Electric pile temperature control method and device, control equipment and storage medium
Automatically adjusting the stack temperature through multi-layer perceptron neural network prediction and optimization algorithm, solving the problem of complex rule definitions and lack of intelligence in fuzzy control, and achieving rapid and accurate control of fuel cell vehicle stack temperature.
Patent Information
- Application Number
- CN202411039411.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-29
AI Technical Summary
The existing stack temperature control method in fuel cell vehicles is based on fuzzy control, the rule definition is complex and dependent on experience, and the control strategy lacks intelligence, making it difficult to respond quickly to dynamic changes, resulting in poor temperature control effect.
A neural network model based on multi-layer perceptron is adopted to predict the stack temperature by training related historical data, and the control strategy is determined in combination with the direction acceleration optimization algorithm, and the stack temperature is automatically adjusted to achieve fast and accurate temperature control.
Without manually defining fuzzy rules, neural networks can quickly predict and control stack temperature, reduce temperature control errors, improve control effects, ensure that the system quickly responds to changes in stack state, and solve the problem of lack of intelligence in fuzzy control.
Smart Images

Figure CN119050410A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of fuel cell temperature control technology, and in particular to a fuel cell stack temperature control method, device, control equipment and storage medium. Background Art
[0002] Fuel cell vehicles refer to vehicles that use fuel cells as their main power source, and have the advantages of high efficiency and environmental protection. During the operation of fuel cell vehicles, the control of the temperature of the stack is crucial. Too high a stack temperature will affect the performance and life of the fuel cell, while too low a temperature will affect its efficiency.
[0003] For the temperature control of the fuel cell stack, the temperature regulation method based on fuzzy control is usually adopted in the related technology. Fuzzy control is to adjust the temperature of the fuel cell stack by setting different cooling water outlet temperature ranges and corresponding control element control rules. However, the temperature regulation based on fuzzy control has the following disadvantages: First, the rule definition is complex and the performance depends on experience: a lot of experience is required to define fuzzy rules in fuzzy control, and the design and debugging of rules are very complicated. Second, the control strategy lacks intelligence: for systems with rapid dynamic changes, the response of fuzzy control is not fast and accurate enough, and the control strategy lacks intelligence. It is difficult to respond to changes in the working state of the fuel cell in real time, resulting in poor temperature control effect. Summary of the invention
[0004] The present application provides a method, device, control equipment and storage medium for controlling the temperature of a battery stack, which can quickly and accurately predict the temperature of the battery stack without manually defining fuzzy rules, ensure that the system can quickly respond to changes in the working state of the battery stack, and improve the temperature control effect of the battery stack.
[0005] In a first aspect, the present application provides a method for controlling a temperature of a fuel cell stack, comprising:
[0006] Acquire at least one type of fuel cell system data related to the temperature of the fuel cell stack at a current moment, and an expected temperature for the fuel cell stack;
[0007] Inputting the at least one type of fuel cell system data into a pre-trained temperature prediction model to obtain a predicted temperature of the fuel cell stack after a preset time period;
[0008] Determining, based on the expected temperature and the predicted temperature, a control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in compliance with a preset target;
[0009] According to the control strategy, adjusting the at least one electronic device to adjust the temperature of the battery stack to the desired temperature;
[0010] The temperature prediction model is a neural network model trained based on a multi-layer perceptron using various types of historical fuel cell system data related to the temperature of the fuel cell stack as input data, wherein the various types include at least one type.
[0011] Optionally, the preset target is that the deviation between the expected temperature and the predicted temperature is minimized, and the control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack when the preset target is met is determined based on the expected temperature and the predicted temperature, including:
[0012] Determining a minimum value of a first objective function based on a directional acceleration optimization algorithm, wherein the first objective function represents a deviation between the desired temperature and the predicted temperature;
[0013] When the first objective function obtains a minimum value, a control strategy corresponding to the at least one electronic device is determined.
[0014] Optionally, the preset target is a deviation between the expected temperature and the predicted temperature, and the energy consumption generated by the at least one electronic device meets corresponding preset conditions, and the control strategy corresponding to the at least one electronic device for adjusting the temperature of the battery stack in accordance with the expected temperature and the predicted temperature is determined in accordance with the expected temperature and the predicted temperature, including:
[0015] Determining a minimum value of a second objective function based on a directional acceleration optimization algorithm, wherein the second objective function represents a weighted sum of a deviation between the desired temperature and the predicted temperature and energy consumption generated by the at least one electronic device;
[0016] When the second objective function obtains a minimum value, a control strategy corresponding to the at least one electronic device is determined.
[0017] Optionally, the multilayer perceptron includes an input layer, a first filter layer, a hidden layer, a second filter layer and an output layer in the order of data processing;
[0018] The input layer is used to extract features from the input data through a fully connected layer having multiple neurons;
[0019] The first filtering layer is used to randomly discard data in neurons of a first proportion from the features extracted by the input layer;
[0020] The hidden layer is used to extract features from the remaining data in the first filtering layer through a fully connected layer having a plurality of neurons;
[0021] The second filtering layer is used to randomly discard data in neurons of a second proportion from the features extracted by the hidden layer;
[0022] The output layer is used to output the predicted temperature of the fuel cell stack through a fully connected layer.
[0023] Optionally, the temperature prediction model is trained through the following steps:
[0024] Inputting the historical fuel cell system data into the input layer, so as to extract features of the historical fuel cell system data through the input layer to obtain first feature data;
[0025] Inputting the first feature data into the first filtering layer, so as to randomly discard data in a first proportion of neurons through the first filtering layer, and obtain second feature data;
[0026] Inputting the second feature data into the hidden layer, so as to perform feature extraction on the second feature data through the hidden layer to obtain third feature data;
[0027] Inputting the third characteristic data into the second filtering layer, so as to randomly discard data in neurons of a second proportion through the second filtering layer, and obtain fourth characteristic data;
[0028] The fourth characteristic data is input into the output layer so as to process the fourth characteristic data through the output layer and output the predicted temperature of the fuel cell stack.
[0029] Optionally, when the multilayer perceptron is trained using the historical data, a back propagation algorithm is used to train the multilayer perceptron, and a cross-validation method and a grid search method are used to optimize the hyperparameters of the multilayer perceptron.
[0030] Optionally, the at least one type includes the pressure of hydrogen when entering the fuel cell stack, the pressure of air when entering the fuel cell stack, and the temperature of cooling water when entering the fuel cell stack;
[0031] The at least one electronic device includes one or more of a cooling fan, a cooling water circulation pump, and an electronic thermostat;
[0032] In addition to the at least one type mentioned above, the multiple types also include one or more of the output power of the cooling fan, the output power of the cooling water circulation pump and the opening of the electronic thermostat. The opening of the electronic thermostat is used to control the flow rate of cooling water entering each cooling circulation system related to the fuel cell stack.
[0033] Optionally, the control strategy corresponding to the at least one electronic device includes one or more information of a target output power of a cooling fan, a target output power of a cooling water circulation pump, and a target opening of the electronic thermostat; and adjusting the at least one electronic device according to the control strategy includes:
[0034] If the at least one electronic device includes the cooling fan, adjusting the cooling fan according to the target output power of the cooling fan;
[0035] If the at least one electronic device includes the cooling water circulation pump, regulating the cooling water circulation pump according to the target output power of the cooling water circulation pump;
[0036] If the at least one electronic device includes the electronic thermostat, the electronic thermostat is adjusted according to the target opening of the electronic thermostat.
[0037] In a second aspect, the present application provides a device for controlling the temperature of a fuel cell stack, comprising:
[0038] A first acquisition module, configured to acquire at least one type of fuel cell system data related to the temperature of the fuel cell stack at a current moment, and an expected temperature for the fuel cell stack;
[0039] A second acquisition module, configured to input the at least one type of fuel cell system data into a temperature prediction model to obtain a predicted temperature of the fuel cell stack after a preset time period;
[0040] A determination module, configured to determine, based on the expected temperature and the predicted temperature, a control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in accordance with a preset target;
[0041] an adjusting module, configured to adjust the at least one electronic device according to the control strategy, so as to adjust the temperature of the battery stack to the desired temperature;
[0042] The temperature prediction model is a neural network model trained based on a multi-layer perceptron using various types of historical fuel cell system data related to the temperature of the fuel cell stack as input data, wherein the various types include at least one type.
[0043] Optionally, the preset target is that the deviation between the expected temperature and the predicted temperature is minimized, and the determination module includes:
[0044] A first determination submodule, configured to determine a minimum value of a first objective function based on a directional acceleration optimization algorithm, wherein the first objective function represents a deviation between the desired temperature and the predicted temperature;
[0045] The second determination submodule is used to determine the control strategy corresponding to the at least one electronic device when the first objective function obtains a minimum value.
[0046] Optionally, the preset target is a deviation between the expected temperature and the predicted temperature, and the energy consumption generated by the at least one electronic device meets corresponding preset conditions, and the determination module includes:
[0047] a third determination submodule, configured to determine a minimum value of a second objective function based on a directional acceleration optimization algorithm, wherein the second objective function represents a weighted sum of a deviation between the expected temperature and the predicted temperature and energy consumption generated by the at least one electronic device;
[0048] The fourth determination submodule is used to determine the control strategy corresponding to the at least one electronic device when the second objective function obtains a minimum value.
[0049] Optionally, the multilayer perceptron includes an input layer, a first filter layer, a hidden layer, a second filter layer and an output layer in the order of data processing;
[0050] The input layer is used to extract features from the input data through a fully connected layer having multiple neurons;
[0051] The first filtering layer is used to randomly discard data in neurons of a first proportion from the features extracted by the input layer;
[0052] The hidden layer is used to extract features from the remaining data in the first filtering layer through a fully connected layer having a plurality of neurons;
[0053] The second filtering layer is used to randomly discard data in neurons of a second proportion from the features extracted by the hidden layer;
[0054] The output layer is used to output the predicted temperature of the fuel cell stack through a fully connected layer.
[0055] Optionally, when the multilayer perceptron is trained using the historical data, a back propagation algorithm is used to train the multilayer perceptron, and a cross-validation method and a grid search method are used to optimize the hyperparameters of the multilayer perceptron.
[0056] Optionally, the at least one type includes the pressure of hydrogen when entering the fuel cell stack, the pressure of air when entering the fuel cell stack, and the temperature of cooling water when entering the fuel cell stack;
[0057] The at least one electronic device includes one or more of a cooling fan, a cooling water circulation pump, and an electronic thermostat;
[0058] In addition to the at least one type mentioned above, the multiple types also include one or more of the output power of the cooling fan, the output power of the cooling water circulation pump and the opening of the electronic thermostat. The opening of the electronic thermostat is used to control the flow rate of cooling water entering each cooling circulation system related to the fuel cell stack.
[0059] Optionally, the control strategy corresponding to the at least one electronic device includes one or more information of a target output power of a cooling fan, a target output power of a cooling water circulation pump, and a target opening of the electronic thermostat; and the adjustment module includes:
[0060] A first regulating submodule, configured to regulate the cooling fan according to a target output power of the cooling fan if the at least one electronic device includes the cooling fan;
[0061] a second regulating submodule, configured to regulate the cooling water circulating pump according to a target output power of the cooling water circulating pump if the at least one electronic device includes the cooling water circulating pump;
[0062] The third regulating submodule is configured to regulate the electronic thermostat according to a target opening of the electronic thermostat if the at least one electronic device includes the electronic thermostat.
[0063] In a third aspect, the present application provides a control device, including:
[0064] processor;
[0065] a memory for storing instructions executable by the processor;
[0066] Wherein, the processor is configured to execute to implement a stack temperature control method as described in the first aspect of the present application.
[0067] In a fourth aspect, the present application provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by a processor of a control device, the control device is enabled to execute a method for controlling a battery stack temperature as described in the first aspect of the present application.
[0068] The present application provides a method for controlling the temperature of a battery stack, firstly predicting the temperature of the battery stack after a preset time by a pre-trained temperature prediction model, then determining the control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in accordance with the preset target based on the predicted temperature and the expected temperature for the battery stack, and finally adjusting at least one electronic device according to the control strategy, thereby adjusting the temperature of the battery stack to the expected temperature. Since the temperature prediction model is a neural network model trained based on a multi-layer perceptron with various types of historical fuel cell system data related to the temperature of the battery stack as input data, it can quickly and accurately predict the temperature of the battery stack after a preset time, reduce the temperature control error of the battery stack, and improve the temperature control effect of the battery stack. Compared with the fuzzy control method in the related art, it has at least the following technical effects:
[0069] First, the neural network can automatically learn the complex relationship between input and output through training data, and realize the rapid prediction of the temperature of the battery stack without manually defining fuzzy rules, which solves the problem of complex rule definition and performance dependence on experience when using fuzzy control;
[0070] Second, the neural network can be combined with the optimization algorithm to quickly predict the output based on the input, realize real-time prediction and control of the temperature of the fuel cell stack, and dynamically adjust the input according to factors such as the expected temperature of the fuel cell stack to ensure that the system can quickly respond to changes in the working state of the fuel cell stack, thus solving the problem of lack of intelligence in the control strategy when using fuzzy control. BRIEF DESCRIPTION OF THE DRAWINGS
[0071] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0072] Figure 1 is a flow chart of a method for controlling temperature of a battery stack shown in one embodiment of the present application;
[0073] Figure 2 is a schematic diagram of a network structure of a multilayer perceptron shown in one embodiment of the present application;
[0074] Figure 3 It is a structural block diagram of a device for controlling the temperature of a battery stack shown in one embodiment of the present application;
[0075] Figure 4 It is a structural schematic diagram of a control device provided in one embodiment of the present application. DETAILED DESCRIPTION
[0076] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0077] The stack temperature control method of the present application is mainly used to control the temperature of the stack in the vehicle. The execution device of the method of the present application is a control system in the vehicle, which can be the central control system of the vehicle or other types of control systems. The control system can be located in the vehicle or in the cloud. The present application does not specifically limit the type and location of the control system.
[0078] For the temperature control of the fuel cell stack, the temperature regulation based on fuzzy control in the related art has at least the following disadvantages: First, the rule definition is complex and the performance depends on experience: a lot of experience is required to define fuzzy rules in fuzzy control, and the design and debugging of rules are very complicated. For complex systems, the number of rules will increase rapidly, making the system difficult to maintain and adjust. Secondly, the performance of fuzzy control is highly dependent on the experience and knowledge of the designer. If the rules are unreasonable, the system performance will be unsatisfactory. Second, the control strategy lacks intelligence: for systems with rapid and dynamic changes, the response of fuzzy control is not fast and accurate enough, the control strategy lacks intelligence, and it cannot be flexibly adjusted according to different working conditions. It is difficult to respond to changes in the working state of the fuel cell in real time, resulting in poor temperature control.
[0079] The stack temperature control method provided in this application can better overcome the problems in the above-mentioned related technologies. Figure 1 1 is a flow chart of a method for controlling the temperature of a battery stack according to an embodiment of the present application. Figure 1 The stack temperature control method in this application includes the following steps:
[0080] Step S11: Acquire at least one type of fuel cell system data related to the temperature of the fuel cell stack at the current moment, and an expected temperature for the fuel cell stack.
[0081] In the embodiment of the present application, the temperature of the fuel cell stack cannot be directly measured. Since the cooling water can fully transfer heat with the fuel cell stack, the outlet temperature of the cooling water, that is, the temperature of the cooling water flowing out of the fuel cell stack, is used as the temperature of the fuel cell stack.
[0082] Among them, at least one type of fuel cell system data related to the temperature of the fuel cell stack may include one or more of the pressure data when hydrogen enters the fuel cell stack, the pressure data when air enters the fuel cell stack, and the temperature when cooling water enters the fuel cell stack, which can be set according to actual needs.
[0083] Step S12: Input at least one type of fuel cell system data into the temperature prediction model to obtain the predicted temperature of the fuel cell stack after a preset period of time. The temperature prediction model is a neural network model trained based on a multi-layer perceptron using multiple types of historical fuel cell system data related to the temperature of the fuel cell stack as input data, and the multiple types include at least one type.
[0084] In this embodiment, the temperature prediction model is based on a multi-layer perceptron. Various types of historical fuel cell system data related to the temperature of the fuel cell stack are used as sample data. The multi-layer perceptron is trained using the sample data to obtain a neural network model, which is used to predict the temperature of the fuel cell stack after a preset time period. The preset time period can be set according to actual needs.
[0085] The following is a detailed explanation of how to determine various types of historical fuel cell system data related to the temperature of the fuel cell stack, which mainly includes the following steps:
[0086] Step 1: Data collection. Collect fuel cell system data during the vehicle road test, including but not limited to: the temperature of the cooling water entering the stack, the temperature of the cooling water flowing out of the stack, the output power of the cooling fan, the output power of the cooling water circulation pump, the pressure of the hydrogen entering the stack, the pressure of the air entering the stack, the vehicle power request, the opening of the electronic thermostat, etc.
[0087] Step 2: Data cleaning. Preprocess the data collected in step 1, which mainly includes the following steps:
[0088] 1) Missing value processing: Due to the large amount of data and storage limitations, the data will be split into multiple files. There may be missing data at the connection between the files, so the missing values need to be filled. After analysis, due to the large amount of data and high collection frequency, the forward data and backward data of the missing data are basically the same, so this application uniformly uses the forward filling method to fill the missing data.
[0089] 2) Feature conversion: CAN bus is a standard protocol for internal communication in automobiles. CAN bus data is communicated in the form of raw frames, which contain information from various sensors and systems. However, directly using these raw CAN frames for analysis and processing is extremely inefficient. Therefore, it is necessary to convert these raw data into standardized normative data for further processing and analysis. This application uses a communication matrix (DBC file) to interpret the raw CAN frames and convert the encoded data in the raw CAN frames into actual physical quantities.
[0090] 3) Data normalization: Due to the large number of feature dimensions, different dimensions, and large numerical differences, data stability is affected. This application uses the Min-Max normalization method to normalize the data to eliminate the differences between different feature dimensions. The conversion formula is as follows:
[0091]
[0092] Among them, x is the sample data to be processed, x max is the maximum value of the sample data, x min is the minimum value of the sample data.
[0093] Step 3: Feature screening. This application uses the Pearson correlation coefficient statistical method to analyze the correlation between each feature and the temperature of the battery stack. Based on the analysis results and combined with the vehicle thermal management system theory, the features that are strongly correlated with the temperature of the battery stack are screened out, for example, including: the temperature when the cooling water enters the battery stack, the output power of the cooling fan, the output power of the cooling water circulation pump, the pressure when the hydrogen enters the battery stack, the pressure when the air enters the battery stack, and the opening of the electronic thermostat, a total of 6 features.
[0094] In the present application, after obtaining multiple features that are strongly correlated with the temperature of the battery stack, sample data for training the multilayer perceptron can be further selected based on some or all of the multiple features (specific features are selected according to actual needs). For example, historical data corresponding to five strongly correlated features, namely, the temperature when the cooling water enters the battery stack, the output power of the cooling fan, the output power of the cooling water circulation pump, the pressure when the hydrogen enters the battery stack, and the pressure when the air enters the battery stack, can be screened out from the historical data of the vehicle as sample data, or historical data corresponding to six strongly correlated features, namely, the temperature when the cooling water enters the battery stack, the output power of the cooling fan, the output power of the cooling water circulation pump, the pressure when the hydrogen enters the battery stack, the pressure when the air enters the battery stack, and the opening of the electronic thermostat, can be screened out from the historical data of the vehicle as sample data.
[0095] It should be noted that since the six strongly correlated features screened out in step 3 are based on the type of data collected in step 1, in actual implementation, multiple types of data to be collected can be selected in step 1 according to actual needs. Therefore, the multiple strongly correlated features screened out in step 3 may be more, not just the six given above.
[0096] In addition, in step 1, some vehicles may not be able to obtain some of the above 6 strongly correlated features. In this case, other features related to these unobtainable features can be used as substitutes. For example, when the pressure of hydrogen (or air) entering the fuel cell stack cannot be obtained, the supply of hydrogen (or air) can be used as a substitute. For another example, when the output power of the cooling fan cannot be obtained, the load of the cooling fan can be used as a substitute. When the output power of the cooling water circulation pump cannot be obtained, the output pressure of the cooling water circulation pump can be used as a substitute.
[0097] Step S13: According to the expected temperature and the predicted temperature, a control strategy corresponding to at least one electronic device for adjusting the temperature of the fuel cell stack is determined in accordance with a preset target.
[0098] After obtaining multiple features (assuming N features) that are strongly correlated with the temperature of the fuel cell stack in step 3 of step S12, the present application selects sample data for training the multilayer perceptron based on some or all of the N features (assuming M features). After the training is completed, based on the temperature prediction model, if the values of P of the features are given, the model can output an expression (the predicted temperature of the fuel cell stack) containing the variables corresponding to the remaining (MP) features. At this time, the historical data corresponding to the M features are the multiple types of historical fuel cell system data in step S12, the data corresponding to the P features are at least one type of fuel cell system data in step S11, and the electronic devices involved in the remaining (MP) features are at least one electronic device in step S13.
[0099] For example, in one embodiment, the temperature prediction model is trained based on the historical data corresponding to the six strongly correlated features selected in step 3 above. If the temperature of the cooling water entering the stack, the pressure of the hydrogen entering the stack, and the pressure of the air entering the stack are given, then the temperature prediction model can output the predicted stack temperature. Since the output power of the cooling fan, the output power of the cooling water circulation pump, and the opening of the electronic thermostat are not given, the predicted stack temperature is an expression containing variables related to the output power of the cooling fan, the output power of the cooling water circulation pump, and the opening of the electronic thermostat. When a situation that meets the preset target is found according to the expected temperature and the predicted temperature, the values of the variables in the expression can be further determined, that is, the specific value of the output power of the cooling fan (the target output power of the cooling fan), the specific value of the output power of the cooling water circulation pump (the target output power of the cooling water circulation pump), and the specific value of the opening of the electronic thermostat (the target opening of the electronic thermostat) can be determined. These specific values can be used to guide the parameter adjustment of the corresponding electronic equipment. For example, the target output power of the cooling fan can be used to guide the power adjustment of the cooling fan, the target output power of the cooling water circulation pump can be used to guide the power adjustment of the cooling water circulation pump, and the target opening of the electronic thermostat can be used to guide the adjustment of the opening of the electronic thermostat. Therefore, in the present application, the control strategy of the corresponding electronic equipment is generated according to the specific values of these determined variables. Naturally, in the present embodiment, at least one electronic device is a cooling fan, a cooling water circulation pump, and an electronic thermostat.
[0100] For example, in another embodiment, the temperature prediction model is obtained by training the historical data corresponding to five strongly correlated features, namely, the temperature when the cooling water enters the battery stack, the output power of the cooling fan, the output power of the cooling water circulation pump, the pressure when the hydrogen enters the battery stack, and the pressure when the air enters the battery stack. If the temperature when the cooling water enters the battery stack, the pressure when the hydrogen enters the battery stack, and the pressure when the air enters the battery stack are given, then the temperature output by the temperature prediction model is an expression containing variables related to the output power of the cooling fan and the output power of the cooling water circulation pump. When a situation that meets the preset target is found based on the expected temperature and the predicted temperature, the target output power of the output power of the cooling fan and the target output power of the cooling water circulation pump can be further determined, and a control strategy for guiding the parameter adjustment of the cooling fan and the cooling water circulation pump can be obtained, that is, the power of the cooling fan is adjusted according to the target output power of the cooling fan, and the power of the cooling water circulation pump is adjusted according to the target output power of the cooling water circulation pump. At this time, at least one electronic device is a cooling fan and a cooling water circulation pump.
[0101] In the present application, assuming that the multiple features strongly correlated with the temperature of the fuel cell stack obtained in step 3 of step S12 are the temperature of the cooling water entering the fuel cell stack, the output power of the cooling fan, the output power of the cooling water circulation pump, the pressure of the hydrogen entering the fuel cell stack, the pressure of the air entering the fuel cell stack, and the opening of the electronic thermostat, then the following situation will exist:
[0102] At least one type of fuel cell system data in step S12 includes the pressure when hydrogen enters the stack, the pressure when air enters the stack, and the temperature when cooling water enters the stack. If at least one electronic device used to adjust the temperature of the stack includes one or more of a cooling fan, a cooling water circulation pump, and an electronic thermostat, then correspondingly, the multiple types of historical fuel cell system data in step S12, in addition to at least one type of fuel cell system data, also include one or more of the output power of the cooling fan, the output power of the cooling water circulation pump, and the opening of the electronic thermostat. The opening of the electronic thermostat is used to control the flow of cooling water entering each cooling circulation system related to the stack. In other words, if at least one electronic device includes a cooling fan, then the multiple types of historical fuel cell system data need to include the output power of the cooling fan, if at least one electronic device includes a cooling water circulation pump, then the multiple types of historical fuel cell system data need to include the output power of the cooling water circulation pump, and if at least one electronic device includes an electronic thermostat, then the multiple types of historical fuel cell system data need to include the opening of the electronic thermostat.
[0103] In order to better describe the method of the present application, unless otherwise specified, in the subsequent embodiments, the multiple types refer to the 6 strongly correlated features obtained in step 3 in step S12, at least one type refers to the pressure of hydrogen entering the fuel cell stack, the pressure of air entering the fuel cell stack, and the temperature of cooling water entering the fuel cell stack, and at least one electronic device refers to a cooling fan, a cooling water circulation pump, and an electronic thermostat.
[0104] Step S14: According to the control strategy, at least one electronic device is adjusted to adjust the temperature of the battery stack to a desired temperature.
[0105] In this embodiment, if the control strategy corresponding to at least one electronic device includes one or more information of the target output power of the cooling fan, the target output power of the cooling water circulation pump, and the target opening of the electronic thermostat, then step S14 may include:
[0106] If at least one electronic device includes a cooling fan, adjusting the cooling fan according to a target output power of the cooling fan;
[0107] If at least one electronic device includes a cooling water circulation pump, the cooling water circulation pump is adjusted according to a target output power of the cooling water circulation pump;
[0108] If the at least one electronic device includes an electronic thermostat, the electronic thermostat is adjusted according to a target opening of the electronic thermostat.
[0109] In this embodiment, the temperature of the battery stack after a preset time is first predicted in real time by a pre-trained temperature prediction model, and then the control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in accordance with the preset target is determined based on the predicted temperature and the expected temperature of the battery stack, and finally at least one electronic device is adjusted according to the control strategy to adjust the temperature of the battery stack to the expected temperature. Since the temperature prediction model is a neural network model trained based on a multi-layer perceptron with various types of historical fuel cell system data related to the temperature of the battery stack as input data, it can quickly and accurately predict the temperature of the battery stack after a preset time, reduce the temperature control error of the battery stack, and improve the temperature control effect of the battery stack. Compared with the fuzzy control method in the related art, it has at least the following technical effects:
[0110] First, the neural network can automatically learn the complex relationship between input and output through training data, and realize the rapid prediction of the temperature of the battery stack without manually defining fuzzy rules, which solves the problem of complex rule definition and performance dependence on experience when using fuzzy control;
[0111] Second, the neural network can be combined with the optimization algorithm to quickly predict the output based on the input, realize real-time prediction and control of the temperature of the fuel cell stack, and dynamically adjust the input according to factors such as the expected temperature of the fuel cell stack to ensure that the system can quickly respond to changes in the working state of the fuel cell stack, thus solving the problem of lack of intelligence in the control strategy when using fuzzy control.
[0112] In combination with the above embodiments, in one implementation, the multilayer perceptron includes an input layer, a first filter layer, a hidden layer, a second filter layer, and an output layer in the order of data processing;
[0113] The input layer is used to extract features from the input data through a fully connected layer with multiple neurons;
[0114] The first filtering layer is used to randomly discard data in the first proportion of neurons in the features extracted by the input layer;
[0115] The hidden layer is used to extract features from the remaining data in the first filtering layer through a fully connected layer with multiple neurons;
[0116] The second filtering layer is used to randomly discard data in the second proportion of neurons in the features extracted by the hidden layer;
[0117] The output layer is used to output the predicted temperature of the battery stack through the fully connected layer.
[0118] On this basis, the temperature prediction model can be trained through the following steps:
[0119] Inputting the historical fuel cell system data into the input layer, so as to extract features of the historical fuel cell system data through the input layer to obtain first feature data;
[0120] Inputting the first feature data into the first filtering layer to randomly discard data in a first proportion of neurons through the first filtering layer to obtain second feature data;
[0121] Inputting the second feature data into the hidden layer, so as to perform feature extraction on the second feature data through the hidden layer to obtain third feature data;
[0122] Inputting the third characteristic data into the second filtering layer to randomly discard data in neurons of a second proportion through the second filtering layer to obtain fourth characteristic data;
[0123] The fourth characteristic data is input into the output layer so as to process the fourth characteristic data through the output layer and output the predicted temperature of the fuel cell stack.
[0124] Among them, Multi-Layer Perceptron (MLP) is a feedforward artificial neural network for supervised learning, which consists of an input layer, one or more hidden layers and an output layer. Each layer contains several neurons, which are connected by weights. Data is passed from the input layer to the hidden layer, processed by the activation function, and then passed to the output layer to finally generate a prediction result. The activation function can use ReLU, Sigmoid, etc. to introduce nonlinearity, so that the multilayer perceptron can fit complex functions. Through the back-propagation algorithm, the multilayer perceptron continuously adjusts the weights to minimize the loss function. Commonly used loss functions are mean square error and cross entropy loss.
[0125] In this application, a multi-layer perceptron is used for time series prediction. The specific structure of the multi-layer perceptron is as follows: Figure 2 shown. Figure 2 It is a schematic diagram of a network structure of a multilayer perceptron shown in one embodiment of the present application.
[0126] exist Figure 2 In the example, the input layer accepts the input sample data and extracts the high-dimensional features in the input sample data through a fully connected layer (using the ReLU activation function) with 128 neurons (the number of neurons can be adjusted according to actual needs).
[0127] The first filter layer is used to randomly discard a first proportion (for example, 20%, which can also be set to other values according to actual needs) of neurons after receiving the data transmitted by the input layer to prevent overfitting.
[0128] The hidden layer is used to receive the data transmitted by the first filtering layer, and then further process and extract features through a fully connected layer (using the ReLU activation function) with 32 neurons (the number of neurons can be adjusted according to actual needs).
[0129] The second filter layer is used to randomly discard neurons of a second proportion (for example, 20%, which can also be set to other values according to actual needs) after receiving the data transmitted by the hidden layer, so as to further prevent overfitting.
[0130] The output layer is used to generate the final prediction result, that is, the predicted temperature of the battery stack, through a fully connected layer after receiving the data transmitted by the second filter layer.
[0131] After building the multi-layer perceptron network structure, the sample data is divided into a training set and a validation set according to a ratio of 8:2 (which can be set according to actual needs), and the temperature prediction model is obtained according to the following steps:
[0132] Step 1: Based on the training set data, the back propagation algorithm is used to train the multilayer perceptron model, and the model parameters are adjusted to minimize the prediction error; and the hyperparameters of the multilayer perceptron are optimized through cross-validation and grid search methods to obtain the best model. During the training process, in order to prevent overfitting, save training resources, and improve the generalization ability of the model, an early stopping mechanism is introduced to automatically decide when to stop training, thereby simplifying the process of hyperparameter adjustment.
[0133] Specifically, this application uses the keras tool to implement the neural network structure, and the hyperparameters of the training process include:
[0134] 1) batch_size is the number of samples for each gradient update;
[0135] 2) epochs: the number of iterations of training model;
[0136] 3) validation_split: the proportion of test data;
[0137] 4) hidden_size: the number of neurons in the hidden layer;
[0138] 5) n_features is the number of input features of the model;
[0139] In one embodiment, the settings of the hyperparameters are as follows: n_features = 5, hidden_size1 = 128, hidden_size2 = 32, validation_split = 0.2, epochs = 50, batch_size = 256. After multiple iterations and adjustments, a neural network model with excellent effects is finally obtained.
[0140] Among them, the specific value of each hyperparameter can be set according to the actual situation. For example, n_features is set according to the number of input features of the model actually used, hidden_size1 is set according to the actual number of neurons in the input layer, and hidden_size2 is set according to the actual number of neurons in the hidden layer.
[0141] The second step is to evaluate the effect of the neural network model through the test set. This application uses the mean square error (mse) to verify the results. The mean square error calculation formula is:
[0142]
[0143] Among them, is the actual value of the i-th sample data, y i is the predicted value of the i-th sample data, and n is the total number of sample data.
[0144] After verifying that the mean square error of the neural network model meets the preset requirements, the neural network model is determined as the final temperature prediction model.
[0145] The third step is to save the tuned temperature prediction model as an h5 format file for subsequent use.
[0146] In combination with the above embodiments, in one implementation, the preset target is to minimize the deviation between the expected temperature and the predicted temperature. Accordingly, in step S13, according to the expected temperature and the predicted temperature, a control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack is determined in accordance with the preset target, which may include:
[0147] Based on the directional acceleration optimization algorithm, determining a minimum value of a first objective function, the first objective function representing a deviation between a desired temperature and a predicted temperature;
[0148] When the first objective function obtains a minimum value, a control strategy corresponding to at least one electronic device is determined.
[0149] In this embodiment, the directional acceleration optimization algorithm is the Powell optimization algorithm, which is a derivative-free optimization algorithm suitable for situations where the objective function is not differentiable or the gradient is difficult to calculate. The algorithm starts from an initial point and performs a one-dimensional line search along a set of initial directions (usually a standard orthogonal basis of unit vectors) to find the point that minimizes the objective function value. After completing a round of line search, the direction set is updated by the conjugate direction method, and then iterates until it converges to a local minimum. The Powell algorithm does not rely on the derivative information of the objective function and has advantages in processing non-smooth or discontinuous objective functions.
[0150] The expression of the first objective function is:
[0151] Loss=(T limit -T pred ) 2
[0152] Among them, T limit is the desired temperature, T pred To predict temperature.
[0153] Based on the above first objective function, the minimum value of the first objective function can be determined by using the Powell optimization algorithm.
[0154] For example, when multiple types of historical fuel cell system data used in training include the pressure of hydrogen entering the fuel cell stack, the pressure of air entering the fuel cell stack, the temperature of cooling water entering the fuel cell stack, the output power of the cooling fan, the output power of the cooling water circulation pump, and the opening of the electronic thermostat, and given the pressure of hydrogen entering the fuel cell stack, the pressure of air entering the fuel cell stack, and the temperature of cooling water entering the fuel cell stack, when the first objective function reaches a minimum value, the target output power of the cooling fan, the target output power of the cooling water circulation pump, and the target opening of the electronic thermostat can be determined, thereby forming a control strategy for the cooling fan, the cooling water circulation pump, and the electronic thermostat.
[0155] In combination with the above embodiments, in one implementation, the preset target is the deviation between the expected temperature and the predicted temperature, and the energy consumption generated by at least one electronic device meets the corresponding preset conditions. On this basis, in step S13, according to the expected temperature and the predicted temperature, the control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in accordance with the preset target is determined, which may include:
[0156] Determining a minimum value of a second objective function based on a directional acceleration optimization algorithm, the second objective function representing a weighted sum of a deviation between a desired temperature and a predicted temperature and energy consumption generated by at least one electronic device;
[0157] When the second objective function obtains a minimum value, a control strategy corresponding to at least one electronic device is determined.
[0158] In the process of obtaining the control strategy, the output power of the cooling fan and the output power of the cooling water circulation pump can also be included in the objective function at the same time, and the corresponding weights are assigned respectively, and the temperature control and comprehensive energy consumption (the energy consumption in this application does not consider the electronic thermostat) are comprehensively considered. Under the condition of ensuring that the temperature error range of the battery stack is acceptable, the comprehensive energy consumption is reduced as much as possible, that is, the overall effect of temperature control and comprehensive energy consumption performance is the best. The best overall effect means achieving the preset target, which also means that the deviation between the expected temperature and the predicted temperature, and the energy consumption generated by at least one electronic device meet the corresponding preset conditions.
[0159] Among them, the comprehensive optimization objective function is the second objective function, and its specific expression is:
[0160] Loss total =(T limit -T pred ) 2 +α(βP fan +P pump )
[0161] Among them, P fan is the output power of the cooling fan, P pump is the output power of the cooling water circulation pump, and α and β are weight coefficients. Through repeated experiments, the values of weight coefficients α and β can be adjusted to achieve a balance between temperature control and energy consumption optimization.
[0162] Based on the above second objective function, the minimum value of the second objective function can be determined by using the Powell optimization algorithm. At this time, the temperature control error is small, the comprehensive energy consumption of the cooling fan and the cooling water circulation pump is small, and the comprehensive effect of the two is the best.
[0163] For example, when multiple types of historical fuel cell system data used in training include the pressure of hydrogen entering the fuel cell stack, the pressure of air entering the fuel cell stack, the temperature of cooling water entering the fuel cell stack, the output power of the cooling fan, the output power of the cooling water circulation pump, and the opening of the electronic thermostat, and given the pressure of hydrogen entering the fuel cell stack, the pressure of air entering the fuel cell stack, and the temperature of cooling water entering the fuel cell stack, if the control strategy is obtained when the minimum value is obtained according to the second objective function, then the temperature control error is small, and the comprehensive energy consumption of the cooling fan and the cooling water circulation pump is also small.
[0164] The present application proposes a method for intelligent control of the temperature of a fuel cell vehicle stack. The method predicts the temperature through a multi-step and multi-input multi-layer perceptron and generates an optimized control strategy using the Powell optimization algorithm. The method can realize intelligent control of the temperature of the stack and energy consumption optimization, and has significant improvements in prediction accuracy, response speed, and energy consumption control. The method has the advantages of more intelligent control strategy, faster response to environmental changes, significantly reduced comprehensive energy consumption, and significantly extended vehicle range.
[0165] Specifically, compared with the prior art, the solution of the present application has at least the following advantages:
[0166] (1) In this application, the neural network automatically learns the complex relationship between input and output through training data, without the need to manually define rules. Through large-scale data training, the neural network can capture the dynamic characteristics of the system and optimize performance. Secondly, the neural network has good generalization ability, can handle different input modes, adapt to various working conditions, and avoid the cumbersomeness and subjectivity of manual rule design.
[0167] (2) Regarding the problem that the control strategy lacks intelligence and is poor at handling dynamic changes when using fuzzy control, in this application, the neural network model combined with the optimization algorithm can predict the output based on the input in real time, and dynamically adjust the input based on the expected output combined with factors such as the vehicle power request, to ensure that the system can quickly respond to changes in operating conditions.
[0168] (3) Regarding the problem of high energy consumption of the cooling system, in this application, during the optimization process, the output power of the cooling fan and the output power of the cooling water circulation pump are simultaneously included in the objective function and assigned corresponding weights respectively. Temperature control and energy consumption optimization are comprehensively considered to reduce the overall energy consumption and extend the vehicle's cruising range while ensuring that the temperature error range of the battery stack is acceptable.
[0169] It should be noted that, for the method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.
[0170] Based on the same inventive concept, an embodiment of the present application provides a stack temperature control device 300. Figure 3 , Figure 3 1 is a structural block diagram of a device for controlling the temperature of a battery stack according to an embodiment of the present application. Figure 3As shown, the device 300 includes:
[0171] A first acquisition module 301 is used to acquire at least one type of fuel cell system data related to the temperature of the fuel cell stack at a current moment, and an expected temperature for the fuel cell stack;
[0172] A second acquisition module 302 is used to input the at least one type of fuel cell system data into a temperature prediction model to obtain a predicted temperature of the stack after a preset time, wherein the temperature prediction model is a neural network model trained based on a multi-layer perceptron using multiple types of historical fuel cell system data related to the temperature of the stack as input data, wherein the multiple types include the at least one type;
[0173] A determination module 303 is used to determine, based on the expected temperature and the predicted temperature, a control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in accordance with a preset target;
[0174] The regulating module 304 is configured to regulate the at least one electronic device according to the control strategy, so as to adjust the temperature of the battery stack to the desired temperature.
[0175] Optionally, the preset target is that the deviation between the expected temperature and the predicted temperature is minimized, and the determination module 303 includes:
[0176] A first determination submodule, configured to determine a minimum value of a first objective function based on a directional acceleration optimization algorithm, wherein the first objective function represents a deviation between the desired temperature and the predicted temperature;
[0177] The second determination submodule is used to determine the control strategy corresponding to the at least one electronic device when the first objective function obtains a minimum value.
[0178] Optionally, the preset target is a deviation between the expected temperature and the predicted temperature, and the energy consumption generated by the at least one electronic device meets corresponding preset conditions, and the determination module 303 includes:
[0179] a third determination submodule, configured to determine a minimum value of a second objective function based on a directional acceleration optimization algorithm, wherein the second objective function represents a weighted sum of a deviation between the expected temperature and the predicted temperature and energy consumption generated by the at least one electronic device;
[0180] The fourth determination submodule is used to determine the control strategy corresponding to the at least one electronic device when the second objective function obtains a minimum value.
[0181] Optionally, the multilayer perceptron includes an input layer, a first filter layer, a hidden layer, a second filter layer and an output layer in the order of data processing;
[0182] The input layer is used to extract features from the input data through a fully connected layer having multiple neurons;
[0183] The first filtering layer is used to randomly discard data in neurons of a first proportion from the features extracted by the input layer;
[0184] The hidden layer is used to extract features from the remaining data in the first filtering layer through a fully connected layer having a plurality of neurons;
[0185] The second filtering layer is used to randomly discard data in neurons of a second proportion from the features extracted by the hidden layer;
[0186] The output layer is used to output the predicted temperature of the fuel cell stack through a fully connected layer.
[0187] Optionally, the temperature prediction model is trained through the following steps:
[0188] Inputting the historical fuel cell system data into the input layer, so as to extract features of the historical fuel cell system data through the input layer to obtain first feature data;
[0189] Inputting the first feature data into the first filtering layer, so as to randomly discard data in a first proportion of neurons through the first filtering layer, and obtain second feature data;
[0190] Inputting the second feature data into the hidden layer, so as to perform feature extraction on the second feature data through the hidden layer to obtain third feature data;
[0191] Inputting the third characteristic data into the second filtering layer, so as to randomly discard data in neurons of a second proportion through the second filtering layer, and obtain fourth characteristic data;
[0192] The fourth characteristic data is input into the output layer so as to process the fourth characteristic data through the output layer and output the predicted temperature of the fuel cell stack.
[0193] Optionally, when the multilayer perceptron is trained using the historical data, a back propagation algorithm is used to train the multilayer perceptron, and a cross-validation method and a grid search method are used to optimize the hyperparameters of the multilayer perceptron.
[0194] Optionally, the at least one type includes the pressure of hydrogen when entering the fuel cell stack, the pressure of air when entering the fuel cell stack, and the temperature of cooling water when entering the fuel cell stack;
[0195] The at least one electronic device includes one or more of a cooling fan, a cooling water circulation pump, and an electronic thermostat;
[0196] In addition to the at least one type mentioned above, the multiple types also include one or more of the output power of the cooling fan, the output power of the cooling water circulation pump and the opening of the electronic thermostat. The opening of the electronic thermostat is used to control the flow rate of cooling water entering each cooling circulation system related to the fuel cell stack.
[0197] Optionally, the control strategy corresponding to the at least one electronic device includes one or more information of a target output power of a cooling fan, a target output power of a cooling water circulation pump, and a target opening of the electronic thermostat; the adjustment module 304 includes:
[0198] A first regulating submodule, configured to regulate the cooling fan according to a target output power of the cooling fan if the at least one electronic device includes the cooling fan;
[0199] a second regulating submodule, configured to regulate the cooling water circulating pump according to a target output power of the cooling water circulating pump if the at least one electronic device includes the cooling water circulating pump;
[0200] The third regulating submodule is configured to regulate the electronic thermostat according to a target opening of the electronic thermostat if the at least one electronic device includes the electronic thermostat.
[0201] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0202] The present application also provides a control device, Figure 4 is a schematic diagram of a control device provided by an embodiment of the present application. Figure 4 , the control device 400 comprises:
[0203] Processor 401;
[0204] A memory 402 for storing instructions executable by the processor;
[0205] The processor 401 is configured to execute to implement a stack temperature control method described in an embodiment of the present application.
[0206] In this application, the control device may be a central control system in a vehicle or other types of control devices. The control device may be located in the vehicle or in the cloud. This application does not specifically limit the type and location of the control device.
[0207] An embodiment of the present application also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by a processor of a control device, the control device is enabled to execute a method for controlling a battery stack temperature as described in an embodiment of the present application.
[0208] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0209] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0210] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0211] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0212] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.
[0213] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.
Claims
1. A method for controlling the temperature of a fuel cell stack, characterized in that: include: Acquire at least one type of fuel cell system data related to the temperature of the fuel cell stack at a current moment, and an expected temperature for the fuel cell stack; Inputting the at least one type of fuel cell system data into a pre-trained temperature prediction model to obtain a predicted temperature of the fuel cell stack after a preset time period, wherein the temperature prediction model is a neural network model trained based on a multi-layer perceptron using multiple types of historical fuel cell system data related to the temperature of the fuel cell stack as input data, wherein the multiple types include the at least one type; Determining, based on the expected temperature and the predicted temperature, a control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in compliance with a preset target; According to the control strategy, the at least one electronic device is adjusted to adjust the temperature of the battery stack to the desired temperature.
2. The method according to claim 1, characterized in that The preset target is that the deviation between the expected temperature and the predicted temperature is minimized, and the control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in accordance with the expected temperature and the predicted temperature is determined in accordance with the preset target, including: Determining a minimum value of a first objective function based on a directional acceleration optimization algorithm, wherein the first objective function represents a deviation between the desired temperature and the predicted temperature; When the first objective function obtains a minimum value, a control strategy corresponding to the at least one electronic device is determined.
3. The method according to claim 1, characterized in that: The preset target is the deviation between the expected temperature and the predicted temperature, and the energy consumption generated by the at least one electronic device reaches the corresponding preset conditions. The control strategy corresponding to the at least one electronic device for adjusting the temperature of the battery stack in accordance with the expected temperature and the predicted temperature is determined in accordance with the preset target, including: Determining a minimum value of a second objective function based on a directional acceleration optimization algorithm, wherein the second objective function represents a weighted sum of a deviation between the desired temperature and the predicted temperature and energy consumption generated by the at least one electronic device; When the second objective function obtains a minimum value, a control strategy corresponding to the at least one electronic device is determined.
4. The method according to claim 1, characterized in that: The multilayer perceptron includes an input layer, a first filter layer, a hidden layer, a second filter layer and an output layer in the order of data processing. The temperature prediction model is obtained by training through the following steps: Inputting the historical fuel cell system data into the input layer, so that the input layer performs feature extraction on the historical fuel cell system data to obtain first feature data; Inputting the first characteristic data into the first filtering layer so that the first filtering layer randomly discards data in a first proportion of neurons to obtain second characteristic data; Inputting the second feature data into the hidden layer so that the hidden layer performs feature extraction on the second feature data to obtain third feature data; Inputting the third characteristic data into the second filtering layer so that the second filtering layer randomly discards data in neurons of a second proportion, thereby obtaining fourth characteristic data; The fourth characteristic data is input into the output layer so that the output layer processes the fourth characteristic data and outputs the predicted temperature of the fuel cell stack.
5. The method according to claim 4, characterized in that When the multilayer perceptron is trained using the historical fuel cell system data, a back propagation algorithm is used to train the multilayer perceptron, and a cross-validation method and a grid search method are used to optimize the hyperparameters of the multilayer perceptron.
6. The method according to any one of claims 1 to 5, characterized in that: The at least one type includes the pressure of hydrogen when entering the fuel cell stack, the pressure of air when entering the fuel cell stack, and the temperature of cooling water when entering the fuel cell stack; the at least one electronic device includes one or more of a cooling fan, a cooling water circulation pump, and an electronic thermostat; in addition to the at least one type, the multiple types also include one or more of the output power of the cooling fan, the output power of the cooling water circulation pump, and the opening of the electronic thermostat, and the opening of the electronic thermostat is used to control the flow rate of cooling water entering each cooling circulation system related to the fuel cell stack.
7. The method according to claim 6, characterized in that The control strategy corresponding to the at least one electronic device includes one or more information of a target output power of a cooling fan, a target output power of a cooling water circulation pump, and a target opening of the electronic thermostat; and adjusting the at least one electronic device according to the control strategy includes: If the at least one electronic device includes the cooling fan, adjusting the cooling fan according to the target output power of the cooling fan; If the at least one electronic device includes the cooling water circulation pump, regulating the cooling water circulation pump according to the target output power of the cooling water circulation pump; If the at least one electronic device includes the electronic thermostat, the electronic thermostat is adjusted according to the target opening of the electronic thermostat.
8. A stack temperature control device, characterized in that: include: A first acquisition module, configured to acquire at least one type of fuel cell system data related to the temperature of the fuel cell stack at a current moment, and an expected temperature for the fuel cell stack; A second acquisition module is used to input the at least one type of fuel cell system data into a pre-trained temperature prediction model to obtain a predicted temperature of the fuel cell stack after a preset time period, wherein the temperature prediction model is a neural network model trained based on a multi-layer perceptron using multiple types of historical fuel cell system data related to the temperature of the fuel cell stack as input data, wherein the multiple types include the at least one type; A determination module, configured to determine, based on the expected temperature and the predicted temperature, a control strategy corresponding to at least one electronic device for adjusting the temperature of the battery stack in accordance with a preset target; The regulating module is used to regulate the at least one electronic device according to the control strategy to adjust the temperature of the battery stack to the desired temperature.
9. A control device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute to implement a stack temperature control method as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by a processor of a control device, the control device is enabled to implement a stack temperature control method as described in any one of claims 1 to 7.
Citation Information
Cited By
Intelligent design method of fuel cell membrane electrode
CN120413721A
Constant-temperature pressure tracing method and system
CN122282194A
