance evaluation of the observer. The change in the current load may cause a change in the relative error between the observer output and the actual value. Therefore, by using MAPE as the evaluation index, we can better understand the relative accuracy of the observer under load changes. Giving a higher weight of 0.4 to MAPE reflects the degree of emphasis on the relative error evaluation under load changes.

[0047] The weight of the maximum error in the variable load stage is taken as 0.3. In the variable load stage, the Max Error of the monitoring observer can provide timely detection of abnormal situations and safety protection. Max Error reflects the possible Max Error value of the observer. When the system load changes, the monitoring of Max Error can help detect potential errors or abnormal situations. Giving the Max Error a moderate weight of 0.3 emphasizes the attention to the accuracy and safety of the observer.

[0048] The weight of the coefficient of determination in the variable load stage is taken as 0.1. R-squared measures the degree of correlation between the observer output and the actual current. In the variable load stage, by evaluating the relationship between the observer output and the actual value, the reliability and consistency of the observer can be understood. Although the weight of R-squared is relatively low at 0.1, it still comprehensively evaluates the performance of the observer.

[0049] Based on the above, the comprehensive evaluation matrix G of the observer accuracy is formed as .

[0050] In the index weights of stability: The weight of the stability in the startup stage is 0.35, is 0.25, is 0.2, is 0.2.

[0051] The weight of the maximum convergence time in the startup stage is taken as 0.35. MCT is the longest time required for the observer to reach the stable state from startup. Giving MCT a higher weight (0.35) is because in the startup stage, quickly reaching the stable state is crucial for the reliability and responsiveness of the system. A shorter MCT indicates that the observer can quickly adapt to system changes and provide accurate output.

[0052] The weight of the average convergence time in the startup stage is taken as 0.25. ACT is the average time required for the observer to reach the stable state in multiple startups. The weight of ACT being 0.25 reflects the stability of the average value. This index contributes to evaluating the overall performance of the observer in the startup stage. A shorter ACT indicates that the observer has consistent stability performance in different startup situations.

[0053] The weight of the maximum jitter amplitude in the startup stage is taken as 0.2. The jitter amplitude refers to the maximum oscillation amplitude of the observer output in the startup stage. Considering the anti-interference ability of the observer, the MJA index is of certain importance. Giving MJA a weight of 0.2 to evaluate the suppression effect of the observer on interference signals. A smaller MJA indicates that the observer has better anti-interference ability and can stably output.

[0054] The weight of the average chattering amplitude in the startup stage is taken as 0.2. AJA is the average value of the chattering amplitude of the observer in multiple startups. The weight given to AJA is 0.2 to ensure that the observer can suppress chattering under different startup conditions. A smaller AJA indicates that the observer has better stability and can reduce the fluctuations of the system output.

[0055] To sum up, such an allocation can comprehensively consider the stability, response speed and anti-interference ability of the observer, and ensure reliable output in the startup stage of the fuel cell.

[0056] The weight of stability in the steady state stage is 0.2, is 0.3, is 0.25, is 0.25.

[0057] The weight of the maximum convergence time in the steady state stage is taken as 0.2. MCT specifies the longest time required for the observer to reach the stable state from any initial state during the steady state operation stage. A relatively low weight of 0.2 is given to MCT because in the steady state stage, the focus is on ensuring that the observer can quickly stabilize, and a smaller MCT can provide a faster response speed and reliability.

[0058] The weight of the average convergence time in the steady state stage is taken as 0.3. ACT is the average time required for the observer to reach the stable state in multiple steady state startups. During the steady state operation stage, the weight of ACT is relatively high at 0.3 to ensure that the observer can have consistent stability under different operating conditions. A shorter ACT indicates that the observer can quickly adapt to system changes and provide a stable output.

[0059] The weight of the maximum chattering amplitude in the steady state stage is taken as 0.25. MJA indicates the maximum oscillation amplitude that occurs during the steady state operation of the observer. Giving MJA a moderate weight of 0.25 reflects the concern for the anti-chattering performance of the observer. A smaller MJA indicates that the observer has better ability to suppress chattering and interference.

[0060] The weight of the average chattering amplitude in the steady state stage is taken as 0.25. AJA is the average amplitude of chattering of the observer in multiple steady state startups. By giving AJA a weight of 0.25, the stability level of the observer under different operating conditions can be evaluated. A smaller AJA indicates that the observer can provide a stable output in the steady state stage.

[0061] The weight of stability in the variable load stage is 0.15, is 0.25, is 0.3, is 0.3.

[0062] The weight of the maximum convergence time in the variable load stage is taken as 0.15. In the variable load stage, the weight of MCT is 0.15, which means that even in the case of extreme load changes, the observer still needs to maintain good stability. The index of MCT pays more attention to the performance under abnormal conditions.

[0063] The weight of the average convergence time in the variable load stage is taken as 0.25. In the variable load stage, the stability and fast response of the observer become crucial. A higher weight of 0.25 is given to ACT because in the variable load stage, the observer needs to be able to quickly adapt to load changes and converge to a stable state to provide accurate output. The weight of ACT is 0.25, considering that both the stability and response speed of the observer are very important during the variable load process.

[0064] The weight of the maximum chattering amplitude in the variable load stage is taken as 0.3; the weight of the average chattering amplitude in the variable load stage is taken as 0.3. The weights of MJA and AJA are both 0.3, indicating that the anti-chattering and anti-interference capabilities of the observer are very important in the variable load stage. This is because load changes may cause system chattering and noise interference, and a high chattering amplitude may lead to unstable output. Therefore, in the variable load stage, the observer needs to have a lower chattering amplitude to ensure stable and accurate output.

[0065] In summary, the weight division is carried out according to the importance of the evaluation indexes of the observer stability under the variable load stage. A higher weight of ACT is required for the observer to quickly adapt to load changes and converge rapidly. The remaining weights consider the stability and anti-chattering ability of the observer in extreme cases. The division of these weights can help evaluate and optimize the performance of the observer in the variable load stage.

[0066] Based on the above, the comprehensive evaluation matrix D of the observer stability is formed as .

[0067] S03: Operate the fuel cell under different working modes, and every Obtain the numerical value of the accuracy index in the observer comprehensive evaluation index system, denoted as , where u takes 1, 2, 3, 4; among them: A1 is the root mean square error value, A2 is the mean absolute percentage error value, A3 is the mean absolute percentage error; A4 is the coefficient of determination; ; ; ; ; where n represents the number of samples collected, m represents the position of each observed value in the dataset, represents the actual value, Represents the observed value using the unscented Kalman filter observer, Represents the predicted value of the observed value using the unscented Kalman filter observer, Represents the average value of the actual value. k represents the number of times of data acquisition; t k Represents the time of the data acquired at the k-th time; t k+1 Represents the time of the data acquired at the (k + 1)-th time; Operate the fuel cell under different working modes, and every Obtain the numerical value of the stability index in the comprehensive evaluation index system of the observer, denoted as , where z takes 1, 2, 3, 4; among them: X1 is the maximum convergence time; X2 is the average convergence time; X3 is the maximum chattering amplitude; X4 is the average chattering amplitude; The numerical calculation formula of the stability index is as follows: ; ; ; ; Among them, Represents the time taken for data convergence in the d-th data acquisition, and e represents the total number of data acquisitions; Is the maximum value of the observed value using the unscented Kalman filter observer, Is the minimum value of the observed value using the unscented Kalman filter observer.

[0068] Normalize the values of the sub-indicators under different working modes and X z To obtain the accuracy normalization value And the normalization value of the sub-indicator of stability , specifically: The normalization value of the sub-indicator of accuracy ; The normalization value of the sub-indicator of stability .

[0069] Obtain the accuracy sub-indicators under the comprehensive evaluation index system of the observer in three stages respectively. Use superscripts 1, 2, and 3 to represent the three stages, where 1 represents the startup stage, 2 represents the steady-state stage, and 3 represents the variable load stage. Represents the value after normalization of the root mean square value after using the observer in the startup stage; specifically as follows: , , ; The normalized values of the sub-indicators of accuracy form the normalized matrix of the accuracy indicator : , the matrix is a 3X4 matrix.

[0070] Similarly, the stability sub-indicators under the comprehensive evaluation index system of the observer are obtained in three stages, respectively: , , ; The normalized values of the stability sub-indicators form the normalized matrix of the stability indicator : , the matrix is a 3X4 matrix.

[0071] S04: The comprehensive evaluation matrix G of the observer accuracy of the sub-indicators, the comprehensive evaluation matrix D of stability, and the obtained normalized matrix of accuracy and the normalized matrix of the stability sub-indicators are comprehensively evaluated to determine the accuracy scores of each sub-indicator and the stability scores under each working mode. That is: Determine the accuracy scores of each sub-indicator under each working mode; let , and a 3X4 matrix is obtained, where in the matrix the is denoted as the v-th column of the matrix. When v = 1, is the first column of the matrix; when v = 2, is the second column of the matrix; when v = 3, is the third column of the matrix; when v = 4, is the fourth column of the matrix.

[0072] = = ; Determine the stability scores of each sub-indicator under each working mode. Let , and a 3X4 matrix is obtained, where in the matrix the is denoted as the w-th column of the matrix. When w = 1, is the first column of the matrix; when w = 2, is the second column of the matrix; when w = 3, is the third column of the matrix; when w = 4, is the fourth column of the matrix.

[0073] = = ; S05: The accuracy scores of each sub-index in each working mode obtained by the SMO observer are denoted as ; the normalization matrix of the accuracy index in each working mode obtained by the SMO observer is denoted as ; the stability scores of each sub-index in each working mode obtained by the SMO observer are denoted as ; the normalization matrix of the stability index in each working mode obtained by the SMO observer is denoted as ; The combination weighting-multiplication synthesis method is used to determine the accuracy sub-index scores of the SMO observer in each working mode, that is, the accuracy score matrix of each sub-index in each working mode obtained by the SMO observer and the normalization value matrix of the observer accuracy are calculated by the combination weighting-multiplication synthesis method. The calculation results are shown in Table 1.

[0074] The combination weighting-multiplication synthesis method is used to determine the stability sub-index scores of the SMO observer in each working mode, that is, the stability score matrix of each sub-index in each working mode obtained by the SMO observer and the normalization matrix of the observer stability are calculated by the combination weighting-multiplication synthesis method. The calculation results are shown in Table 2.

[0075] The accuracy scores of each sub-index in each working mode obtained by the UKF observer are denoted as , and the normalization matrix of the accuracy index in each working mode obtained by the UKF observer is denoted as ; the normalization matrix of the stability index in each working mode obtained by the UKF observer is denoted as ; The combination weighting-multiplication synthesis method is used to determine the accuracy sub-index scores of the UKF observer in each working mode, that is, the accuracy score matrix of each sub-index in each working mode obtained by the UKF observer and the normalization matrix of the observer accuracy are calculated by the combination weighting-multiplication synthesis method. The calculation results are shown in Table 1.

[0076] The combination weighting-multiplication synthesis method is used to determine the stability sub-index scores of the UKF observer in each working mode, that is, the stability score matrix of each sub-index in each working mode obtained by the UKF observer Observer stability normalization matrix The combined weighting-multiplication synthesis method is used for calculation. The calculation results are shown in Table 2.

[0077] Table 1: Accuracy sub-index scores

[0078] In the table , the subscript 1 represents the root mean square error value of the accuracy sub-index; the superscript 1 represents the startup phase, and SMO represents the accuracy score obtained using the SMO observer. It can be represented by letters , which represents in the SMO observer The accuracy score of the $u$-th sub-index in the phase. Similarly , which represents in the UKF observer The accuracy score of the $u$-th sub-index in the phase.

[0079] Table 2: Stability sub-index scores

[0080] , the subscript 1 represents the maximum convergence time of the stability sub-index; the superscript 1 represents the startup phase, and SMO represents the accuracy score obtained using the SMO observer. It can be represented by letters , which represents in the SMO observer The stability score of the $z$-th sub-index in the phase. Similarly , which represents in the UKF observer The stability score of the $z$-th sub-index in the phase.

[0081] The calculated The comprehensive score using the SMO observer in the phase And the comprehensive score using the UKF observer , and the specific formulas are as follows: ; ; As shown in Table 3.

[0082] Table 3: Comprehensive scores

[0083] In the table, Represents the comprehensive score using the SMO observer in the startup phase.

[0084] S06: Establish a fusion observer control strategy based on the comprehensive scores $F$ of each observer under each working mode.

[0085] At In the [stage], compare the comprehensive score of the state estimation using the SMO observer with the comprehensive score of the state estimation using the UKF observer , if , and , then in the [stage], use the SMO observer as the observer for the fuel cell state estimation to conduct performance evaluation; if , and , and , then use the fusion observation strategy; If , and , then in the [stage], use the UKF observer as the observer for the fuel cell state estimation to conduct performance evaluation; if , and , and , then use the fusion observation strategy; a is the set value; Use the fusion observation strategy, as shown in Figure 2 below, specifically as follows: Calculate the accuracy score and stability score of using the SMO observer and the UKF observer respectively. The calculation formulas are as follows: ; ; In the formula: is the accuracy score of using the SMO observer in the [stage]; is the stability score of using the SMO observer in the [stage]; is the accuracy score of using the UKF observer in the [stage]; is the stability score of using the UKF observer in the [stage]; is the accuracy score of using the UKF observer in the [stage]; is the stability score of using the UKF observer in the [stage]; is the accuracy score of using the UKF observer in the [stage]; is the stability score of using the UKF observer in the [stage]; When > , and > , then use the SMO observer for performance evaluation; When < , and < , then use the UKF observer for performance evaluation; When > , and < , then use the SMO observer for accuracy performance evaluation and the UKF observer for stability performance evaluation; When < , and > If so, the UKF observer is used for accuracy performance evaluation, and the SMO observer is used for stability performance evaluation.

[0086] Embodiment S01: Division of working modes. According to the change of working current, the working modes are divided into a starting stage, a steady-state stage, and a variable load stage.

[0087] S02: Establish a comprehensive evaluation index system, including constructing a comprehensive evaluation matrix G for the accuracy of the observer and constructing a comprehensive evaluation matrix D for the stability of the observer; Among them, the comprehensive evaluation matrix G for the accuracy of the observer is: G = ; The comprehensive evaluation matrix D for the stability of the observer is D = ; S03: Operate the fuel cell with different hydrogen supply methods under different working conditions, and obtain the numerical values of the accuracy index and the stability index in the comprehensive evaluation index system of the observer every , and perform normalization processing on them.

[0088] At a certain moment, the normalized matrix of the accuracy and the normalized matrix of the stability of the SMO observer are obtained as follows: ; ; At a certain moment, the normalized matrix of the accuracy and the normalized matrix of the stability of the UKF observer are obtained as follows: ; ; S04: Conduct a comprehensive evaluation on the comprehensive evaluation matrix G of the accuracy of the observer of the sub-index, the comprehensive evaluation matrix D of the stability, the obtained normalized matrix J of the accuracy, and the normalized matrix H of the stability of the sub-index, and determine the accuracy score and the stability score of each sub-index under each working mode.

[0089] The accuracy scores and the stability scores of each sub-index under each working mode obtained through the SMO observer are respectively: = ; = ; Accuracy scores of each sub - index under each working mode obtained by the UKF observer and stability scores are respectively = ; = ; S05: Combine the accuracy scores and stability scores under each working mode and the normalization values of each observer, and use the combined weighting - multiplication synthesis method to calculate the accuracy scores, stability scores, and comprehensive scores of the sub - indexes of each observer under each working mode.

[0090] Use the combined weighting - multiplication synthesis method to determine the accuracy sub - index scores of the SMO observer under each working mode, that is, multiply the accuracy score matrix of each sub - index under each working mode obtained by the SMO observer by the observer accuracy normalization value matrix using the combined weighting - multiplication synthesis method; use the combined weighting - multiplication synthesis method to determine the accuracy sub - index scores of the UKF observer under each working mode, that is, multiply the accuracy score matrix of each sub - index under each working mode obtained by the UKF observer by the observer accuracy normalization matrix using the combined weighting - multiplication synthesis method. The calculation results are shown in Table 4.

[0091] Table 4: Accuracy scores

[0092] Use the combined weighting - multiplication synthesis method to determine the stability sub - index scores of the SMO observer under each working mode, that is, multiply the stability score matrix of each sub - index under each working mode obtained by the SMO observer by the observer stability normalization matrix using the combined weighting - multiplication synthesis method. Use the combined weighting - multiplication synthesis method to determine the stability sub - index scores of the UKF observer under each working mode, that is, multiply the stability score matrix of each sub - index under each working mode obtained by the UKF observer by the observer stability normalization matrix using the combined weighting - multiplication synthesis method. The calculation results are shown in Table 5.

[0093] Table 5: Stability scores

[0094] Calculated Comprehensive score using the SMO observer in the startup stage and the comprehensive score using the UKF observer , as shown in Table 6.

[0095] Table 6: Comprehensive score

[0096] S06: Establish a fusion observer control strategy based on the comprehensive score F of each observer under each working mode.

[0097] In the startup stage, , and , then use the UKF observer for performance evaluation in this startup stage; In the steady state stage, , and , then use the fusion observation strategy: Calculate the accuracy score and stability score of using the SMO observer and the UKF observer respectively. The calculation formulas are as follows: ; ; That is, > , and < , then use the SMO observer for accuracy performance evaluation and the UKF observer for stability performance evaluation; In the variable load stage, , and , then use the fusion observation strategy: ; ; That is, > , and < , then use the SMO observer for accuracy performance evaluation and the UKF observer for stability performance evaluation.

[0098] A fuel cell state evaluation device, including a processor and a memory. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the fuel cell state estimation method as described above are run.

[0099] A storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the fuel cell state estimation method are run.

[0100] It should be understood that although this specification is described according to various embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other implementation manners that can be understood by those skilled in the art.

[0101] The series of detailed descriptions listed above are only specific descriptions of the feasible embodiments of the present invention, and they are not intended to limit the protection scope of the present invention. Any equivalent embodiments or modifications made without departing from the technical spirit of the present invention should be included within the protection scope of the present invention.

Claims

1. A fuel cell state estimation method, characterized in that, It includes the following steps: Work mode division: According to the change of working current, the work mode is divided into a startup stage, a steady-state stage, and a variable load stage; Establish a comprehensive evaluation index system, which includes a comprehensive evaluation matrix G for the accuracy of the observer and a comprehensive evaluation matrix D for stability; the comprehensive evaluation matrix G for the accuracy of the observer includes the weights of the accuracy indexes in different working stages, and the comprehensive evaluation matrix D for the stability of the observer includes the weights of the stability indexes in different working stages; Operate the fuel cell in different stages, obtain the values of the accuracy indexes and stability indexes in the comprehensive evaluation index system of the observer, and perform normalization processing; the values of the accuracy indexes include root mean square error value, mean absolute percentage error value, mean absolute percentage error, and coefficient of determination; the values of the stability indexes include maximum convergence time, average convergence time, maximum chattering amplitude, and average chattering amplitude; the normalized values of the sub-indexes of accuracy form a normalized matrix J of accuracy indexes, and the normalized values of the sub-indexes of stability form a normalized matrix H of stability indexes; The comprehensive evaluation matrix G of the observer accuracy of the sub-indicators, the comprehensive evaluation matrix D of the stability, and the obtained matrix with normalized accuracy And the matrix with normalized sub-indicators of stability Carry out a comprehensive evaluation to determine the accuracy scores of the sub-indicators under each working mode And the stability scores ; Use the combined weighting-multiplication synthesis method to determine the accuracy sub-index scores and stability sub-index scores of the SMO observer in each working mode; use the combined weighting-multiplication synthesis method to determine the accuracy sub-index scores and stability sub-index scores of the UKF observer in each working mode; Determine the comprehensive scores of using the SMO observer and the UKF observer in different stages; Based on the comprehensive scores of each observer in different stages, establish a control strategy for the fusion observer.

2. The fuel cell state estimation method according to claim 1, wherein The comprehensive evaluation matrix G of the observer accuracy is expressed as: , represents the weight of the u-th accuracy index in the stage; u = 1 represents the root mean square error; u = 2 represents the mean absolute percentage error; u = 3 represents the maximum error; u = 4 represents the coefficient of determination; , represents the starting stage, represents the steady state stage, represents the variable load stage; The comprehensive evaluation matrix D of the observer stability is expressed as: , represents the weight of the z-th stability index in the stage; z = 1 represents the maximum convergence time; z = 2 represents the average convergence time; z = 3 represents the maximum chattering amplitude; z = 4 represents the average chattering amplitude.

3. The fuel cell state estimation method according to claim 1, wherein Operate the fuel cell at different stages to obtain the values of the accuracy index in the comprehensive evaluation index system of the observer , , where: A1 is the root mean square error value, A2 is the mean absolute percentage error value, A3 is the mean absolute percentage error; A4 is the coefficient of determination; ； ； ； ； where n represents the number of samples collected, and m represents the position of each observation in the dataset, represents the actual value, represents the observed value using the unscented Kalman filter observer, represents the predicted value of the observed value using the unscented Kalman filter observer, represents the average value of the actual values; k represents the number of times of data collection; t k represents the time of the data collected at the k-th time; t k+1 represents the time of the data collected at the (k + 1)-th time; Operate the fuel cell at different stages to obtain the numerical values of the stability indicators in the comprehensive evaluation index system of the observer , where: X1 is the maximum convergence time; X2 is the average convergence time; X3 is the maximum chattering amplitude; X4 is the average chattering amplitude; ； ； ； ； Among them, represents the time taken for data convergence in the d-th data acquisition, and e represents the total number of data acquisitions; is the maximum value of the observations using the unscented Kalman filter observer, is the minimum value of the observations using the unscented Kalman filter observer; The values of sub - indicators at different stages and are normalized. The normalized values of sub - indicators for accuracy form the accuracy index normalization matrix J, and the normalized values of sub - indicators for stability form the stability index normalization matrix , where: ； ； In the formula: represents the normalized value of the u-th accuracy index after using the observer in the represents the normalized value of the z-th stability index after using the observer in the 4. The fuel cell state estimation method according to claim 1, wherein Determine the accuracy scores of each sub-index under each working mode , ; The accuracy scores of each sub-index under each working mode are expressed as , which is expressed as the v-th column of the matrix; Determine the stability scores of each sub-index under each working mode , ; The accuracy scores of each sub-index under each working mode are expressed as , which is expressed as the w-th column of the matrix.

5. The fuel cell state estimation method according to claim 1, characterized in that Use the combined weighting-multiplication synthesis method to determine the accuracy sub-index scores and stability sub-index scores of the SMO observer in each working mode; use the combined weighting-multiplication synthesis method to determine the accuracy sub-index scores and stability sub-index scores of the UKF observer in each working mode. Specifically: The accuracy scores of each sub - index under each working mode obtained by the SMO observer are denoted as ; The normalization matrix of the accuracy index under each working mode obtained by the SMO observer is denoted as ; The stability scores of each sub - index under each working mode obtained by the SMO observer are denoted as ; The normalization matrix of the stability index under each working mode obtained by the SMO observer is denoted as ; The accuracy score matrix of each sub - index under each working mode obtained by the SMO observer and the observer accuracy normalization value matrix are calculated using the combined weighting - multiplication synthesis method to obtain the accuracy score of the \(u\) - th sub - index in the stage of the SMO observer, denoted as ; The stability score matrix of each sub - index under each working mode obtained by the SMO observer and the observer stability normalization matrix are calculated using the combined weighting - multiplication synthesis method to obtain the stability score of the \(z\) - th sub - index in the stage of the SMO observer, denoted as ; The accuracy scores of each sub-index under each working mode obtained by the UKF observer are denoted as , and the normalization matrix of the accuracy index under each working mode obtained by the UKF observer is denoted as ; the normalization matrix of the stability index under each working mode obtained by the UKF observer is denoted as ; the accuracy score matrix of each sub-index under each working mode obtained by the UKF observer and the observer accuracy normalization matrix are calculated using the combined weighting-multiplication synthesis method to obtain the accuracy score of the u-th sub-index in the stage of the UKF observer, denoted as ; the stability score matrix of each sub-index under each working mode obtained by the UKF observer and the observer stability normalization matrix are calculated using the combined weighting-multiplication synthesis method to obtain the stability score of the z-th sub-index in the stage of the UKF observer, denoted as .

6. The fuel cell state estimation method according to claim 5, characterized in that, Determine the comprehensive scores of using the SMO observer and the UKF observer in different stages. Specifically: According to the accuracy score of the u-th sub-index and the stability score of the z-th sub-index in the phase of the SMO observer, determine the comprehensive score of the SMO observer in the phase: ； According to the accuracy score of the u-sub-index in the UKF observer during the stage and the stability score of the z-sub-index determine the comprehensive score of the UKF observer in the ； In the formula: For the comprehensive score using the SMO observer in the For the comprehensive score using the SMO observer in the 7. The fuel cell state estimation method according to claim 1, characterized in that Based on the comprehensive scores of each observer in different stages, establish a control strategy for the fusion observer. Specifically: In the stage, compare the comprehensive score of the state estimation using the SMO observer with the comprehensive score of the state estimation using the UKF observer , If and , then in the phase, the SMO observer is used as the observer for fuel cell state estimation to evaluate the performance; If and , then use the fusion observation strategy; If and , then in the stage, the UKF observer is used as the observer for fuel cell state estimation to evaluate the performance; If and , then use the fusion observation strategy; a is a set value.

8. The fuel cell state estimation method according to claim 7, wherein The specific fusion observation strategy is: Calculate the accuracy scores and stability scores of using the SMO observer and the UKF observer respectively. The calculation formulas are as follows: ； ； Wherein: is the accuracy score of using the SMO observer in the phase; is the stability score of using the SMO observer in the phase; is the accuracy score of using the UKF observer in the phase; is the stability score of using the UKF observer in the phase; When > and > , the SMO observer is used for performance evaluation; When < and < , the UKF observer is used for performance evaluation; When > and < then use the SMO observer for accuracy performance evaluation and the UKF observer for stability performance evaluation; When < and > then use the UKF observer for accuracy performance evaluation and the SMO observer for stability performance evaluation.

9. A state evaluation device, characterized in that, It includes a processor and a memory, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the fuel cell state estimation method according to any one of claims 1-8 are run.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps in the fuel cell state estimation method according to any one of claims 1-8 are run.

Images