• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The present invention discloses an intelligent control system and method with real-time global optimization for a fuel cell bus, wherein at the fuel cell bus trip starting point, the vehicle driving communication unit downloads the prediction model parameters to the fuel cell vehicle controller VCU; and wherein the battery management system and the real-time engine power calculation module obtain the real-time SOC of the battery pack and the real-time power, respectively, and wherein the optimal SOC prediction model module obtains the predicted value of the optimal SOC reference trajectory of a next operating state segment, and wherein the MPC Prediction control module obtains the performance reference value, and the above parameters are respectively input to the fuel cell control unit to judge the working state of the fuel cell. In the driving process, the bus constantly uploads the segmented vehicle driving status information through the vehicle driving communication unit, after the completion of each trip, through the vehicle driving status information uploaded in real time, the cloud analysis workstation performs incremental learning and training to update the optimal SOC prediction model. The present invention can accurately and flexibly control the fuel cell bus in real time to reduce fuel consumption.
    公开号:CH717533B1
    申请号:CH01054/20
    申请日:2020-03-13
    公开日:2022-01-14
    发明作者:Hu Donghai；Wang Jing；He Hongwen；Yi Fengyan；Zhou Jiaming；Gao Jiangping；Li Zhongyan；Liu Xinlei
    申请人:Univ Jiangsu；
    IPC主号:
    专利说明:

    technical field
    The present invention relates to the technical field of energy management of new energy vehicles, in particular to an intelligent control system and method with real-time global optimization for a fuel cell bus.
    State of the art
    With the steady increase in the number of vehicles in China, the energy and environmental pressures of the automobile industry are also increasing. Due to the increasing external dependency on non-renewable energies such as oil, the implementation of energy substitution is extremely urgent; therefore, the hydrogen energy enters the public view with a high calorific value, an ample reserve and the excellent environmental friendliness; From the application of hydrogen, the fuel cell vehicle has become one of the major research directions, according to the statistics, over 40 automakers in the Chinese market are involved in the production of hydrogen fuel cell vehicles. On the other hand, the use of public transport also significantly reduces the pressure on energy and the environment, due to which it is imperative to spread the use of fuel cell electric buses in the market.
    Machine learning is a younger branch of artificial intelligence research and artificial intelligence science, its main research object is artificial intelligence, especially improving the performance of specific algorithms in empirical learning; incremental learning is a dynamic and step-by-step updating algorithm that refers to not recreating all knowledge bases when new data is added, but only training based on the original knowledge base for the updates caused by the new data will; this is also more in line with the principles of human thinking and can avoid repeated learning on massive data. In actual databases, the amount of data often increases gradually. Because of this, in the face of new data, the learning method is said to be able to make some changes to the trained system to learn the knowledge contained in the new data, and the time cost of changing a trained system is usually lower than the cost of retraining of a system.
    The prior art relates to an energy management method for plug-in hybrid vehicles based on deep-boosted learning, this invention has the following shortcomings: 1) the invention does not concern changing the rules involved due to data changes in the deep-boosted Learning is only about performing dimension reduction and fusion processing using the case of using massive data, when new data is added, a system should be retrained; 2) the invention does not concern a dynamically and stepwise updated algorithm, the data in the database change dynamically when there is new data, the learning method should be able to make certain changes for the trained system to learn the knowledge contained in the new data. The prior art further relates to an intelligent prediction-based energy management method for plug-in hybrid vehicles, this invention has the following shortcomings: 1) the use of deep learning for model prediction has a greater impact on the timeliness and accuracy of the database search, as a result, the forecast range can only be reached in the short term; 2) if there is a large difference from the target driving route, the model should be reconstructed.
    Because of this, it has become a relatively urgent research direction to reflect the changes in data conveniently and effectively. It is of great practical importance to develop an efficient, accurate and flexible intelligent control system and method with real-time global optimization for a fuel cell bus.
    content of the invention
    The present invention provides an intelligent control system and method with global real-time optimization for a fuel cell bus, wherein during the driving process of the fuel cell bus, the vehicle controller in combination with an optimal SOC prediction model controls the working state of the fuel cell in the event of new vehicle driving condition information performs retraining using the incremental learning model in the cloud workstation, updates the model parameters and timely downloads them to the VCU of the vehicle controller.
    The present invention comprises an intelligent control system with global real-time optimization for a fuel cell bus, a vehicle running communication unit, a vehicle running condition information prediction and analysis unit, a fuel cell vehicle control unit and a vehicle running condition information acquisition unit, wherein the fuel cell vehicle control unit has a signal connection with the prediction and a vehicle running state information analysis unit, and wherein the fuel cell vehicle control unit further forms a signal connection with the vehicle running state information acquisition unit; and wherein the vehicle running state information prediction and analysis unit acquires the prediction model parameters and downloads them to the fuel cell vehicle control unit to perform an update of the optimal SOC prediction model; and wherein the fuel cell vehicle control unit controls the working state of the fuel cell.
    In one embodiment, the fuel cell vehicle control unit controls the working state of the fuel cell based on the power reference value and the real-time power of the drive motor, the predicted value of the optimal SOC reference trajectory of a next operating state segment of the vehicle, and the real-time SOC value of the battery pack.
    In one embodiment, the real-time SOC of the battery pack is detected by the battery management system BMS in real-time.
    In one embodiment, the predicted value of the optimal SOC reference trajectory of a next operating state segment is obtained by the formula Y*=min+f(X)(max-min), where the optimal SOC prediction model module in the fuel cell vehicle control unit calculates the characteristic parameters of the current operating state segment, where f(X) is the regression function, max is the maximum value of the sample data, min is the minimum value of the sample data, and Y* is the predicted value of the optimal SOC reference trajectory of a next operating state segment.
    In one embodiment, the power reference value of the drive motor is calculated by the MPC prediction control module in the fuel cell vehicle control unit in accordance with the predicted value of the optimal SOC reference trajectory of a next operating state segment.
    In one embodiment, the real-time engine power is calculated by a real-time engine power calculation module.
    According to the invention, the prediction model parameters are obtained by transferring the characteristic parameters calculated by the characteristic parameter calculation module to the incremental learning model and training them; wherein the characteristic parameter calculation module receives the vehicle running condition information and optimal SOC reference trajectory sent by the speed information receiving module and the dynamic planning module.
    In one embodiment, the incremental learning model trains the characteristic parameters to obtain a regression model SVM1 and a support vector set SV1, integrating new vehicle driving condition information synthesized with the support vector set SV1 into the new data of the sample database and training continues to get a completely new regression model SVM2 and a completely new support vector set SV2 as the final model.
    An intelligent control method with global real-time optimization for a fuel cell bus includes the following steps: Step 1: Perform training for the obtained vehicle driving condition information and optimal SOC reference trajectory using the incremental learning model to obtain the parameters of the optimal SOC prediction model downloading them to the optimal SOC prediction model module to perform an optimal SOC prediction model update; Step 2: inputting the characteristic parameters of the current operating state segment into the optimal SOC prediction model module to output the predicted value of the optimal SOC reference trajectory of a next operating state segment; Step 3: inputting the predicted value of the optimal SOC reference trajectory of a next operating state segment to the MPC prediction control module to obtain the power reference value of the engine; Step 4: the fuel cell control unit FCU judges the SOC value and the output and controls the working state of the fuel cell; Step 5: the vehicle controller VCU judges whether the bus has completed a trip, if the trip is not completed, it returns to Step 2 and the process circulates; when the trip is finished, the cloud analysis workstation downloads the updated prediction model parameters again to the vehicle controller VCU to perform an update of the optimal SOC prediction model.
    In particular, step 4 is preferably as follows: judging whether the real-time SOC value of the battery pack is greater than the maximum value SOCmax of the predicted value of the optimal SOC reference trajectory of a next operating state segment, if SOC>SOCmaxist, it is judged whether the real-time power P is smaller than the power reference value Pmin, if P < Pmin, the fuel cell control unit FCU is turned off and charging is stopped; if P≥Pmin, it is further judged whether the real-time power P is greater than the power reference value Pmax, if P>Pmax, the fuel cell control unit FCU starts working, if Pmin≤P≤Pmax, the state of the fuel cell control unit FCU remains unchanged; if SOC < SOCminor SOCmin≤SOC≤SOCmax. and P > Pmaxist, the fuel cell control unit FCU enables the fuel cell stack to be switched off and discharged, otherwise the state remains unchanged.
    The present invention has the following advantages: 1) Avoiding repeated learning on massive data. When adding new data, there is no need to rebuild all knowledge bases, only training for the updates caused by the new data is performed based on the original model. 2) Saving cloud storage space and reducing costs. Due to the use of incremental learning, the original training data can be deleted after the training is completed to avoid data accumulation and achieve cost savings. 3) Constantly improve the operating condition adaptability. After the vehicle completes a trip, the cloud performs incremental training on the new data and downloads the trained model to the vehicle, continuously improving the vehicle's operating state adaptability over the same commute.
    Brief description of the drawings
    Figure 1 shows a structural view of an intelligent control system with global real-time optimization for a fuel cell bus according to an embodiment of the present invention. FIG. 2 shows a schematic structural view of a fuel cell bus according to an embodiment of the present invention. Figure 3 shows a schematic diagram of the control principle for the entire vehicle fuel cell bus according to an embodiment of the present invention. FIG. 4 shows a schematic diagram of the generation principle of an optimal SOC prediction model according to an embodiment of the present invention. FIG. 5 shows a structural view of a support vector machine-based incremental learning model according to an embodiment of the present invention. FIG. 6 shows a workflow diagram of an intelligent controller with global real-time optimization for a fuel cell bus according to an embodiment of the present invention. Figure 7 shows a case analysis diagram of a sudden operating condition of an embodiment of the present invention.
    Reference List
    1 Fuel cell bus 2 Vehicle controller VCU 3 Wireless communication system 4 Satellite 5 Base station 6 Wire communication system 7 Cloud analysis workstation 8 Incremental learning model 9 Optimal SOC prediction model module 10 MPC prediction control module 11 Fuel cell control unit FCU 12 Fuel cell hydrogen storage tank 13 Fuel cell stack 14 Speed sensor 15 Acceleration sensor 16 Drive motor real-time power calculation module 17 Motor controller MCU 18 Drive motor 19 Battery management system BMS 20 Speed information receiving module 21 Dynamic planning module 22 Characteristic parameter calculation module
    Detailed Embodiments
    In connection with figures, the structure and the working principle of an intelligent control system and method with global real-time optimization for a fuel cell bus according to the present invention are explained in more detail below.
    As shown in Figures 1 and 2, an intelligent control system with global real-time optimization for a fuel cell bus according to an embodiment of the present invention, comprising a vehicle running communication unit, a vehicle running state information prediction and analysis unit, a fuel cell vehicle control unit and a vehicle running state information acquisition unit. The vehicle driving communication unit comprises a wireless communication system 3, a satellite 4, a base station 5 and a wired communication system 6, wherein the vehicle driving condition information prediction and analysis unit comprises a cloud analysis workstation 7, and the fuel cell vehicle control unit comprises a vehicle controller VCU 2, a fuel cell control unit FCU 11, a fuel cell hydrogen storage tank 12, a fuel cell stack 13, a motor controller MCU 17 and a drive motor 18, and wherein the fuel cell hydrogen storage tank 12 provides the fuel to the fuel cell stack 13; and wherein the vehicle running condition information acquisition unit comprises a speed sensor 14, an acceleration sensor 15 and a real-time drive motor power calculation module 16, and wherein the speed sensor 14 and the acceleration sensor 15 each form a signal connection with the vehicle controller VCU 2, and the real-time drive motor power calculation module 16 forms a signal connection with the fuel cell control unit FCU 11 and the engine control unit MCU 17 .
    As shown in Figure 2, the vehicle controller VCU 2, the fuel cell control unit FCU 11, the drive motor 18, the motor controller MCU 17, the speed sensor 14, the acceleration sensor 15, the real-time power calculation module of the drive motor 16, the fuel cell hydrogen storage tank 12 and the fuel cell stack 13 are arranged at the upper part of the fuel cell bus 1, respectively.
    While the vehicle is driving, the vehicle controller VCU 2 is connected to the satellite 4 through the wireless communication system 3, the satellite 4 being connected to the base station 5 through the wireless communication system 3, and the base station 5 through the wired communication system 6 connected to the cloud analysis workstation 7; As shown in Figure 4, in the cloud analysis workstation 7, a speed information receiving module 20 and dynamic planning module 21 connected to each other are arranged, wherein the speed information receiving module 20 and the dynamic planning module 21 are respectively connected to the characteristic parameter calculation module 22, and the characteristic parameter calculation module 22 is connected to the incremental learning model 8; at the driving start point, the fuel cell bus 1 downloads the prediction model parameters to the vehicle controller VCU 2 through the vehicle driving communication system, wherein in the vehicle controller VCU an interconnected optimal SOC prediction model 9 and MPC prediction control module 10 are arranged, at the end the fuel cell control unit FCU 11 integrates the power reference value and the Real-time performance of the drive motor 18 with each other and outputs the operating state of the fuel cell.
    As shown in Figure 3, the vehicle controller VCU 2 comprises an optimal SOC prediction model module 9 and an MPC prediction control module 10, which are sequentially connected; wherein the module of the optimal SOC prediction model 9 receives the characteristic parameters of the current operating state segment in order to obtain a predicted value of the optimal SOC reference trajectory of a next operating state segment with the formula Y*=min+f(X)(max-min) and this ans MPC prediction control module 10 , based on this, the MPC prediction control module 10 determines the power reference value of the drive motor, which is transmitted to the fuel cell control unit FCU 11 . The battery management system BMS 19 detects the SOC value of the battery pack in real time, with the calculation module of the real-time power of the drive motor 16 forming a signal connection with the motor control MCU 17, and with the motor control MCU 17 detecting the speed and torque of the drive motor 18 in order to thus to get the real-time performance; and wherein the fuel cell control unit FCU 11 uses the power reference value and the real-time power of the drive motor 18, the predicted value of the optimal SOC reference trajectory of a next operating state segment and the real-time SOC value of the battery pack to control the turning on, turning off and maintaining of the fuel cell.
    As shown in Figure 4, the speed information receiving module 20 in the cloud analysis workstation 7 sends the vehicle driving condition information to the dynamic planning module 21 to obtain an optimal SOC reference trajectory; wherein the speed information receiving module 20 and the dynamic planning module 21 then send the vehicle running state information and the optimal SOC reference trajectory to the characteristic parameter calculation module 22, and wherein the characteristic parameter calculation module 22 transmits the calculated characteristic parameters to the incremental learning model 8 to obtain the prediction model parameters (including dual parameters α, α*, RBF core function, deviation b) to obtain; and wherein the fuel cell bus 1 downloads the parameters of the trained prediction model to the optimal SOC prediction model module 9 in the vehicle controller VCU 2 at the driving start point through the vehicle driving communication system; and in the running process of the vehicle, the vehicle controller VCU 2 divides the operating state into equal operating state segments TS in accordance with the time segments and uploads them to the speed information receiving module 20 . After the bus 1 completes each trip, using the operating state data of a new trip (including the speed and acceleration) and its optimal SOC reference trajectory, it acquires the characteristic parameters to perform incremental learning, with the bus generating new prediction model parameters and applying them to downloads the optimal SOC prediction model module 9 to update the optimal SOC prediction model therein.
    After the cloud analysis workstation 7 undergoes training by the incremental learning model 8 each time, since the incremental learning technology is used, the original training patterns can be deleted to result in a large amount of data collection and an increase in the training load and production costs avoid.
    As shown in Figure 5, the incremental learning model 8 trains the characteristic parameters to obtain a regression model SVM1 and a support vector set SV1, integrating new vehicle driving condition information synthesized with the support vector set SV1 into the new data of the sample database, and where training is continued to obtain a completely new regression model SVM2 and a completely new support vector set SV2 as final model. Using the characteristic parameter of the current operating state segment as the input of the incremental learning model and using the characteristic parameter of the optimal SOC reference trajectory of a next operating state segment as the output of the model, using a large amount of the inputs and outputs, the incremental learning model is trained to form a mapping relationship between the characteristic parameters of the current operating state segment and the characteristic parameters of the optimal SOC reference trajectory of a next operating state segment, thereby realizing an estimation and prediction for the characteristic parameters of the optimal SOC reference trajectory of a next operating state segment. The optimal SOC reference trajectory of a next operating state segment Yi+1 can be expressed as follows:
    The parameters in the above formula each represent several features of the optimal SOC reference trajectory of the operating state segment: maximum SOC value, minimum SOC value, SOC standard deviation, maximum SOC rate of change and average SOC.SOCmax= max:SOCj(2)SOCmin= min:SOCj(3) Kmax= max: Kj(7)
    Where n stands for the number of data points in the operating state segment, Δt for the data point time interval and K for the SOC change rate, j=1.2,...,n,.
    In order to realize the prediction of the optimal SOC reference trajectory of a next operating state segment, the characteristic parameters of the current operating state segment are recorded as follows:
    The parameters in the above formula each represent the vehicle running condition information of the current running condition segment: maximum speed, minimum speed, maximum acceleration, minimum acceleration and average speed.Vmax= max: Vj(9) amax= max: aj(11)amin= min: aj(12)
    Here V stands for the speed of the Hrennstott cell bus 1 and a for the acceleration of the fuel cell bus 1.
    Before training the incremental learning, the dynamic planning module 21 shall be used to calculate the road spectrum information of the vehicle and thus obtain the optimal SOC reference trajectory. The dynamic planning mainly includes the recursive method and the inverse method, in the recursive method, from the first stage, a front-to-back recursion is carried out by means of a state transition equation, the principle is as follows:
    Here, k stands for the stage number, Sk for the state variable, uk for the control variable, rk for the stage index function, fk for the optimal index function and Tk for the state transition function.
    Using power battery SOC as a state variable, the total stride length is divided into steps of m number and 1 s step length. The stage index function rk is the energy consumption of the kth stage, calculated as follows:
    where f(Pfc) stands for the energy consumption of the fuel cell at an output power Pfc; where is the equivalent energy consumption of the power battery; and where Z includes the dynamic working efficiency of the fuel cell, the DC-DC converter and the power battery, which can be calculated by experiments or a model of the equivalent circuits.
    The state transition equation from the kth stage to the k+1 stage is:
    Pb_k stands for the output power of the power battery, Ub for the bus voltage and Cb for the power battery capacity. The control parameter is the output power Pfc_k of the fuel cell in the kth stage, and the constraints of the state variables and control variables are:
    Here, Pfc_max represents the maximum output power of the fuel cell, and Pb_min and Pb_max represent the maximum charge and discharge power of the power battery; the optimization goal is to find the optimal control variable Pfc_k throughout the driving cycle such that the energy consumption J is minimal:
    The state variable SOC is divided into N nodes in the range of SOCmin and SOCmax in accordance with a certain step length, and each node stores the optimal trajectory with which the node is reached. The calculation process of node i in the kth step is as follows: first find out all the nodes that can be transferred to node i in the k-1th step under constraints and calculate the cumulative energy consumption fk of these state transitions such that the minimum state transition is the optimal strategy with which the kth step traverses node i. Recursion to the endpoint of the cycle, finding the node with minimum fk, where the optimal SOC reference trajectory can be found out through the trajectory information stored by the node.
    Using the characteristic parameter of the current operating state segment X as the input of the incremental learning model and using the optimal SOC reference trajectory of a next operating state segment Yi+1 as the output of the incremental learning model. In order to reduce the network prediction error caused by a relatively large difference in magnitude between the input data, the input data is normalized and the range of values of the normalized data is [0,1]. With the dispersion standardization method, the normalization is performed, the conversion formula is as follows:
    Here, max stands for the maximum value of the sample data, min for the minimum value of the sample data, X for the original training data (including a large number of characteristic parameters of the current operating state segment Xi), and X* for the normalized data.
    After using the SVM model to predict the optimal SOC reference trajectory of a next operating state segment, it is also necessary to denormalize the prediction results with equation (3) so that the predicted data corresponds to the actual range and importance. Correspondingly, the expression of the optimal SOC reference trajectory of a next operating state segment is as follows:
    In the formula, φ is a non-linear mapping from the input space to the high-level feature space; the weight Wi and the deviation b are obtained by the following formula:
    In the formula, W={W1,W2,..,Wi,...WN}, are the regularization parts, in the second point, the empirical risk measured by the insensitive loss function Lε in the following formula is ε is the maximum error allowed for the regression; C is the weight parameter used to balance the two and is called the regularization parameter.
    To obtain Wi and b, equation (6) is transformed by the RBF kernel function K(Xi,Xj) to:
    In the formula, α and α* are the dual parameters.
    Because of this, the regression function is converted to the following precise form:
    After the prediction output of the incremental learning model is obtained, the predicted value is denormalized and reset to the predicted value of the optimal SOC reference trajectory of a next operating state segment:Y* = min + ƒ(X)(max-min) (25)
    See Figure 6, an intelligent control method with a global real-time optimization for a fuel cell bus, comprising the following steps: Step 1: The speed sensor 14 and the acceleration sensor 15 first detect the vehicle driving operating status information of the fuel cell bus 1 and sends it to the vehicle controller VCU 2 , the vehicle controller VCU 2 sends it to the speed information receiving module 20, and the speed information receiving module 20 transmits it to the dynamic planning module 21 to obtain an optimal SOC reference trajectory of a next operating state. The characteristic parameter calculation module 22 receives the vehicle running state information sent by the speed information receiving module 20 and the dynamic planning module 21 and optimal SOC reference trajectory of a next running state, and calculates their characteristic parameters, respectively. Step 2: Perform training for the obtained vehicle driving condition information and optimal SOC reference trajectory by means of the incremental learning model 8 to obtain the parameters of the optimal SOC prediction model and download them to the optimal SOC prediction model module 9 . Step 3: input the characteristic parameters of the current operating state segment into the optimal SOC prediction model module 9 to output the predicted value of the optimal SOC reference trajectory of a next operating state segment; Step 4: Inputting the predicted value of the optimal SOC reference trajectory of a next operating state segment into the MPC prediction control module 10 and outputting the corresponding speed and torque, so as to obtain the reference value for the power P, which contains a maximum value Pmax and a minimum value Pmin. Step 5: The battery management system BMS 19 monitors the real-time SOC value of the battery pack and compares it with the predicted value of the optimal SOC reference trajectory of a next operating state segment TSi+1 output in step 3. Step 6: The real-time engine power calculation module 16 measures the real-time power P using the speed and torque of the engine 18 and compares it with the engine power reference value obtained in step 4. Step 7: According to a power tracking energy control strategy, the fuel cell control unit FCU 11 summarizes the SOC value and the power, which is specifically as follows: first, it is judged whether the real-time SOC value of the battery pack is greater than the maximum value SOCmax of the predicted value of the optimal SOC reference trajectory of a next operating state segment, if SOC>SOCmax, it is judged whether the real-time power P is smaller than the power reference value Pmin, if P<Pmin, the fuel cell control unit FCU 11 is turned off, and charging is stopped; if P≥ Pminist, it is further judged whether the real-time power is greater than the power reference value Pmax, if P>Pmaxist, the fuel cell control unit FCU 11 starts to work, if Pmin≤P≤Pmaxist, the state of the fuel cell control unit FCU 11 remains unchanged ; if SOC < SOCmin or SOCmin≤SOC≤SOCmax and P > Pmax, the fuel cell control unit FCU 11 allows the fuel cell stack 13 to be turned off and discharged, otherwise the state remains unchanged. Step 8: the vehicle controller VCU 2 judges whether the fuel cell bus 1 has reached the end point and completed a trip, if the trip is not completed, it returns to Step 3 and the process circulates; when the trip is over, the cloud analysis workstation 7 completes incremental learning for the new training data through the incremental learning model 8 and downloads the updated prediction model parameters again to the optimal SOC prediction model module 9 in the vehicle controller VCU 2 to obtain a update the optimal SOC prediction model.
    The workflow according to an embodiment of the present invention is explained in more detail below in connection with FIG operating state segment a controller is started; the characteristic parameters of the current operating state segment are input to the optimal SOC prediction model module 9 to obtain a predicted value of the optimal SOC reference trajectory of a next operating state segment, which is transmitted as input to the MPC prediction control module 10, the MPC prediction control module 10 providing a performance reference value of the driving motor, and wherein the battery management system module BMS 19 monitors the SOC value of the battery pack of the fuel cell vehicle in real time, and wherein the real-time power calculation module of the driving motor 16 monitors the speed and the torque in the driving process of the vehicle by the motor controller MCU 17 in order to thus to calculate the real-time performance of the drive motor; the real-time SOC value of the battery pack and the real-time power of the traction motor that were measured, and the predicted value of the optimal SOC reference trajectory of a next segment and the power reference value of the traction motor, which were calculated by the optimal SOC prediction model module 9 and the MPC predicted and outputted by the prediction control module 10 are collectively input to the fuel cell control unit FCU 11, and according to a power tracking energy control strategy, the working state of the fuel cell is judged to complete the control of a next operating state segment. In the driving process, the fuel cell bus 1 constantly uploads the vehicle driving status information of the respective segments through the vehicle driving communication system, after the completion of each trip, the cloud analysis workstation 7 performs incremental learning and training through the vehicle driving status information uploaded in real time. If a reconfiguration of the original travel route leads to a change in road working conditions, the uploaded vehicle running condition information will also change accordingly. After completing the trip, the cloud analysis workstation 7 performs incremental learning and training on the new data using the incremental learning model 8 , then the model parameters are updated, and the trained model parameters are downloaded to the fuel cell vehicle controller VCU 2 . When the vehicle runs again under the same sudden operating condition, it can adapt to the change of the new environment.
    The above series of detailed explanations relate only to the practicable embodiments of the present invention and do not limit the scope of the present invention, any equivalent embodiments or changes that do not depart from the gist of the present invention shall be construed as of are considered to be within the scope of the present invention.
    权利要求:
    Claims (8)
    [1]
    1. Intelligent control system with global real-time optimization for a fuel cell bus, characterized in that it comprises a vehicle driving communication unit, a vehicle driving condition information prediction and analysis unit, a fuel cell vehicle control unit and a vehicle driving condition information acquisition unit, wherein the fuel cell vehicle control unit comprises a vehicle controller (2) and a fuel cell control unit (11) wherein the vehicle running communication unit enables signal connection between the fuel cell vehicle control unit and the vehicle running state information prediction and analysis unit, and signal connection between the fuel cell vehicle control unit and the vehicle running state information acquisition unit; and wherein the vehicle running state information prediction and analysis unit is configured to calculate prediction model parameters and download them to the fuel cell vehicle control unit to perform an update of an optimal SOC prediction model; and wherein the fuel cell control unit is adapted to control the working state of the fuel cell, wherein the intelligent control system is adapted to obtain the predictive model parameters by transmitting and training characteristic parameters calculated by a characteristic parameter calculation module (22) to an incremental learning model (8). will; wherein the characteristic parameter calculation module (22) receives the vehicle running condition information and optimal SOC reference trajectory sent by a speed information receiving module (20) and a dynamic planning module (21).
    [2]
    2. Intelligent control system with global real-time optimization for a fuel cell bus according to claim 1, characterized in that the fuel cell vehicle control unit is designed to, based on a power reference value and a real-time power of a drive motor (18), a predicted value of an optimal state of charge reference trajectory of a next operating state segment of the Fuel cell bus and a real-time state of charge value of a battery pack to control the working state of the fuel cell.
    [3]
    3. Intelligent control system with global real-time optimization for a fuel cell bus according to claim 2, characterized in that a battery management system (19) is designed to detect a real-time state of charge of the battery pack in real time.
    [4]
    4. Intelligent control system with a global real-time optimization for a fuel cell bus according to claim 2, characterized in that the predicted value of the optimal state of charge reference trajectory of the next operating state segment is defined by the formula Y* = min+ f(X)(max- min), where a module of the optimal state of charge prediction model (9) in the fuel cell vehicle control unit is designed to receive the characteristic parameters of the current operating state segment, where f(X) is the regression function, max is the maximum value of the sample data, min is the minimum value of the sample data and Y* is the predicted value of the optimal State of charge reference trajectory of a next operating state segment is.
    [5]
    5. Intelligent control system with global real-time optimization for a fuel cell bus according to claim 4, characterized in that a prediction control module (10) in the fuel cell vehicle control unit is adapted to calculate the power reference value of the drive motor (18) in accordance with the predicted value of the optimal state of charge reference trajectory of the to calculate the next operating state segment.
    [6]
    6. Intelligent control system with global real-time optimization for a fuel cell bus according to claim 1, characterized in that the incremental learning model (8) is adapted to train the characteristic parameters to obtain a regression model and a support vector set, with new vehicle driving operating status information being integrated , which are synthesized with the support vector set to the new sample database data, and training continues to obtain an entirely new regression model and an entirely new support vector set as the final model.
    [7]
    7. Intelligent control method with global real-time optimization for a fuel cell bus, characterized in that it comprises the following steps:Step 1: Performing training for the obtained vehicle driving condition information and optimal state of charge reference trajectory using the incremental learning model (8) to obtain the optimal state of charge prediction model parameters, downloading them to the optimal state of charge prediction model module (9) to update the optimal state of charge prediction model to carry out;Step 2: inputting the characteristic parameters of the current operating state segment into the optimal state of charge prediction model module (9) to output the predicted value of the optimal state of charge reference trajectory of a next operating state segment;Step 3: inputting the predicted value of the optimal state of charge reference trajectory of a next operating state segment into the prediction control module (10) to obtain the power reference value of the drive motor;Step 4: the fuel cell control unit (11) judges the state of charge value and the output and controls the working state of the fuel cell;Step 5: the vehicle controller (2) judges whether the fuel cell bus has completed a trip, if the trip is not completed, it returns to Step 2 and the process circulates; when the trip is finished, the cloud analysis workstation (7) downloads the updated prediction model parameters again to the vehicle controller (2) in order to update the optimal state of charge prediction model.
    [8]
    8. Intelligent control method with global real-time optimization for a fuel cell bus according to claim 7, characterized in that step 4 is as follows: judging whether the real-time state of charge value SOC of the battery pack is greater than the maximum value SOCmaxdes predicted value of the optimal state of charge reference trajectory of a next Operating state segment is when SOC>SOCmax, it is judged whether the real-time power P is smaller than the power reference value Pmin when P<Pmin. , the fuel cell control unit (11) is turned off and charging is stopped; if P≥Pmin, it is further judged whether the real-time power P is greater than the power reference value Pmax, if P>Pmax, the fuel cell control unit (11) starts to work, if Pmin≤P≤Pmax, the state of the fuel cell control unit (11 ) unchanged; if SOC<SOCmin or S0Cmin≤SOC≤SOCmax and P>Pmax, the fuel cell control unit (11) enables the fuel cell stack (13) to be switched off and discharged, otherwise the state remains unchanged.
    类似技术:
    公开号 | 公开日 | 专利标题
    CH717533B1|2022-01-14|Intelligent control system and method with real-time global optimization for a fuel cell bus.
    DE102009048821A1|2011-04-14|Method for determining range of vehicle, particularly electric vehicle, involves considering parameter during determination of range, which is obtained from information relating to scheduled or actually traveled driving route
    DE102018116826A1|2019-01-17|VEHICLE WITH MODEL-BASED ROUTE ENERGY FORECASTING, CORRECTION AND OPTIMIZATION
    DE102012106268A1|2013-01-17|A method and system for determining a target charge state for charging a battery in a vehicle using external electric power
    DE102015203803A1|2015-09-10|Outsourcing parameter identification through the use of cloud computing resources
    CN106427990B|2018-09-28|Hybrid power system and its energy management method
    DE102004053608A1|2005-08-04|Vehicle energy management system using predictions
    DE102016101100A1|2016-08-04|Intelligent energy management to improve the life of an electrically charged battery
    CN105882648A|2016-08-24|Hybrid power system energy management method based on fuzzy logic algorithm
    DE102015200402A1|2015-07-23|A system and method for controlling battery performance based on predicted battery energy usage
    DE102011085454A1|2013-05-02|Method for controlling a hybrid drive with an internal combustion engine and an electric motor and an electrical energy store, in particular for a rail vehicle, control device and hybrid drive
    DE102017117207A1|2018-02-08|SYSTEM AND METHOD FOR SELECTING A CHARGE SOURCE FOR AN ELECTRIC VEHICLE
    CN104348205A|2015-02-11|SOC-SOH |-based distributed BMS |
    DE102017201226A1|2018-07-26|Method for operating a data evaluation system for motor vehicle functions and data evaluation system
    CN105098852A|2015-11-25|Electric automobile battery changing device and charge management method thereof
    WO2018215273A1|2018-11-29|Method for controlling a drive device of a hybrid vehicle and hybrid vehicle
    DE112010003356T5|2012-06-21|Identification and use of reserve energy in energy storage systems
    CN111581850A|2020-08-25|Full-period power battery management system applying digital twinning technology
    DE102014002973A1|2015-06-18|Method for charging batteries and converter for charging
    DE102013209837A1|2014-11-27|Recuperation arrangement with control unit
    DE102020120589A1|2021-02-18|ELECTRIFIED VEHICLE CONFIGURED TO SELECTIVELY INCREASE AN ENERGY RECOVERY THRESHOLD BASED ON FRICTION FORECAST, AND THE APPROPRIATE PROCEDURE
    DE102011119904A1|2013-06-06|Electrical power supply for electrical car and hybrid car, has secondary circuit inductively attached to primary circuit, and internal voltage sources or voltage sinks and external voltage sources or -sinks attached to secondary circuit
    DE102014222073A1|2016-05-04|Method and device for operating a motor vehicle with an electrical energy store
    WO2014048463A1|2014-04-03|Device having a stationary buffer battery for charging electrical energy accumulators and method
    CN109239485B|2020-10-30|Energy storage tramcar super capacitor fault identification method and system based on BP neural network
    同族专利:
    公开号 | 公开日
    CN110852482A|2020-02-28|
    CN110852482B|2020-12-18|
    WO2021073036A1|2021-04-22|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    CN107818377A|2016-09-12|2018-03-20|法乐第网络科技有限公司|Vehicle global optimization control method, system, vehicle and cloud computing platform based on cloud computing platform|
    JP6930592B2|2017-03-01|2021-09-01|日本電気株式会社|Warning, Irregular, Unfavorable Mode Predictors and Methods|
    CN107425534B|2017-08-25|2020-07-31|电子科技大学|Micro-grid scheduling method based on optimization of storage battery charging and discharging strategy|
    CN108197726B|2017-12-11|2021-11-09|浙江科技学院|Family energy data optimization method based on improved evolutionary algorithm|
    CN109927709B|2017-12-15|2020-08-28|郑州宇通客车股份有限公司|Vehicle driving route working condition determining method, energy management method and system|
    CN108427985B|2018-01-02|2020-05-19|北京理工大学|Plug-in hybrid vehicle energy management method based on deep reinforcement learning|
    CN108177648B|2018-01-02|2019-09-17|北京理工大学|A kind of energy management method of the plug-in hybrid vehicle based on intelligent predicting|
    CN110309978A|2019-07-09|2019-10-08|南京邮电大学|Neural network photovoltaic power prediction model and method based on the adjustment of secondary dynamic|
    CN110852482B|2019-10-15|2020-12-18|江苏大学|Real-time global optimization intelligent control system and method for fuel cell bus|CN110852482B|2019-10-15|2020-12-18|江苏大学|Real-time global optimization intelligent control system and method for fuel cell bus|
    CN113428049B|2021-08-26|2021-11-09|北京理工大学|Fuel cell hybrid vehicle energy management method considering battery aging inhibition|
    法律状态:
    2021-12-15| PL| Patent ceased|
    优先权:
    申请号 | 申请日 | 专利标题
    CN201910977702.8A|CN110852482B|2019-10-15|2019-10-15|Real-time global optimization intelligent control system and method for fuel cell bus|
    PCT/CN2020/079244|WO2021073036A1|2019-10-15|2020-03-13|Real-time global optimization intelligent control system and method for fuel cell bus|
    [返回顶部]