A multi-stage fuel cell on-board response method based on power prediction

Abstract

The application discloses a kind of fuel cell multi-section type vehicle-mounted response method based on power prediction.According to vehicle operating state and driving instruction analysis real-time demand power of whole vehicle, and combine power battery state and fuel cell state to carry out output power pre-distribution;With fuel cell output power sequence as main behavior sequence, whole vehicle output power sequence as related variable sequence, adopt NGBM model to predict subsequent output power of fuel cell.According to the difference between predicted power and current actual power, it is judged whether there is large load demand, and normal response mode or early response mode is selected;When early response, the latter response stage of current response period and the former response stage of next response period are continuously used for power regulation, and the power difference is absorbed or supplemented by power battery, so as to reduce the impact of fuel cell large load change and consider the dynamic response performance of whole vehicle.

Description

Technical Field

[0001] This invention relates to the field of fuel cell vehicle control and energy management technology, specifically a multi-stage on-board response method for fuel cells based on power prediction. Background Technology

[0002] Fuel cell vehicles typically employ a hybrid power system consisting of a fuel cell and a power battery. The fuel cell provides the primary energy, while the power battery handles instantaneous power compensation. Due to the dynamic response time inherent in the electrochemical reactions and gas supply processes of the fuel cell, its output power cannot quickly adapt to changes in the vehicle's power demand. When the vehicle experiences rapid acceleration or incline driving, a significant increase in load on the fuel cell within a short period can easily cause noticeable voltage fluctuations and exacerbate the performance degradation of components such as the catalyst layer and membrane electrode assembly, ultimately affecting the durability of the fuel cell system.

[0003] Current fuel cell vehicle energy management typically allocates power between the fuel cell and the battery based on the current power demand of the vehicle and the state of charge of the battery, adjusting the fuel cell output as the power demand changes. Due to the limitations of the fuel cell's permissible load variation rate, when the vehicle experiences significant power changes, simply increasing the fuel cell's instantaneous response speed is insufficient to simultaneously ensure both power performance and durability. Furthermore, frequent and large load variations may further shorten the fuel cell's lifespan. While the battery can withstand certain short-term power differences, it still requires careful coordination with the fuel cell's response timing.

[0004] Therefore, there is an urgent need for an on-board response method that can predict the subsequent power changes of fuel cells in advance and actively adjust the timing and response process of fuel cell power adjustment based on the prediction results. This would enable the fuel cell to adjust its power in advance before the actual arrival of large load changes, meet the power requirements of the vehicle without significantly increasing the load change rate of the fuel cell, and reduce the adverse effects of large and rapid load changes on the durability of the fuel cell.

[0005] A search revealed a Chinese patent document disclosing an energy management method for fuel cell buses based on short-term power smoothing prediction [Application No.: 202210869006.7, Publication No.: CN115230485B]. This patent collects the motor power demand of the bus from the vehicle's CAN bus, using the motor output power as a representation of the driving conditions. It first uses a first-order exponential smoothing method to process historical power demand, and then combines an LSTM neural network to predict short-term power demand. Subsequently, it establishes a multi-objective cost function that includes factors such as hydrogen consumption, power source lifespan, and power battery SOC trajectory, and uses a dynamic programming algorithm in the prediction time domain to obtain the optimal sequence of control variables to achieve energy allocation between the fuel cell and the power battery.

[0006] While the comparative patent achieves predictive energy management for fuel cell vehicles through short-term power prediction and features the ability to predict motor power demand using first-order exponential smoothing and LSTM, and optimize power allocation between the fuel cell and the power battery through dynamic programming, this invention does not primarily focus on energy allocation optimization within the prediction time domain. Instead, it uses the fuel cell output power sequence as the main behavior sequence and the vehicle output power sequence as the relevant variable sequence, directly predicting the subsequent output power of the fuel cell through an NGBM model. It then uses the change between the predicted power and the current actual power to determine large load changes. When a large load change condition is reached, the latter stage of the current response cycle and the former stage of the next response cycle are continuously used for advance power adjustment of the fuel cell, thereby gaining additional load change time. The power battery absorbs or supplements the temporary power difference formed by the advance adjustment. The comparative patent lacks this "power prediction—large load change threshold determination—multi-segment advance response timing switching" fuel cell vehicle response mechanism. Summary of the Invention

[0007] To address the problems existing in the prior art, the purpose of this invention is to provide a multi-stage on-board response method for fuel cells based on power prediction.

[0008] A multi-stage on-board response method for fuel cells based on power prediction, characterized by comprising: S1. Analyze the real-time power demand of the vehicle based on the vehicle's operating status and driving commands; S2. Based on the real-time power demand, power battery status and fuel cell status, pre-allocate the output power demand of the fuel cell and power battery. S3. Using the fuel cell output power sequence as the main behavior sequence and the vehicle output power sequence as the related variable sequence, the NGBM(1,2) model is used to obtain the predicted value of fuel cell power. S4. Calculate the magnitude of the difference between the predicted value and the current actual power, and compare it with the large load threshold. S5. Based on the comparison results, the direction of power change, and the load change capacity of the fuel cell, determine the multi-stage response mode; when it is determined that the load change needs to be done in advance, the current cycle's post-response stage and the next cycle's pre-response stage are used continuously for power regulation; otherwise, the post-response stage is used for power stabilization, and the power excess or deficiency is absorbed or supplemented by the power battery.

[0009] Preferably, in step S1, the vehicle operating status includes vehicle speed, and the driving command includes gear information, accelerator pedal information, and brake pedal information; In step S2, the state of the power battery includes the state of charge (SOC), dischargeable power, rechargeable power, and fault state; the state of the fuel cell includes the dischargeable power and fault state. The pre-allocation of the output demand power is aimed at enabling the fuel cell to operate in the high-efficiency range, and is carried out using fuzzy control methods, neural network methods, rule-based control strategies, optimization-based control strategies, or learning-based control strategies.

[0010] The above technical solution can rationally pre-allocate the vehicle's power demand based on the vehicle's current operating status, driving commands, and the real-time status of the power battery and fuel cell, so that the fuel cell and power battery can each undertake the output power demand that matches their working capacity, while making the fuel cell operate in the most efficient range.

[0011] Specifically, vehicle speed, gear information, accelerator pedal information, and brake pedal information reflect the vehicle's current operating status and the driver's driving or braking needs; the state of charge (SOC), dischargeable power, rechargeable power, and fault status of the power battery reflect its current charging and discharging capabilities; and the dischargeable power and fault status of the fuel cell reflect its current power output capabilities. Based on this, pre-allocating the output demand power can determine reasonable initial power requirements for the fuel cell and the power battery.

[0012] In practical applications, power pre-allocation can be performed using fuzzy control methods, neural network methods, rule-based control strategies, optimization-based control strategies, or learning-based control strategies, depending on the specific form of the vehicle control system. This ensures that the fuel cell operates within a high-efficiency range as much as possible and provides a power allocation basis for subsequent fuel cell power prediction, large load change judgment, and multi-stage response control.

[0013] Preferably, step S3 includes obtaining the output power of the fuel cell at the current moment and all previous moments to form the original output power sequence of the fuel cell: ; Obtain the output power of the vehicle at the current moment and all previous moments to construct the original output power sequence of the vehicle: ; in, This indicates the unaccumulated output power sequence of a fuel cell, indicated by the superscript. Represents the original sequence, subscript Indicates fuel cell; Indicates the first The actual output power of the fuel cell at each sampling time. The sampling time sequence number; This represents the unaccumulated sequence of vehicle output power, with subscripts indicating the order. Indicates the entire vehicle; Indicates the first The actual output power of the vehicle at each sampling time; This indicates the number of data points contained in each raw output power sequence, and In automotive applications The range is 4 to 10; the two original output power sequences use the same sampling time and the same number of data points, and are updated using a metabolic update method in each prediction; the NGBM(1,2) model can be extended to the NGBM(1,N) model to introduce at least one of vehicle speed and power battery state of charge (SOC) as additional related variable sequences.

[0014] The above technical solution can establish a corresponding power sequence by utilizing the actual output power of the fuel cell and the vehicle at multiple sampling times, both current and previous. This allows the prediction model to reflect both the power variation of the fuel cell itself and the impact of the vehicle's output power variation on the fuel cell's output power.

[0015] Specifically, the original output power sequence of the fuel cell is used as the main behavior sequence, and the original output power sequence of the vehicle is used as the relevant variable sequence. Both use the same sampling time and number of data points, so that the power data under the same sequence number corresponds to the same vehicle operating time. After each new power sampling is completed, an earlier data is removed and the latest data is added through a metabolic update method, so that the power sequence involved in the prediction always reflects the vehicle's operating status over the most recent period.

[0016] In practical applications, using 4 to 10 data points can leverage the grey prediction model's suitability for modeling with limited data, reducing the data requirements for on-board online computation. Furthermore, depending on the prediction needs, NGBM(1,2) can be extended to NGBM(1,N) to further incorporate relevant variables such as vehicle speed or battery SOC, improving the model's ability to represent changes in vehicle operating status.

[0017] Preferably, the original output power sequence of the fuel cell and the original output power sequence of the vehicle are respectively accumulated once to generate the following: ; ; as well as: ; ; in, This represents the fuel cell's first-accumulation sequence obtained after the original output power sequence of the fuel cell is accumulated once. (Superscript) This represents the sequence generated by a single accumulation. Indicates from the first sampling time to the... The cumulative value of the actual output power of the fuel cell at each sampling time; This represents the vehicle's first-accumulation sequence obtained after the original output power sequence of the vehicle is accumulated once. Indicates from the first sampling time to the... The cumulative value of the actual output power of the vehicle at each sampling time; ∑ represents the summation operation. This represents the sampling time sequence number during the summation process. This indicates the sampling time number currently participating in the accumulation calculation; the accumulation is used to reduce random fluctuations in the original output power sequence. And calculate the nearest neighbor mean of the first accumulation sequence of the fuel cell according to the following formula: ; in, Indicates the first The nearest neighbor mean of the fuel cell corresponding to the sampling time is derived from the first sampling time. The cumulative value of the fuel cell at each sampling time. The cumulative value of the fuel cell at the first sampling time. The arithmetic mean is obtained.

[0018] The above technical solution can reduce the random fluctuations in the original output power data of fuel cells and vehicles by generating them in one step, making the development trend contained in the original power sequence more obvious, and using the nearest mean to establish a data foundation suitable for subsequent grey model calculations.

[0019] Specifically, the output power of a vehicle during actual operation will constantly change with driving demands, and the original sampling sequence is prone to short-term random fluctuations. By accumulating the power data before each sampling time, the impact of random changes at a single sampling point on the overall sequence trend can be reduced; further, by calculating the average of two adjacent accumulated values, the nearest neighbor mean of the fuel cell's first accumulated sequence can be obtained.

[0020] In practical applications, the above-mentioned single-accumulation generation and adjacent mean calculation can both be completed using the existing power sampling data of the vehicle controller. This improves the regularity of the power data sequence without adding additional sensors, and provides a data foundation for the establishment of the whitening differential equation and parameter identification of the NGBM(1,2) model.

[0021] Preferably, the whitening differential equation of the NGBM(1,2) model is: ; in, Represents a continuous-time variable; This indicates the time sequence of a single accumulation of fuel cell data. The corresponding continuous-time function at that point; The continuous time function represents time. The rate of change; This indicates the cumulative sequence of the entire vehicle in time. The corresponding continuous-time function at that point; The evolution coefficient is used to characterize the evolution of the fuel cell output power sequence itself. These are the coefficients of relevant variables, used to characterize the weight of the influence of the vehicle's output power on the fuel cell's output power; , which is a nonlinear coefficient used to characterize the degree of influence of the nonlinear term in the fuel cell output power; It is the Bernoulli index, and ,when hour, This represents the nonlinear quadratic term of the cumulative power of the fuel cell in one operation. Discretizing the whitening differential equation yields: ; in, Indicates the first The actual output power of the fuel cell at each sampling time; This represents the product of the nearest neighbor mean and the development coefficient of a fuel cell. This represents the relevant variable terms after the cumulative power of the entire vehicle in one operation is weighted by the relevant variable coefficients; This represents the nonlinear term resulting from the calculation of the nearest neighbor mean of the fuel cell using the Bernoulli exponent and weighted by nonlinear coefficients.

[0022] The above technical solution can simultaneously consider the evolution law of fuel cell output power, the correlation between vehicle output power and fuel cell output power, and the nonlinear characteristics of fuel cell power change during the fuel cell power prediction process.

[0023] Specifically, the development coefficient Coefficients used to describe the development trend of fuel cell power sequences themselves. Used to describe the impact of the vehicle's output power on the fuel cell's output power; nonlinear coefficient. Bernoulli index This is used to describe the nonlinear variation of the fuel cell power sequence. By having the above parameters participate in the modeling, the model can simultaneously utilize information from the main behavior sequence and the related variable sequences.

[0024] In practical applications, after discretizing the whitening differential equation, the parameters can be directly identified using the fuel cell and vehicle power data collected by the vehicle controller according to the sampling period, making the NGBM(1,2) model adaptable to the requirements of on-board discrete data processing and online prediction.

[0025] Preferably, an output vector is constructed based on the discretized equation. Data matrix and the vector of parameters to be estimated : ; ; ; in, It is an n-1 dimensional column vector formed by sequentially arranging the original output power of the fuel cell from the 2nd to the nth sampling time. The data matrix contains n-1 rows and 3 columns. The first column of the data matrix is ​​the negative of the nearest neighbor mean of the fuel cell at each sampling time. The second column is the cumulative power of the entire vehicle at each sampling time. The third column is the nearest neighbor mean of the fuel cell at each sampling time. Power of; To include the development coefficient Coefficients of relevant variables and nonlinear coefficients The column vector of parameters to be estimated; The parameters are identified using the least squares method according to the following formula: ; in, Represents the vector of parameters to be estimated The estimated value; the superscript T indicates the matrix transpose operation; The product of the transpose of the data matrix and the data matrix is ​​represented; the superscript -1 indicates the matrix inversion operation; by Development coefficients were obtained respectively Coefficients of relevant variables and nonlinear coefficients The estimated value.

[0026] The above technical solution can transform the NGBM(1,2) discretization equation into a parameter identification problem based on vehicle historical power sampling data, and obtain the development coefficient, related variable coefficient and nonlinear coefficient corresponding to the current power sequence through the least squares method.

[0027] Specifically, data matrix It simultaneously incorporates the nearest neighbor mean of the fuel cell, the cumulative power of the entire vehicle, and the nonlinear term of the nearest neighbor mean of the fuel cell, enabling the parameter identification process to simultaneously reflect the power evolution of the fuel cell itself, the correlation with the vehicle power, and the degree of nonlinear change. Using the least squares method, the estimation result of the parameter vector to be estimated can be obtained based on existing sampled data.

[0028] In practical applications, when new power data is added to the original power sequence, the output vector and data matrix can be reconstructed based on the updated data, and the parameters can be re-identified so that the model parameters can be updated with the recent operating power status of the vehicle, providing a parameter basis for power prediction at the next sampling time.

[0029] Preferably, based on the development coefficient obtained from the identification Coefficients of relevant variables and nonlinear coefficients Solving the whitening differential equation yields the time response function of the fuel cell's cumulative power over a single charge: ; in, This represents the predicted cumulative power of the fuel cell at the (k+1)th sampling time, denoted by "". "Indicates the predicted value; This represents the first data point in the original output power sequence of the fuel cell; and For Bernoulli index A definite index; This represents the cumulative power of the entire vehicle at the (k+1)th sampling time. This represents the ratio of the coefficient of the relevant variable to the development coefficient; This represents the ratio of the nonlinear coefficient to the growth coefficient. Represents the natural constant; Indicated by the development coefficient and sampling time sequence number A defined index term; The predicted cumulative power of the fuel cell is restored by subtraction according to the following formula: ; in, This represents the predicted original output power of the fuel cell at the (k+1)th sampling time after cumulative subtraction and restoration. and These represent the predicted cumulative power of the fuel cell at the (k+1)th and kth sampling times, respectively; the initial conditions are: ; The initial condition indicates that the predicted cumulative power of the fuel cell at the first sampling time is equal to the actual output power of the fuel cell at the first sampling time. The power difference in step S4 is calculated according to the following formula: ; in, This represents the change in the predicted power value of the fuel cell at the next sampling time relative to the actual output power at the current sampling time. This represents the predicted output power of the fuel cell at the (k+1)th sampling time. Indicates the first The current actual output power of the fuel cell at each sampling time; when A positive value indicates a predicted load increase demand. A negative value indicates a predicted load reduction requirement, and the determination of whether to adjust the fuel cell output power in advance is based on the comparison between the power difference and the large load change threshold.

[0030] The above technical solution can restore the cumulative power prediction result calculated by the NGBM(1,2) model to the original output power prediction value of the fuel cell with actual physical meaning, and determine the direction and magnitude of subsequent fuel cell power changes by the difference between the prediction value and the current actual output power.

[0031] Specifically, the cumulative power prediction value at the next sampling time is obtained by solving the whitening differential equation, and then the original output power prediction value of the fuel cell is restored using the cumulative subtraction operation. When this occurs, it indicates that there is a predicted demand for increased fuel cell capacity in the future; when This indicates that there is a predicted demand for load reduction in fuel cells; at the same time, through... The magnitude of the change in power is used to determine whether the load condition has been met.

[0032] In practical applications, the above prediction results can form a judgment before the actual power demand occurs, enabling the vehicle controller to determine in advance whether the fuel cell should adopt a normal response mode or an early response mode, providing a direct basis for the selection of subsequent multi-stage response timing.

[0033] Preferably, the large load change threshold is determined by a dynamic load change test of the fuel cell based on the voltage drop amplitude and voltage recovery time during the fuel cell load increase process. During the fuel cell load ramp-up process, each response cycle is divided into a pre-response phase, a mid-response phase, and a post-response phase. The pre-response phase is used for normal power regulation, the mid-response phase is used for power stabilization, and the post-response phase is used for power stabilization when there is no need for early load change. When early load change is required, the post-response phase and the pre-response phase of the next response cycle are used together for continuous power regulation. The number of response phases and the duration of each response phase are adjusted according to actual needs. For rated power fuel cells, This indicates the rated power of the fuel cell. The response period is 9 seconds, with the durations of the pre-response, mid-response, and post-response phases each being 3 seconds. The load-up rate is 30 kW / s, the load-down rate is 60 kW / s, and the maximum load-up power achievable during the pre-response phase is 90 kW. The post-response phase of the current response cycle and the pre-response phase of the next response cycle constitute a continuous power regulation time of 6 seconds, enabling a maximum load-up of 180 kW. (From 20%) Load increased to 90% The time was 5.6 seconds, from 90% Reduce load to 10% The time is 3.2 seconds; this will exceed the fuel cell's rated power by 35%. The load change demand is determined as a large load change demand; if the load reduction rate can meet the load reduction demand within the pre-response phase, no early load change will be carried out; if the load reduction rate cannot meet the load reduction demand within the pre-response phase, the same early load change method as the load increase process will be adopted.

[0034] The above technical solution can determine the large load threshold that matches the actual dynamic performance of the fuel cell through dynamic load testing. When a large load demand is predicted, the post-response phase of the current response cycle can be used in advance for the next power change, so that it forms a continuous power adjustment time with the pre-response phase of the next response cycle, thereby increasing the available time for the fuel cell to complete a large power change.

[0035] Specifically, in normal response mode, the fuel cell performs normal power regulation in the first 3 seconds, and power stabilization in the middle and last 3 seconds. When a large load change demand is predicted, the last 3 seconds of the current cycle are converted from power stabilization time into advance regulation time, and connected with the first 3 seconds of the next cycle to form a 6-second continuous power regulation time. Taking a 240kW fuel cell as an example, it can complete a maximum load increase of 180kW within 6 seconds at a load increase rate of 30kW / s, while increasing the load from 48kW to 216kW only requires 168kW, which can be completed within 5.6 seconds.

[0036] In practical applications, this approach does not require simply increasing the instantaneous load-up rate of the fuel cell to meet high power demands. Instead, it obtains lead time through power prediction and extends the available time for load changes using a multi-stage response. Temporary power surplus during the advance load-up period can be absorbed by the power battery, while when the vehicle's power demand has increased but the fuel cell has not yet reached its target power, the power battery supplements the power deficiency, thus balancing fuel cell durability and vehicle dynamic response capabilities.

[0037] Compared with the prior art, the present invention has the following advantages: 1. This invention predicts the subsequent output power of the fuel cell and compares the predicted result with the current actual output power, enabling the vehicle controller to identify potential large power changes in advance. Compared to adjusting only after the actual power demand occurs, this method provides the fuel cell with a certain response lead time, allowing it to better match the vehicle's power requirements while maintaining a relatively smooth power change process, and helps reduce the adverse effects of frequent large load changes on the fuel cell's operating status.

[0038] 2. This invention divides the response cycle of the fuel cell into different response stages and switches between normal response and early response based on power prediction results. When a large power change demand occurs, the later response stage of the current cycle and the earlier response stage of the next cycle can be used continuously for power regulation, while the power battery absorbs or supplements the temporary power difference. This control method utilizes the respective response characteristics of the fuel cell and the power battery for coordinated regulation, which helps to balance the fuel cell's durability and the vehicle's dynamic response requirements. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the overall framework of the fuel cell vehicle energy management strategy of the present invention; Figure 2 These are test diagrams of the variable load characteristics of the fuel cell under different current loading rates according to the present invention; Figure 3 This is a schematic diagram of the fuel cell power prediction process based on the nonlinear grey Bernoulli prediction model NGBM of this invention; Figure 4 This is a schematic diagram of the overall process of the multi-stage on-board response method for fuel cells based on power prediction according to the present invention. Figure 5 This is a schematic diagram of the multi-segment on-board response timing of the normal response mode and the early response mode of the fuel cell of the present invention; Figure 6 This is a schematic diagram of the dynamic selection of multi-stage response modes of fuel cells and the power compensation process of the power battery in this invention. Detailed Implementation

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments. It should be understood that the following content is used to illustrate the implementation of the present invention and is not intended to limit the scope of protection of the present invention.

[0041] Working principle Overall Control Principles like Figure 1 As shown, this invention is based on the original vehicle energy management architecture of fuel cell vehicles, and introduces a "power prediction - large load change judgment - multi-stage response" control link between the power pre-allocation stage and the fuel cell power execution stage.

[0042] The system includes: a vehicle power demand analysis module to obtain the real-time power demand of the vehicle based on the vehicle's operating status and driving commands; a power pre-allocation module to preliminarily allocate the output power demand borne by the fuel cell and the power battery based on the real-time power demand of the vehicle, the fuel cell status, and the power battery status; a power prediction module to predict the power change trend of the fuel cell at the next moment based on the historical output power of the fuel cell and the historical output power of the vehicle; a large load change judgment module to determine whether the fuel cell output power needs to be adjusted in advance based on the prediction results; and a multi-stage response module to switch between normal response mode and advance response mode based on the judgment results.

[0043] Therefore, this invention does not cancel or replace the original power pre-allocation strategy of the vehicle, but rather adjusts the timing and response method of the fuel cell power command based on the original power pre-allocation results, so that the fuel cell can start power regulation before the actual arrival of large load changes.

[0044] After the fuel cell is pre-adjusted, there may be a temporary discrepancy between the actual output power of the fuel cell and the current power demand of the vehicle. When the power output of the fuel cell is higher than the current power demand of the vehicle, the power battery absorbs the power difference; when the output power of the fuel cell is temporarily lower than the power demand of the vehicle, the power battery makes up the power difference, thereby ensuring that the output power of the vehicle received by the motor controller can meet the driving needs of the vehicle.

[0045] Analysis principle of vehicle power demand The present invention first obtains the current operating status of the vehicle and the driving system control commands.

[0046] The current operating status of the vehicle includes at least the vehicle speed; the driving system control commands include at least gear information, accelerator pedal information, and brake pedal information. The vehicle power demand analysis module analyzes the above information to determine the vehicle's current required driving or braking power, thus obtaining the vehicle's real-time power demand.

[0047] To avoid confusion with the actual output power, in this embodiment, the real-time power demand of the vehicle is denoted as: ,in, This indicates the current power demand of the vehicle, obtained from the analysis of the vehicle's operating status and driving commands.

[0048] The specific analytical algorithm for the vehicle's required power can adopt the required power calculation method in the vehicle's original vehicle controller. The focus of this invention is not to limit the specific calculation model of the required power, but to use the real-time required power of the vehicle obtained by analysis to implement subsequent power pre-allocation, power prediction and multi-stage response control.

[0049] Power pre-distribution principle of fuel cells and power batteries The power pre-allocation module receives the real-time power demand of the entire vehicle. It also obtains the status of the power battery and the fuel cell.

[0050] The status of a power battery includes its state of charge (SOC), dischargeable power, rechargeable power, and fault status; the status of a fuel cell includes its dischargeable power and fault status.

[0051] The power pre-allocation module integrates the above information to obtain the initial output power demand of the fuel cell and the initial output power demand of the power battery. To distinguish between the demand power and the actual output power obtained from subsequent sampling, they are denoted as: ,and ,in, This indicates the output power demand allocated to the fuel cell by the power pre-allocation module; This indicates the output power required by the power pre-allocation module for the power battery.

[0052] Power pre-allocation can prioritize fuel economy as the primary control objective, ensuring that the fuel cell operates within its highest efficiency range. Specifically, it can employ fuzzy control methods, neural network methods, rule-based control strategies, optimization-based control strategies, or learning-based control strategies.

[0053] Power pre-allocation This refers to the initial power demand of the fuel cell. This initial power demand also needs to undergo power prediction, large load change determination, and multi-stage response processing according to the present invention to determine whether the fuel cell should adjust in a normal manner or prematurely enter a power regulation state.

[0054] Reasons for setting up predictive control and multi-stage response Fuel cells have relatively soft output characteristics, and their output power response speed is slower than that of power batteries. When fuel cells experience frequent and significant load changes, the electrochemical reaction rate, temperature, and gas supply conditions are prone to large fluctuations, which in turn cause output voltage fluctuations and accelerate the performance degradation of key components such as the catalyst layer and membrane electrode assembly.

[0055] Figure 2 The variable load characteristic test results show that as the fuel cell current loading rate increases, the fuel cell voltage fluctuation also increases. Therefore, simply increasing the fuel cell loading rate to meet instantaneous power demands is detrimental to fuel cell durability.

[0056] Power batteries are characterized by fast response speed and the ability to perform short-duration high-power charge and discharge, which can play a role in peak shaving and valley filling during periods of slow power change in fuel cells. In the disclosed scheme, the pulse discharge time of the power battery can typically reach 30-60 seconds. Therefore, this invention utilizes the power battery to handle short-duration power differences and uses power prediction to enable the fuel cell to start power changes in advance, thereby meeting the subsequent power demand of the vehicle without significantly increasing the fuel cell load change rate.

[0057] Its basic control idea is: instead of increasing the instantaneous load rate of the fuel cell to obtain dynamic response capability, it obtains the time advance by prediction, and then uses multi-stage response to extend the time when the fuel cell can be used to adjust power.

[0058] Fuel cell power prediction principle This invention uses the actual output power sequence of the fuel cell as the main behavior sequence and the actual output power sequence of the whole vehicle as the relevant variable sequence, and uses the nonlinear grey Bernoulli prediction model NGBM(1,2) to predict the output power of the fuel cell at the next sampling time.

[0059] In NGBM(1,2), "1" indicates the use of a first-order grey differential equation, and "2" indicates that the model includes a sequence of principal behaviors and a sequence of related variables. Compared with a univariate model that only uses the historical power of the fuel cell itself, introducing the vehicle output power sequence can reflect the impact of changes in vehicle power on changes in fuel cell power.

[0060] (1) Obtain the original power sequence Obtain the actual output power of the fuel cell at the current sampling time and at each sampling time before the current sampling time to construct the original output power sequence of the fuel cell: ; in: This represents the original output power sequence of the fuel cell; superscript This indicates that the sequence is the original sequence without any accumulation or generation processing; Subscript Indicates fuel cell; Indicates the first The actual output power of the fuel cell collected at each sampling time; Indicates the sampling time number in the original power sequence; This indicates the number of data points included in the original power sequence, and In automotive applications, You can choose 4 to 10.

[0061] Obtain the actual output power of the vehicle corresponding to the fuel cell power data to form the original output power sequence of the vehicle: ; in: This represents the original output power sequence of the entire vehicle; Subscript Indicates the entire vehicle; Indicates the first The actual output power of the whole vehicle collected at each sampling time; The original output power sequence of the vehicle and the original output power sequence of the fuel cell have the same number of data points and corresponding sampling times to ensure that the data with the same sequence number in the two sequences can represent the output state of the fuel cell and the output state of the vehicle at the same time.

[0062] After each new power sampling is completed, the two original power sequences are updated by removing the oldest historical data from the original power sequence and adding the most recently acquired data to the original power sequence, so that the original power sequences always remain unchanged. Each data point. In this way, the predictive model can be continuously updated online using the vehicle's recent operational data.

[0063] (2) Generate the original power sequence by accumulating it once. The original output power sequences of the fuel cell and the vehicle are accumulated once to generate the fuel cell accumulated sequence and the vehicle accumulated sequence, respectively. ; in: ; In the formula, Indicates from the first sampling time to the... The cumulative value of the actual output power of the fuel cell at each sampling time; ∑ represents the summation operation; Indicates the sampling time sequence number that participated in the accumulation; Indicates the current accumulated sampling time number; superscript This indicates that a sequence is generated by accumulating data once.

[0064] The cumulative sequence for the entire vehicle is as follows: ; in: ; In the formula, Indicates from the first sampling time to the... The cumulative value of the actual output power of the whole vehicle at each sampling time.

[0065] A single cumulative generation can reduce the random fluctuations in the original power data and make the changing trend of the original power sequence more obvious, so as to establish a grey prediction model.

[0066] (3) Generate the nearest neighbor mean sequence of fuel cells Calculate the nearest neighbor mean based on the first cumulative sequence of the fuel cell: ; in: Indicates the first The nearest neighbor mean of the fuel cell at each sampling time; Indicates the first The cumulative value of the fuel cell at each sampling time; This represents the cumulative value of the fuel cell at the (k-1)th sampling time.

[0067] The nearest neighbor mean is used to approximate the background value of the cumulative power of the fuel cell between two adjacent sampling times, and is used to convert the continuous whitening differential equation into a discrete form.

[0068] (4) Establish the whitening differential equation of NGBM(1,2) The whitening differential equation for the NGBM(1,2) model is: ; in: Represents a continuous-time variable; This represents the continuous-time function corresponding to a single accumulation sequence of a fuel cell; This represents the rate of change of the cumulative power of a fuel cell over time. The development coefficient is used to characterize the evolution of the fuel cell output power sequence itself. This represents the continuous-time function corresponding to a single cumulative sequence of the entire vehicle. The coefficients represent the relevant variables, used to characterize the weight of the influence of the vehicle's output power on the fuel cell's output power. This represents the nonlinear coefficient, used to characterize the degree of influence of the nonlinear variation term of fuel cell power on the model output; Represents the Bernoulli index, and In this embodiment, We can choose 2, at this point It is a nonlinear quadratic term.

[0069] In the above equation, Reflecting the trend of fuel cell power variation itself, This reflects the external impact of changes in overall vehicle power on fuel cell power. It reflects the nonlinear characteristics of the fuel cell power change process.

[0070] (5) Model parameter identification Discretizing the whitening differential equation, we get: ; In the formula: Indicates the first The actual output power of the fuel cell at each sampling time; This represents the product of the development coefficient and the nearest neighbor mean of the fuel cell. This represents the relevant variable term formed after weighting the cumulative power of the entire vehicle by the relevant variable coefficients; This represents the nonlinear term formed by calculating the nearest mean of the fuel cell using the Bernoulli exponent and then weighting it with nonlinear coefficients.

[0071] According to the 2nd to the 3rd The output vector is constructed from the data at each sampling time: ; in, It is a column vector containing n-1 elements, where each element represents the actual output power of the fuel cell at the 2nd to nth sampling time.

[0072] Constructing the data matrix: ; in, This is a data matrix with n-1 rows and 3 columns; the first column is the negative of the nearest neighbor mean of the fuel cell at each sampling time; the second column is the cumulative power of the entire vehicle at the corresponding sampling time; and the third column is the nearest neighbor mean of the fuel cell at the corresponding sampling time. Power of 1.

[0073] Construct the vector of parameters to be estimated: ; in, In order, they include the development coefficient. Coefficients of relevant variables and nonlinear coefficients .

[0074] The model parameters are estimated using the least squares method: ; in: Represents the parameter vector The estimated value; The superscript T indicates matrix transpose; This represents the product of the transpose of the data matrix and the data matrix. The superscript -1 indicates matrix inversion; Depend on Obtaining the development coefficient Coefficients of relevant variables and nonlinear coefficients The estimated value.

[0075] (6) Calculate the predicted power value of the fuel cell Solving the whitening differential equation based on the identified model parameters yields the time response function of the fuel cell's single-stage cumulative power: ; in: This represents the predicted cumulative power of the fuel cell at the (k+1)th sampling time. symbol" " indicates that the corresponding parameter is the model's predicted value; This represents the actual output power data of the first fuel cell in the current modeling sequence; This represents the cumulative power of the entire vehicle at the (k+1)th sampling time. This represents the ratio of the coefficient of the relevant variable to the development coefficient; This represents the ratio of the nonlinear coefficient to the growth coefficient. Represents the natural constant; Indicated by the development coefficient and sampling time sequence number The index terms that are jointly determined.

[0076] The cumulative power prediction value is then reduced and restored: ; in: This represents the predicted original output power of the fuel cell at the (k+1)th sampling time. This represents the predicted cumulative power of the fuel cell at the (k+1)th sampling time. This represents the predicted cumulative power of the fuel cell at the k-th sampling time.

[0077] The initial conditions are: ; That is, the predicted cumulative power of the fuel cell at the first sampling time is equal to the first actual output power data in the current modeling sequence.

[0078] Large load change judgment principle The power prediction value of the fuel cell at the next sampling time is compared with the current actual output power to obtain the power change: ; in: This represents the change in the predicted power value of the fuel cell at the next sampling time relative to the current actual output power; This represents the predicted output power of the fuel cell at the (k+1)th sampling time. Indicates the first The current actual output power of the fuel cell at each sampling time.

[0079] when When, it indicates that there is a predicted demand for increased fuel cell capacity; when This indicates that there is a predicted demand for load reduction in fuel cells.

[0080] Preset large load threshold The large load change threshold can be determined based on dynamic load change tests of the fuel cell. During the test, the voltage drop amplitude and voltage recovery time of the fuel cell under different loading amplitudes and loading rates are recorded, and the large load change threshold is determined according to the voltage fluctuation range that the fuel cell can withstand.

[0081] During the load increase process, if the positive power change exceeds the large load threshold, it is determined that the fuel cell output power needs to be adjusted in advance; if the positive power change is less than the large load threshold, it is determined that no adjustment is needed in advance, and the fuel cell operates in the normal response mode.

[0082] The purpose of the large load change determination is not to decide whether the fuel cell should respond to the vehicle's demand, but rather to determine whether the fuel cell should adopt a normal response timing or an advanced response timing. That is, regardless of whether the large load change threshold is exceeded, the fuel cell must respond to the power demand; the difference is that under large load change conditions, the power adjustment action is advanced to the later response stage of the current response cycle.

[0083] Multi-segment vehicle response principle like Figure 3 As shown, each response cycle of a fuel cell is divided into a pre-response phase, a mid-response phase, and a post-response phase.

[0084] Let a complete response cycle be The durations of the pre-response phase, the mid-response phase, and the post-response phase are as follows: , and ,but: ;

[0085] The pre-response stage is used for normal power regulation, the mid-response stage is used for power stabilization, and the post-response stage is selected for power stabilization or early power regulation based on the large load determination results.

[0086] (1) Normal response mode When the change in fuel cell power does not reach the large load change threshold, adopt Figure 3 The dynamic adjustment is shown as 1.

[0087] During the pre-response phase of the current response cycle, the fuel cell adjusts its output power according to the preset load change rate; after entering the mid-response phase, the fuel cell maintains its current output power; after entering the post-response phase, since the predicted power change in the next stage does not reach the large load change threshold, the fuel cell continues to maintain its current output power.

[0088] Therefore, in normal response mode, the post-response phase of the current response cycle is the power stabilization time, and the fuel cell will perform the next normal adjustment according to the new power demand in the pre-response phase of the next response cycle.

[0089] (2) Early response mode When the power change of the fuel cell reaches or exceeds the large load change threshold, the following measures are adopted: Figure 3 The dynamic adjustment is shown in Figure 2.

[0090] After the fuel cell completes normal power regulation in the pre-response phase and achieves power stabilization in the mid-response phase of the current response cycle, it begins the next power regulation ahead of schedule in the post-response phase of the current response cycle. After entering the next response cycle, the fuel cell continues to perform power regulation in the pre-response phase of the next response cycle.

[0091] Therefore, the post-response phase of the current response cycle is connected to the pre-response phase of the next response cycle to form a continuous power regulation time. Compared to starting regulation only in the pre-response phase of the next response cycle, the early response mode increases the time the fuel cell can use to complete large load changes, thus eliminating the need to increase the fuel cell load change rate to meet large power change demands.

[0092] (3) Response parameters of a 240kW fuel cell With rated power as: Taking a fuel cell as an example, the response period is set to 9 seconds, where: ; ; ; The fuel cell has a loading rate of 30 kW / s, so the maximum loading power that the fuel cell can achieve during the 3-second initial response phase is 90 kW.

[0093] When the early response mode is adopted, the last 3 seconds of the current response cycle and the first 3 seconds of the next response cycle together constitute a 6-second continuous power adjustment time, which can achieve a maximum continuous load increase of 180kW at a load increase rate of 30kW / s.

[0094] When the fuel cell is upgraded from 20% to 90% of its rated power (i.e., from 48kW to 216kW), the power change is 70% of the rated power. At an upgrade rate of 30kW / s, the upgrade time is 5.6s. With the early response feature, the fuel cell can begin upgrading earlier in the later response phase of the current response cycle, allowing the upgrade process to continue continuously across the current and next response cycles, reducing the instantaneous power gap when the actual power demand arrives.

[0095] The fuel cell has a deload rate of 60 kW / s. It takes 3.2 seconds to deload from 90% of the rated power to 10% of the rated power, i.e., from 216 kW to 24 kW. Since a pre-response phase lasts for 3 seconds, it can achieve a deload of close to 180 kW. Therefore, in this embodiment, the deload process does not need to employ pre-load change.

[0096] When the deload rate of the fuel cell in a practical application is low and the required deload cannot be completed within the normal adjustment phase, the same principle as the load increase process can be applied, using the post-response phase of the current response cycle and the pre-response phase of the next response cycle as the continuous deload time.

[0097] Power compensation principle of power battery When a fuel cell uses an advance response mode, its output power changes earlier than the actual power demand of the vehicle. Therefore, during the advance adjustment period, a temporary power difference will occur between the actual output power of the fuel cell and the real-time power demand of the vehicle.

[0098] When the fuel cell increases its load ahead of schedule but the vehicle's power demand has not yet increased, the power battery receives the excess power generated by the fuel cell; when the vehicle's power demand has increased but the fuel cell has not yet completed its load increase, the power battery outputs power to the drive system to supplement the fuel cell's insufficient power.

[0099] Through coordinated control between the fuel cell and the power battery, the fuel cell varies its power at a rate beneficial to durability, while the power battery undertakes the task of power balancing for short periods. This transforms the vehicle's rapid dynamic power demands into relatively gradual power changes in the fuel cell and short-duration charge-discharge processes in the power battery.

[0100] Online cyclic control principle The present invention performs the following control process cyclically during vehicle operation: First, real-time data collection is performed on gear position, pedal information, vehicle speed, power battery status, and fuel cell status, and the real-time power demand of the vehicle is analyzed. Second, power pre-allocation between the fuel cell and the power battery is performed. Then, the actual output power of the fuel cell and the actual output power of the vehicle are collected, and the data sequence of the NGBM(1,2) model is updated. Next, the model parameters are identified based on the updated data sequence, and the predicted fuel cell power value at the next sampling time is calculated. After that, the predicted value is compared with the current actual output power of the fuel cell to determine whether the large load threshold has been reached. Finally, based on the determination result, a normal response mode or an early response mode is selected, and the power battery compensates for the temporary difference between the fuel cell output power and the vehicle's power demand.

[0101] After completing one control cycle, the latest collected power data is re-input into the prediction model to enter the next prediction and judgment process, thus forming a closed-loop control process of "demand analysis - power pre-allocation - power prediction - large load judgment - multi-stage response - actual power feedback".

[0102] This invention provides advance time through power prediction, converts the advance time into the actual usable power adjustment time of the fuel cell through multi-stage response, and compensates for the power imbalance generated during the advance adjustment process through the power battery, thereby taking into account both the dynamic response requirements of the vehicle and the durability requirements of the fuel cell.

[0103] Example 1: Online prediction of fuel cell power and on-board response control This embodiment uses a 49-ton fuel cell heavy-duty truck as the implementation object. The vehicle's power system includes a fuel cell, a power battery, a motor controller, a vehicle controller, a fuel cell controller, and a power battery management system. The rated power of the fuel cell... The power output is 240kW, with the fuel cell's load-up rate set at 30kW / s and load-down rate set at 60kW / s.

[0104] The vehicle controller acquires the vehicle's current speed, gear information, accelerator pedal information, and brake pedal information, and analyzes the real-time power demand of the vehicle based on the above vehicle operating status and driving commands. At the same time, the vehicle controller acquires the state of charge (SOC), discharge power, rechargeable power, and fault status of the power battery, and also acquires the discharge power and fault status of the fuel cell.

[0105] The vehicle controller pre-allocates the output power of the fuel cell and the power battery based on the real-time power demand of the vehicle, the status of the power battery, and the status of the fuel cell, thus obtaining the required output power of the fuel cell. and the power output demand of the power battery During the power pre-allocation process, the fuel cell is prioritized to operate in its high-efficiency operating range, while the power battery covers the short-term power difference between the fuel cell's output power and the vehicle's real-time power demand.

[0106] This embodiment uses a nonlinear grey Bernoulli prediction model NGBM(1,2) to predict the power of the fuel cell at the next sampling time. The number of data points used in the model is n=4, and the Bernoulli exponent is... .

[0107] The actual output power of the fuel cell was continuously acquired at four sampling times to form the original output power sequence of the fuel cell: ; The actual output power of the vehicle at four consecutive sampling times corresponding to the actual output power of the fuel cell is acquired to form the original output power sequence of the vehicle: ; in, Indicates the first The actual output power of the fuel cell at each sampling time. Indicates the first The actual output power of the vehicle at each sampling time. .

[0108] The original output power sequences of the fuel cell and the original output power sequences of the vehicle are generated by summing them once each: ; ; in, ; Indicates the time from the first sampling moment to the second sampling moment. The cumulative value of fuel cell output power at each sampling time; Indicates the time from the first sampling moment to the second sampling moment. The cumulative value of the vehicle's output power at each sampling time.

[0109] Calculate the nearest neighbor mean based on the first cumulative sequence of the fuel cell: ; in, ; Indicates the first The mean value of the nearest neighbors of the fuel cell at each sampling time.

[0110] Based on the original output power of the fuel cell, the mean of the fuel cell's nearest neighbors, and the cumulative power of the entire vehicle, an output vector is constructed. and data matrix : ; ; The model parameter vector to be identified is: ; in, This is the development coefficient, used to reflect the evolution of the fuel cell output power sequence itself; These are the coefficients of relevant variables, used to reflect the weight of the influence of the vehicle's output power on the fuel cell's output power; This is a nonlinear coefficient used to reflect the degree of influence of the nonlinear term in the fuel cell output power.

[0111] Model parameters were identified using the least squares method. ; Based on the identified parameters, the predicted cumulative power of the fuel cell at the 5th sampling time is calculated according to the aforementioned NGBM(1,2) time response function. Then, the cumulative power is reduced to obtain the predicted output power of the fuel cell at the 5th sampling time. .

[0112] Calculate the predicted power change: ; in, This represents the change in the predicted output power of the fuel cell at the next sampling time relative to the current actual output power. When this occurs, it indicates that the fuel cell has a need for increased load capacity; when This indicates that there is a need to reduce the load on the fuel cell.

[0113] In this embodiment, the large load change threshold is set to 35% of the fuel cell's rated power, that is: ; in, This indicates the threshold for large load changes.

[0114] The vehicle controller selects either a normal response mode or an early response mode based on the comparison between the predicted power change and the large load threshold. After completing this prediction, it acquires the actual power data at the 5th sampling time and updates the original power sequence according to the metabolic update method. ; ; The updated power sequence is then used for the next fuel cell power prediction and multi-stage response determination, thus forming an online cyclic control process.

[0115] Example 2: Early response under high load conditions This embodiment uses the 49-ton fuel cell heavy-duty truck described in Embodiment 1. Fuel cell rated power... The power is 240kW, the fuel cell response cycle is set to 9s, and one response cycle is divided into a pre-response stage, a middle response stage and a post-response stage, with each response stage lasting 3s.

[0116] The pre-response phase is used for normal power regulation of the fuel cell, the mid-response phase is used for power stabilization of the fuel cell, and the post-response phase is selected for power stabilization or early power regulation based on the power prediction results.

[0117] The fuel cell is currently only 20% of its rated power, that is: ; Based on the NGBM(1,2) model prediction described in Example 1, the subsequent output power demand of the fuel cell is 90% of the rated power, i.e.: ; The predicted power change of the fuel cell is: ; The maximum load threshold in this embodiment is 84kW, because: ; Therefore, the vehicle controller determined that the fuel cell had a significant load increase requirement and executed an early response control.

[0118] During the first 3 seconds of the current response cycle, the fuel cell performs normal power regulation corresponding to the current cycle; during the middle 3 seconds of the current response cycle, the fuel cell maintains stable output power; when entering the last 3 seconds of the current response cycle, the vehicle controller no longer uses the latter response phase as the power stabilization time, but instead controls the fuel cell to perform the load-lifting action corresponding to the next response cycle in advance.

[0119] The fuel cell increases its load at a rate of 30 kW / s during the last 3 seconds of the current response cycle, and continues to increase its load at the same rate during the first 3 seconds of the next response cycle. The last 3 seconds of the current response cycle and the first 3 seconds of the next response cycle together form a continuous power regulation time of 6 seconds.

[0120] The last 3 seconds of the current response cycle and the first 3 seconds of the next response cycle form a continuous power adjustment time of 6 seconds. During this continuous power adjustment time, at a load rate of 30 kW / s, the maximum load power that the fuel cell can achieve is: ; in, This indicates the upload power that a fuel cell can achieve within a 6-second continuous power adjustment time.

[0121] To increase the fuel cell power output from 48kW to 216kW, the actual required power increase is: ; because: ; Therefore, the 6-second continuous power adjustment time is sufficient to meet the load increase requirements. The actual time required for the fuel cell to increase from 48kW to 216kW is: ; When the fuel cell output power reaches 216kW, it stops increasing the load and enters a stable power state.

[0122] During the 5.6s fuel cell load increase process, when the fuel cell loads ahead of schedule but the real-time power demand of the vehicle has not yet increased accordingly, the power battery absorbs the increased output power of the fuel cell. When the real-time power demand of the vehicle has increased but the fuel cell has not yet reached the target power of 216kW, the power battery supplements the power difference between the output power of the fuel cell and the real-time power demand of the vehicle.

[0123] After the fuel cell reaches the target power of 216kW, it enters a power stabilization state, and the power battery resumes the corresponding charging and discharging state according to the pre-allocation result of the vehicle power.

[0124] Through the above control, the fuel cell can start loading ahead of time by taking advantage of the post-response phase of the current response cycle without increasing the loading rate by 30kW / s, so that most of the loading process is completed before the actual high power demand of the vehicle arrives.

[0125] Example 3: Normal response under significant load reduction conditions This embodiment uses the 49-ton fuel cell heavy-duty truck described in Embodiment 1, with a fuel cell rated power of... The power is 240kW, and the unloading rate is 60kW / s.

[0126] The fuel cell is currently outputting 90% of its rated power, that is: ; Based on the NGBM(1,2) model, the subsequent output power demand of the fuel cell is predicted to decrease to 10% of the rated power, i.e.: ; The predicted power change of the fuel cell is: ; The negative sign indicates that the fuel cell needs to be deloaded, and the deload power amplitude is: ; Although the stated load reduction power amplitude is greater than the large load change threshold of 84kW, the fuel cell load reduction rate is 60kW / s, and the time required to complete the 192kW load reduction is: ; During the first 3 seconds of normal power adjustment in the response cycle, the load reduction power that the fuel cell can achieve is: ; After the initial 3 seconds of load reduction, the fuel cell's output power decreased from 216kW to: ; At this point, the difference between the fuel cell output power and the target power of 24kW is: ; The fuel cell continues to reduce the remaining 12kW at a rate of 60kW / s, and the time required is: ; The fuel cell reduced its power output from 216kW to 24kW in 3.2 seconds. Since the fuel cell can complete the load reduction within a timeframe close to a normal power regulation phase, this embodiment does not utilize the later response phase of the previous response cycle to implement early load reduction, but instead performs normal load reduction after the start of the current response cycle.

[0127] During fuel cell load reduction, when the fuel cell output power temporarily exceeds the vehicle's real-time power demand, the power battery reduces its discharge power or absorbs the excess output power of the fuel cell to maintain the power balance of the vehicle's DC bus. The fuel cell enters a stable power state after its output power decreases to 24kW.

[0128] When the actual load reduction rate of the fuel cell is lower than 60kW / s used in this embodiment, or when the predicted load reduction power cannot be completed within the normal power adjustment phase, the vehicle controller sets the post-response phase of the current response cycle as the early load reduction phase, and continuously uses the post-response phase of the current response cycle and the pre-response phase of the next response cycle for fuel cell load reduction.

[0129] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0130] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-stage on-board response method for fuel cells based on power prediction, characterized in that, include: S1. Analyze the real-time power demand of the vehicle based on the vehicle's operating status and driving commands; S2. Based on the real-time power demand, power battery status and fuel cell status, pre-allocate the output power demand of the fuel cell and power battery. S3. Using the fuel cell output power sequence as the main behavior sequence and the vehicle output power sequence as the related variable sequence, the NGBM(1,2) model is used to obtain the predicted value of fuel cell power. S4. Calculate the magnitude of the difference between the predicted value and the current actual power, and compare it with the large load threshold. S5. Determine the multi-stage response mode based on the comparison results, the power change direction, and the variable load capacity of the fuel cell; When it is determined that a load change needs to be made in advance, the current cycle's post-response phase and the next cycle's pre-response phase are used continuously for power regulation; otherwise, the post-response phase is used for power stabilization, and any excess or insufficient power is absorbed or supplemented by the power battery.

2. The fuel cell multi-segment on-board response method based on power prediction according to claim 1, characterized in that: In step S1, the vehicle operating status includes vehicle speed, and the driving command includes gear information, accelerator pedal information, and brake pedal information. In step S2, the state of the power battery includes the state of charge (SOC), dischargeable power, rechargeable power, and fault state; the state of the fuel cell includes the dischargeable power and fault state. The pre-allocation of the output demand power is aimed at enabling the fuel cell to operate in the high-efficiency range, and is carried out using fuzzy control methods, neural network methods, rule-based control strategies, optimization-based control strategies, or learning-based control strategies.

3. The fuel cell multi-segment on-board response method based on power prediction according to claim 1, characterized in that, Step S3 includes obtaining the output power of the fuel cell at the current moment and all previous moments, forming the original output power sequence of the fuel cell: ； Obtain the output power of the vehicle at the current moment and all previous moments to construct the original output power sequence of the vehicle: ； in, This indicates the unaccumulated output power sequence of a fuel cell, indicated by the superscript. Represents the original sequence, subscript Indicates fuel cell; Indicates the first The actual output power of the fuel cell at each sampling time. The sampling time sequence number; This represents the unaccumulated sequence of vehicle output power, with subscripts indicating the order. Indicates the entire vehicle; Indicates the first The actual output power of the vehicle at each sampling time; This indicates the number of data points contained in each raw output power sequence, and In automotive applications The range is 4 to 10; the two original output power sequences use the same sampling time and the same number of data points, and are updated using a metabolic update method in each prediction; the NGBM(1,2) model can be extended to the NGBM(1,N) model to introduce at least one of vehicle speed and power battery state of charge (SOC) as additional related variable sequences.

4. The fuel cell multi-segment on-board response method based on power prediction according to claim 3, characterized in that, The original output power sequence of the fuel cell and the original output power sequence of the vehicle are respectively accumulated once to generate the following: ； ； as well as: ； ； in, This represents the fuel cell's first-accumulation sequence obtained after the original output power sequence of the fuel cell is accumulated once. (Superscript) This indicates the sequence generated by a single accumulation. Indicates from the first sampling time to the... The cumulative value of the actual output power of the fuel cell at each sampling time; This represents the vehicle's first-accumulation sequence obtained after the original output power sequence of the vehicle is accumulated once. Indicates from the first sampling time to the... The cumulative value of the actual output power of the vehicle at each sampling time; ∑ represents the summation operation. This represents the sampling time sequence number during the summation process. This indicates the sampling time number currently participating in the accumulation calculation; the accumulation is used to reduce random fluctuations in the original output power sequence. And calculate the nearest neighbor mean of the first accumulation sequence of the fuel cell according to the following formula: ； in, Indicates the first The nearest neighbor mean of the fuel cell corresponding to the sampling time is derived from the first sampling time. The cumulative value of the fuel cell at each sampling time. The cumulative value of the fuel cell at the first sampling time. The arithmetic mean is obtained.

5. The fuel cell multi-segment on-board response method based on power prediction according to claim 4, characterized in that, The whitening differential equation of the NGBM(1,2) model is: ； in, Represents a continuous-time variable; This indicates the time sequence of a single accumulation of fuel cell data. The corresponding continuous-time function at that point; The continuous time function represents time. The rate of change; This indicates the cumulative sequence of the entire vehicle in time. The corresponding continuous-time function at that point; The evolution coefficient is used to characterize the evolution of the fuel cell output power sequence itself. These are the coefficients of relevant variables, used to characterize the weight of the influence of the vehicle's output power on the fuel cell's output power; is a nonlinear coefficient used to characterize the degree of influence of the nonlinear term in the fuel cell output power; It is the Bernoulli index, and ,when hour, This represents the nonlinear quadratic term of the cumulative power of the fuel cell in one operation. Discretizing the whitening differential equation yields: ； in, Indicates the first The actual output power of the fuel cell at each sampling time; This represents the product of the nearest neighbor mean and the development coefficient of a fuel cell. This represents the relevant variable terms after the cumulative power of the entire vehicle has been weighted by the relevant variable coefficients; This represents the nonlinear term resulting from the calculation of the nearest neighbor mean of the fuel cell using the Bernoulli exponent and weighted by nonlinear coefficients.

6. The fuel cell multi-segment on-board response method based on power prediction according to claim 5, characterized in that, Construct the output vector based on the discretization equation. Data matrix and the vector of parameters to be estimated : ； ； ； in, It is an n-1 dimensional column vector formed by sequentially arranging the original output power of the fuel cell from the 2nd to the nth sampling time. The data matrix contains n-1 rows and 3 columns. The first column of the data matrix is ​​the negative of the nearest neighbor mean of the fuel cell at each sampling time. The second column is the cumulative power of the entire vehicle at each sampling time. The third column is the nearest neighbor mean of the fuel cell at each sampling time. Power of; To include the development coefficient Coefficients of relevant variables and nonlinear coefficients The column vector of parameters to be estimated; The parameters are identified using the least squares method according to the following formula: ； in, Represents the vector of parameters to be estimated The estimated value; the superscript T indicates the matrix transpose operation; The product of the transpose of the data matrix and the data matrix is ​​represented; the superscript -1 indicates the matrix inversion operation; by Development coefficients were obtained respectively Coefficients of relevant variables and nonlinear coefficients The estimated value.

7. The fuel cell multi-segment on-board response method based on power prediction according to claim 6, characterized in that, Based on the identified development coefficient Coefficients of relevant variables and nonlinear coefficients Solving the whitening differential equation yields the time response function of the fuel cell's cumulative power over a single charge: ； in, This represents the predicted cumulative power of the fuel cell at the (k+1)th sampling time, denoted by "". "Indicates the predicted value; This represents the first data point in the original output power sequence of the fuel cell; and For Bernoulli index A definite index; This represents the cumulative power of the entire vehicle at the (k+1)th sampling time. This represents the ratio of the coefficient of the relevant variable to the development coefficient; This represents the ratio of the nonlinear coefficient to the growth coefficient. Represents the natural constant; Indicated by the development coefficient and sampling time sequence number A defined index term; The predicted cumulative power of the fuel cell is restored by subtraction according to the following formula: ； in, This represents the predicted original output power of the fuel cell at the (k+1)th sampling time after cumulative subtraction and restoration. and These represent the predicted cumulative power of the fuel cell at the (k+1)th and kth sampling times, respectively; the initial conditions are: ； The initial condition indicates that the predicted cumulative power of the fuel cell at the first sampling time is equal to the actual output power of the fuel cell at the first sampling time. The power difference in step S4 is calculated according to the following formula: ； in, This represents the change in the predicted power value of the fuel cell at the next sampling time relative to the actual output power at the current sampling time. This represents the predicted output power of the fuel cell at the (k+1)th sampling time. Indicates the first The current actual output power of the fuel cell at each sampling time; when A positive value indicates a predicted load increase demand. A negative value indicates a predicted load reduction requirement, and the determination of whether to adjust the fuel cell output power in advance is based on the comparison between the power difference and the large load change threshold.

8. The fuel cell multi-segment on-board response method based on power prediction according to claim 7, characterized in that, The large load change threshold is determined through dynamic load change tests of the fuel cell based on the voltage drop amplitude and voltage recovery time during the fuel cell load increase process. During the fuel cell load ramp-up process, each response cycle is divided into a pre-response phase, a mid-response phase, and a post-response phase. The pre-response phase is used for normal power regulation, the mid-response phase is used for power stabilization, and the post-response phase is used for power stabilization when there is no need for early load change. When early load change is required, the post-response phase and the pre-response phase of the next response cycle are used together for continuous power regulation. The number of response phases and the duration of each response phase are adjusted according to actual needs. For rated power fuel cells, This indicates the rated power of the fuel cell. The response period is 9 seconds, with the durations of the pre-response, mid-response, and post-response phases each being 3 seconds. The load-up rate is 30 kW / s, the load-down rate is 60 kW / s, and the maximum load-up power achievable during the pre-response phase is 90 kW. The post-response phase of the current response cycle and the pre-response phase of the next response cycle constitute a continuous power regulation time of 6 seconds, enabling a maximum load-up of 180 kW. (From 20%) Load increased to 90% The time was 5.6 seconds, from 90% Reduce load to 10% The time is 3.2 seconds; this will exceed the fuel cell's rated power by 35%. The load change demand is determined as a large load change demand; if the load reduction rate can meet the load reduction demand within the pre-response phase, no early load change will be carried out; if the load reduction rate cannot meet the load reduction demand within the pre-response phase, the same early load change method as the load increase process will be adopted.

Images