Method, device, equipment and medium for estimating the cruising range of a fuel cell vehicle

Through LSTM model and transfer learning technology, the problems of operating mode, operating condition changes and insufficient data samples in the range estimation of fuel cell vehicles are solved, the prediction accuracy and generalization capabilities are improved, and more accurate range prediction is achieved.

CN115453394BActive Publication Date: 2024-11-05TSINGHUA UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211117828.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-09-14
Publication Date
2024-11-05
Estimated Expiration
2042-09-14

AI Technical Summary

Technical Problem

The existing fuel cell vehicle range estimation methods fail to accurately consider vehicle operating mode, operating conditions and power changes, and lack of some vehicle model data samples, resulting in weak generalization capabilities and affecting the prediction accuracy.

Method used

Using a range prediction model based on LSTM, the continuous driving data at the current moment is obtained, input into the preset model and iteratively calculates it. The model is trained in combination with historical data to enhance generalization ability, and fine-tune the model using transfer learning technology to improve prediction accuracy .

Benefits of technology

It improves the accuracy of fuel cell vehicle range prediction, enhances the generalization ability of the model, solves the problems of operating mode, operating conditions and insufficient data samples, and achieves more accurate real-vehicle data prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115453394B_ABST
    Figure CN115453394B_ABST
Patent Text Reader

Abstract

The present application relates to the field of industrial big data mining technology, and in particular to a method, device, equipment and medium for estimating the cruising range of a fuel cell vehicle. The method includes: obtaining the continuous driving data of the fuel cell vehicle at the current moment, inputting it into a preset cruising range prediction model, and outputting the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment, and inputting it into the preset cruising range prediction model as new continuous driving data for iterative calculation, thereby obtaining the cruising range estimation result of the fuel cell vehicle. Thus, the estimation method in the related art reduces the accuracy of changes in the vehicle's operating mode, operating conditions, power, etc., lacks data samples of some models, and has weak generalization ability. Through the estimation method of the cruising range of fuel cell vehicles based on LSTM and transfer learning, the generalization ability of the prediction model is enhanced, thereby improving the accuracy of real vehicle data prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of industrial big data mining, and in particular to a method, device, equipment and medium for estimating the cruising range of a fuel cell vehicle. Background Art

[0002] Traditional vehicles are mainly powered by fossil energy such as coal and oil. A large amount of fossil energy will produce a lot of pollutants and greenhouse gases during use, thus aggravating the world's environmental pollution and global warming. At the same time, fossil energy, as a non-renewable energy, has limited surplus after years of mining. Therefore, it is necessary to develop other energy sources to power vehicles.

[0003] Fuel cell vehicles are driven by the electricity generated by the chemical reaction of hydrogen and oxygen and the charging of power batteries. As a renewable energy source, hydrogen has high energy efficiency and low pollution, which are in line with the current needs of solving environmental pollution, climate warming caused by high carbon emissions, and fuel energy depletion. At present, domestic fuel cell vehicles are mainly used in the transportation and public transportation industries. Under the existing layout of hydrogen refueling stations, fuel cell vehicles need to strictly plan driving routes and energy use, and remind drivers of the current vehicle's range so that they can refuel in a planned manner.

[0004] In related technologies, most of them calculate the current range based on the current hydrogen consumption rate, utilization rate and remaining amount combined with the current vehicle speed, or perform regression calculation of the remaining mileage based on a large amount of operating condition characteristic data, fuel cell and hydrogen quality data.

[0005] However, this estimation method also has some problems, such as not using real vehicle data for prediction and verification, not considering the accuracy of simple estimation methods for changes in vehicle operating modes, operating conditions, and power, lack of data samples for some vehicle driving types, and the limitations of deep learning models with large amounts of data due to their weak generalization capabilities for various vehicle models. Summary of the invention

[0006] The present application provides a method, device, equipment and medium for estimating the cruising range of a fuel cell vehicle to solve the problems that the estimation method in the related art will reduce the accuracy of changes in the vehicle's operating mode, operating conditions and power, lack of data samples of some vehicle types and weak generalization ability.

[0007] The first embodiment of the present application provides a method for estimating the cruising range of a fuel cell vehicle, comprising the following steps:

[0008] Obtaining continuous driving data of the fuel cell vehicle at the current moment;

[0009] Inputting the continuous driving data at the current moment into a preset cruising range prediction model, and outputting the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment;

[0010] The estimated vehicle energy consumption and estimated speed at the next moment are used as new continuous driving data and input into the preset cruising range prediction model for iterative calculation until the preset iteration conditions are met to obtain the cruising range estimation result of the fuel cell vehicle.

[0011] According to one embodiment of the present application, before inputting the continuous driving data at the current moment into the preset cruising range prediction model, the method further includes:

[0012] Acquire historical continuous driving data of a plurality of target fuel cell vehicles, wherein the continuous driving data includes at least one of continuous driving time, hydrogen storage tank temperature, hydrogen storage tank pressure, hydrogen storage tank volume, remaining capacity of power battery, vehicle driving speed and battery pack output power;

[0013] Generating training samples according to the historical continuous driving data, and using the training samples to train a basic model for cruising range to obtain an initial cruising range prediction model;

[0014] The initial cruising range prediction model is fine-tuned and trained based on the continuous driving data to obtain the preset cruising range prediction model.

[0015] According to an embodiment of the present application, generating training samples according to the historical continuous driving data includes:

[0016] Based on a preset hydrogen mass calculation formula, the remaining hydrogen mass at each moment in the historical continuous driving data is calculated to obtain first driving data;

[0017] Performing nonlinear fitting on the first driving data to obtain a fitting curve, calculating the variance of each data in the historical continuous driving data with respect to the fitting curve, and cleaning out the data that meets a preset cleaning condition based on the variance to obtain second driving data;

[0018] Based on a preset data sample format, the second driving data is structured to construct data sets of different working conditions with the same data volume, and the data sets are normalized by a preset conversion function to obtain the training samples.

[0019] According to one embodiment of the present application, the cruising range basic model is:

[0020] i t =σ(W xi x t +W hi ht-1 +W ci c t-1 +b i );

[0021] f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f );

[0022] c t =f t c t-1 +i t tanh(W xc x t +W hc h t-1 +b c );

[0023] o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o );

[0024] h t =o t tanh(c t );

[0025] Among them, σ represents the sigmoid function, i t represents the input gate, f t represents the forget gate, o t Indicates the output, c t represents the unit activation vector, h t is the hidden layer unit, b is the deviation value of each item, t is the vehicle running time, tanh is the activation function, x t is the data at the current moment, h t-1 is the output data of the previous moment, W xi is the weight matrix of the input gate and the input feature vector, W hi is the weight matrix of the input gate and the hidden layer unit vector, W ci is the weight matrix of the input gate and the activation vector, b i is the input gate deviation value, W xf is the weight matrix of the forget gate and the input feature vector, W hf is the weight matrix of the forget gate and the hidden layer unit vector, W cfis the weight matrix of the forget gate and the activation vector, b f is the forget gate bias value, W xc is the weight matrix of the unit activation vector and the input feature vector, W hc is the weight matrix of the unit activation vector and the hidden layer unit vector, b c is the unit activation vector bias value, W xo is the weight matrix of the output gate and the input feature vector, W ho is the weight matrix of the output gate and the hidden layer unit vector, W co is the weight matrix of the output gate and the hidden layer unit vector, b o is the output gate deviation value.

[0026] According to one embodiment of the present application, the preset iteration condition is:

[0027]

[0028]

[0029] Among them, M t0 is the remaining hydrogen mass at the current moment, SOC t0 is the remaining power of the power battery at the current moment, δ M is the vehicle energy consumption estimate for the next moment, δ SOC is the vehicle speed estimate for the next moment, t0 is the current moment, and n is the number of iterations.

[0030] According to one embodiment of the present application, obtaining the estimated result of the cruising range of the fuel cell vehicle includes:

[0031] Based on a preset range estimation formula, an estimated range result of the fuel cell vehicle is obtained, wherein the preset range estimation formula is:

[0032]

[0033] Among them, v t is the current vehicle speed, v t+1 is the vehicle speed at the next moment, and Δt is the time difference between the two moments.

[0034] According to the method for estimating the cruising range of a fuel cell vehicle of the embodiment of the present application, the continuous driving data of the fuel cell vehicle at the current moment is obtained, input into a preset cruising range prediction model, and the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment are output, and the data is input into the preset cruising range prediction model as new continuous driving data for iterative calculation, thereby obtaining the cruising range estimation result of the fuel cell vehicle. Thus, the estimation method in the related art reduces the accuracy of changes in the vehicle's operating mode, operating conditions, power, etc., lacks data samples of some models, and has weak generalization ability. Through the method for estimating the cruising range of a fuel cell vehicle based on LSTM (Long Short-Term Memory, long short-term memory network) and transfer learning, the generalization ability of the prediction model is enhanced, thereby improving the accuracy of the prediction of real vehicle data.

[0035] A second aspect of the present application provides a device for estimating the cruising range of a fuel cell vehicle, comprising:

[0036] An acquisition module, used to acquire the continuous driving data of the fuel cell vehicle at the current moment;

[0037] An output module, used to input the continuous driving data at the current moment into a preset cruising range prediction model, and output the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment;

[0038] The calculation module is used to input the estimated vehicle energy consumption and estimated speed at the next moment as new continuous driving data into the preset cruising range prediction model for iterative calculation until the preset iteration conditions are met, thereby obtaining the cruising range estimation result of the fuel cell vehicle.

[0039] According to one embodiment of the present application, before inputting the continuous driving data at the current moment into the preset cruising range prediction model, the output module further includes:

[0040] An acquisition unit, used to acquire historical continuous driving data of a plurality of target fuel cell vehicles, wherein the continuous driving data includes at least one of continuous driving time, hydrogen storage tank temperature, hydrogen storage tank pressure, hydrogen storage tank volume, remaining capacity of power battery, vehicle driving speed and battery pack output power;

[0041] A training unit, configured to generate training samples according to the historical continuous driving data, and use the training samples to train a basic model of the cruising range to obtain an initial cruising range prediction model;

[0042] A fine-tuning unit is used to fine-tune and train the initial cruising range prediction model based on the continuous driving data to obtain the preset cruising range prediction model.

[0043] According to one embodiment of the present application, the training unit is specifically used for:

[0044] Based on a preset hydrogen mass calculation formula, the remaining hydrogen mass at each moment in the historical continuous driving data is calculated to obtain first driving data;

[0045] Performing nonlinear fitting on the first driving data to obtain a fitting curve, calculating the variance of each data in the historical continuous driving data with respect to the fitting curve, and cleaning out the data that meets a preset cleaning condition based on the variance to obtain second driving data;

[0046] Based on a preset data sample format, the second driving data is structured to construct data sets of different working conditions with the same data volume, and the data sets are normalized by a preset conversion function to obtain the training samples.

[0047] According to one embodiment of the present application, the cruising range basic model is:

[0048] i t =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i );

[0049] f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f );

[0050] c t =f t c t-1 +i t tanh(W xc x t +W hc h t-1 +b c );

[0051] o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o );

[0052] h t =o t tanh(c t );

[0053] Among them, σ represents the sigmoid function, i t represents the input gate, f t represents the forget gate, o t Indicates the output, c t represents the unit activation vector, h t is the hidden layer unit, b is the deviation value of each item, t is the vehicle running time, tanh is the activation function, x t is the data at the current moment, h t-1 is the output data of the previous moment, W xi is the weight matrix of the input gate and the input feature vector, W hi is the weight matrix of the input gate and the hidden layer unit vector, W ci is the weight matrix of the input gate and the activation vector, b i is the input gate deviation value, W xf is the weight matrix of the forget gate and the input feature vector, W hf is the weight matrix of the forget gate and the hidden layer unit vector, W cf is the weight matrix of the forget gate and the activation vector, b f is the forget gate bias value, W xc is the weight matrix of the unit activation vector and the input feature vector, W hc is the weight matrix of the unit activation vector and the hidden layer unit vector, b c is the unit activation vector bias value, W xo is the weight matrix of the output gate and the input feature vector, W ho is the weight matrix of the output gate and the hidden layer unit vector, W co is the weight matrix of the output gate and the hidden layer unit vector, b o is the output gate deviation value.

[0054] According to one embodiment of the present application, the preset iteration condition is:

[0055]

[0056]

[0057] Among them, M t0 is the remaining hydrogen mass at the current moment, SOC t0 is the remaining power of the power battery at the current moment, δ M is the vehicle energy consumption estimate for the next moment, δ SOCis the vehicle speed estimate for the next moment, t0 is the current moment, and n is the number of iterations.

[0058] According to one embodiment of the present application, the computing module includes:

[0059] A calculation unit is used to obtain a cruising range estimation result of the fuel cell vehicle based on a preset cruising range estimation formula, wherein the preset cruising range estimation formula is:

[0060]

[0061] Among them, v t is the current vehicle speed, v t+1 is the vehicle speed at the next moment, and Δt is the time difference between the two moments.

[0062] According to the cruising range estimation device of the fuel cell vehicle of the embodiment of the present application, by obtaining the continuous driving data of the fuel cell vehicle at the current moment, inputting it into the preset cruising range prediction model, and outputting the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment, it is input as new continuous driving data into the preset cruising range prediction model for iterative calculation, thereby obtaining the cruising range estimation result of the fuel cell vehicle. Thus, the estimation method in the related art reduces the accuracy of changes in the vehicle's operating mode, operating conditions, power, etc., lacks data samples of some models, and has weak generalization ability. Through the estimation method of the cruising range of the fuel cell vehicle based on LSTM and transfer learning, the generalization ability of the prediction model is enhanced, thereby improving the accuracy of the prediction of the actual vehicle data.

[0063] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for estimating the cruising range of a fuel cell vehicle as described in the above embodiment.

[0064] The fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the method for estimating the cruising range of a fuel cell vehicle as described in the above embodiment.

[0065] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0067] Figure 1 A flowchart of a method for estimating the cruising range of a fuel cell vehicle according to an embodiment of the present application;

[0068] Figure 2 It is an overall framework flow chart according to an embodiment of the present application;

[0069] Figure 3 A schematic diagram of the LSTM model structure according to an embodiment of the present application;

[0070] Figure 4 Schematic diagram of a block diagram of a cruising range estimation device for a fuel cell vehicle according to an embodiment of the present application;

[0071] Figure 5 Schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0072] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0073] The following describes the range estimation method, device, equipment and medium of the fuel cell vehicle proposed according to the embodiment of the present invention with reference to the accompanying drawings. In view of the problem that the estimation method in the related art mentioned in the above background technology will reduce the accuracy of the changes in the vehicle's operating mode, operating conditions and power, lack of data samples of some vehicle types and weak generalization ability, the present application provides a method for estimating the range of a fuel cell vehicle, in which the continuous driving data of the fuel cell vehicle at the current moment is obtained, input into a preset range prediction model, and output the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment, and input it as new continuous driving data into the preset range prediction model for iterative calculation, thereby obtaining the range estimation result of the fuel cell vehicle. Thus, the estimation method in the related art will reduce the accuracy of the changes in the vehicle's operating mode, operating conditions and power, lack of data samples of some models and weak generalization ability, and the like, and the estimation method of the range of a fuel cell vehicle based on LSTM and transfer learning is used to enhance the generalization ability of the prediction model, thereby improving the accuracy of the prediction of the actual vehicle data.

[0074] Specifically, Figure 1 It is a flow chart of a method for estimating the cruising range of a fuel cell vehicle according to an embodiment of the present application.

[0075] like Figure 1 As shown, the method for estimating the cruising range of a fuel cell vehicle includes the following steps:

[0076] In step S101, the continuous driving data of the fuel cell vehicle at the current moment is obtained.

[0077] Specifically, Figure 2 As shown, the embodiment of the present application first needs to collect platform data and continuous driving data of the fuel cell vehicle at the current moment through the real-time device of the vehicle-mounted power system, save the collected real-time data, and then upload it to the unified computing platform through the vehicle's 32960 and other communication protocol types.

[0078] Secondly, the long-term data of multiple fuel cell vehicles are collected through the communication protocol type between the platform and the vehicle as the original data of the basic model, and the real-time data of each vehicle for nearly 2 days is saved in the vehicle's software system and uploaded to the unified computing platform as the original data of the terminal model.

[0079] In step S102, the continuous driving data at the current moment is input into a preset cruising range prediction model, and the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment are output.

[0080] Among them, the preset cruising range prediction model can be set accordingly by technical personnel in this field according to the actual prediction function of the vehicle cruising range estimation method, and is not specifically limited here.

[0081] Specifically, after acquiring the continuous driving data of the fuel cell at the current moment, the embodiment of the present application inputs the data into a preset range prediction model, and outputs the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment.

[0082] Furthermore, in some embodiments, before the continuous driving data at the current moment is input into the preset cruising range prediction model, it also includes: obtaining historical continuous driving data of multiple target fuel cell vehicles, wherein the continuous driving data includes at least one of continuous driving time, hydrogen storage tank temperature, hydrogen storage tank pressure, hydrogen storage tank volume, remaining capacity of the power battery, vehicle driving speed and battery pack output power; generating training samples based on the historical continuous driving data, and using the training samples to train the cruising range basic model to obtain an initial cruising range prediction model; fine-tuning the training of the initial cruising range prediction model based on the continuous driving data to obtain a preset cruising range prediction model.

[0083] Specifically, before inputting the continuous driving data at the current moment into the preset cruising range prediction model, the embodiment of the present application also needs to collect historical continuous driving data of multiple target fuel cell vehicles, mainly including continuous driving time, hydrogen storage tank temperature, hydrogen storage tank pressure, hydrogen storage tank volume, remaining capacity of the power battery, vehicle driving speed and at least one of the battery pack output power.

[0084] Furthermore, after collecting the historical continuous driving data of multiple target fuel cell vehicles, the embodiment of the present application also needs to collect long-term data of multiple fuel cell vehicles through the communication protocol between the platform and the vehicle as the original data of the basic model of the cruising range, that is, training samples, and then use the training samples to train the basic model of the cruising range to obtain an initial cruising range prediction model, and then save the real-time data of each vehicle for nearly 2 days in the vehicle's software system as the original data of the terminal model, that is, to obtain the preset cruising range prediction model.

[0085] Furthermore, in some embodiments, training samples are generated based on historical continuous driving data, including: based on a preset hydrogen mass calculation formula, calculating the remaining hydrogen mass at each moment in the historical continuous driving data to obtain first driving data; performing nonlinear fitting on the first driving data to obtain a fitting curve, and calculating the variance of each data in the historical continuous driving data with respect to the fitting curve, and based on the variance, cleaning out data that meets a preset cleaning condition to obtain second driving data; based on a preset data sample format, performing structured processing on the second driving data to construct a data set of different operating conditions with the same data volume, and performing data normalization processing on the data set through a preset conversion function to obtain training samples.

[0086] Specifically, Figure 2 As shown, in the process of generating training samples according to the historical continuous driving data, the embodiment of the present application needs to perform data preprocessing on the above-collected data. First, data conversion needs to be performed, that is, based on the preset hydrogen mass calculation formula, the remaining hydrogen mass at each time node in the historical continuous driving data is calculated, so as to obtain the first driving data. The specific expression is as follows:

[0087]

[0088] Among them, m is the molar mass of hydrogen molecules (g / mol), V is the total volume of the fuel cell vehicle hydrogen storage tank (L), R is the common gas constant ((MPa*L) / (Mol*K)), P is the pressure of the fuel cell vehicle hydrogen storage tank (MPa), T is the temperature inside the fuel cell vehicle hydrogen storage tank (K), and Z is the hydrogen compression factor at the current pressure and temperature.

[0089] Secondly, if Figure 2As shown, after the data conversion is performed, the embodiment of the present application needs to perform data cleaning on the obtained first driving data. The main method includes: first, removing the outliers of the data, and then performing nonlinear fitting on the data points of the speed, cumulative mileage, hydrogen quality and power of the power battery of the fuel cell vehicle to obtain the fitting curve equation, and calculating the variance of each data in the historical continuous driving data to the fitting curve, and based on the variance, cleaning the data that meets the preset cleaning conditions, that is, when the variance is too large, judging that the data value is abnormal and removing it, thereby obtaining the second driving data.

[0090] Again, if Figure 2 As shown, the embodiment of the present application needs to perform structured processing on the second driving data based on a preset data sample format, so as to construct data sets of different working conditions with the same data volume.

[0091] Specifically, the embodiment of the present application can divide the vehicle into three driving states according to the ratio of the sum of the current hydrogen mass of all hydrogen cylinders to the sum of the hydrogen mass of all hydrogen cylinders when they are full: a multi-hydrogen driving state, a low-hydrogen driving state, and a pure electric driving state. Then, each adjacent 30 time units of data points constitute a data sample, and so on, thereby forming a data set that is always continuous in time; at the same time, the data at the starting point is changed to construct a new data sample of 30 time units, and so on, thereby forming a new set of new data sets that are always continuous in time. At this time, the new data set can double the amount of data, and then continue to follow the above method until a data set of different working conditions with equivalent data volume is constructed. Among them, the format of generating a set of data samples is as follows:

[0092]

[0093] Y n =[v i+30 M i+30 SOC i+30 ]

[0094] Where i is time (s), v i is the current speed of the vehicle at time i (m / s), M i is the current remaining hydrogen amount of the vehicle at time i (kg), SOC i is the current power battery capacity of the vehicle at time i (%), P i is the battery output power at time i (W). It should be noted that X n Each row in the table represents the data of a vehicle at a certain point in time, and the two rows above and below it are continuous.

[0095] Finally, if Figure 2As shown, in the data used in the embodiments of the present application, the data distribution of different vehicles is different due to different vehicle models, performance degradation of vehicle energy systems, drivers, driving sections, etc. Therefore, it is necessary to normalize all relevant parameters separately through a conversion function to obtain training samples.

[0096] Optionally, the method adopted in the embodiment of the present application is to use linear normalization to perform linear transformation on the original data so that the result falls into the interval [0,1], wherein the conversion function can be expressed as:

[0097]

[0098] Among them, b is the original data, b max is the maximum value of the data, b min is the minimum value of the data.

[0099] Furthermore, if Figure 2 As shown, after the data is preprocessed, the embodiment of the present application needs to train the basic model of the cruising range. Since RNN (Recurrent Neural Network) is different from other neural networks, it is a neural network used to process sequence data, and based on recursive operation logic, the output data h of the recurrent neural network at a certain time step = t t , based on the current data x t and the output data h at the previous moment t-1 , when the output h at the last moment is calculated t When t0 is t, the network contains information from all moments from t0 to t. Therefore, in the application of fuel cell vehicle driving range, when the time axis is accumulated for a long time, the timestep will be very large, resulting in t The back propagation process to h0 may cause gradient explosion or gradient disappearance problems.

[0100] Therefore, if Figure 3 As shown, in order to solve this problem, the embodiment of the present application adopts LSTM recurrent neural network, that is, three gate units (gates) are added to the RNN unit, namely input gate, forget gate and output gate. Through a sigmoid neural layer and a point-by-point multiplication operation, the gate can selectively allow information to pass. Among them, the specific formula of the LSTM recurrent neural network model is as follows:

[0101] i t =σ(W xi x t +W hi h t-1+W ci c t-1 +b i )

[0102] f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f )

[0103] c t =f t c t-1 +i t tanh(W xc x t +W hc h t-1 +b c )

[0104] o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o )

[0105] h t =o t tanh(c t )

[0106] Among them, σ represents the sigmoid function, i t represents the input gate, f t represents the forget gate, o t Indicates the output, c t represents the unit activation vector, h t is the hidden layer unit, b is the deviation value of each item, t is the vehicle running time, tanh is the activation function, x t is the data at the current moment, h t-1 is the output data of the previous moment, W xi is the weight matrix of the input gate and the input feature vector, W hi is the weight matrix of the input gate and the hidden layer unit vector, W ci is the weight matrix of the input gate and the activation vector, b i is the input gate deviation value, W xf is the weight matrix of the forget gate and the input feature vector, W hf is the weight matrix of the forget gate and the hidden layer unit vector, W cfis the weight matrix of the forget gate and the activation vector, b f is the forget gate bias value, W xc is the weight matrix of the unit activation vector and the input feature vector, W hc is the weight matrix of the unit activation vector and the hidden layer unit vector, b c is the unit activation vector bias value, W xo is the weight matrix of the output gate and the input feature vector, W ho is the weight matrix of the output gate and the hidden layer unit vector, W co is the weight matrix of the output gate and the hidden layer unit vector, b o is the output gate deviation value.

[0107] Furthermore, the LSTM recurrent neural network model adopted in the embodiment of the present application is transmitted to the vehicle terminal system via the communication network, and terminal fine-tuning training data preparation is performed.

[0108] Specifically, the data preprocessing method based on the cruising range basic model in the embodiment of the present application preprocesses the real-time data of the fuel cell vehicle, that is, after data conversion, data cleaning, structured processing and normalization processing, since the data set does not need to be segmented and reorganized based on the data ratio when the structured processing is performed at the vehicle terminal, only the time difference δ of each starting point needs to be calculated. t0 and data length δ t Therefore, the embodiment of the present application needs to transfer learning of the basic model and fine-tune training on the terminal. It should be noted that the transfer model only needs to achieve the goal of fitting the data of the vehicle, and does not need to consider the generalization ability of the model to multiple vehicles. The maximum value b of the vehicle is used in the normalization process. max , minimum value b min And other parameters.

[0109] Furthermore, the embodiment of the present application uses a pre-trained model that has been trained on a large data set before, keeps the original structure and initial weights of the model unchanged during the fine-tuning process, and then retrains based on the new data set to fine-tune the vehicle data. In other words, the embodiment of the present application pre-processes the collected real-time data, i.e., converts, cleans, structures, and normalizes it, and inputs it into the fine-tuning model to obtain the vehicle energy consumption estimate and speed estimate at the next moment to achieve higher accuracy. Among them, the expressions for the vehicle energy consumption estimate and speed estimate at the next moment are respectively:

[0110] δ M =M t+1 -M t0

[0111] δ SOC =SOC t+1 -SOCt0

[0112] v t+1

[0113] Among them, M t+1 The remaining hydrogen mass at the next moment, SOC t+1 is the remaining power of the power battery at the next moment, v t+1 is the vehicle speed at the next moment, M t0 is the remaining hydrogen mass at the current moment, SOC t0 The remaining power of the power battery at the current moment.

[0114] In step S103, the estimated vehicle energy consumption and estimated speed at the next moment are used as new continuous driving data and input into a preset cruising range prediction model for iterative calculation until the preset iteration conditions are met to obtain the estimated cruising range of the fuel cell vehicle.

[0115] Specifically, the embodiment of the present application uses the estimated vehicle energy consumption and estimated speed at the next moment obtained above as new continuous driving data, and inputs them into the preset cruising range prediction model for iterative calculation, wherein the preset iteration condition is to iterate the prediction of the next time node multiple times until the remaining energy is cleared, that is, the expression is satisfied:

[0116]

[0117]

[0118] Among them, M t0 is the remaining hydrogen mass at the current moment, SOC t0 is the remaining power of the power battery at the current moment, δ M is the vehicle energy consumption estimate for the next moment, δ SOC is the vehicle speed estimate for the next moment, t0 is the current moment, and n is the number of iterations.

[0119] Furthermore, in some embodiments, obtaining an estimated result of the cruising range of the fuel cell vehicle includes: obtaining an estimated result of the cruising range of the fuel cell vehicle based on a preset cruising range estimation formula.

[0120] Specifically, when the embodiment of the present application iterates the prediction of the next time node multiple times until the remaining energy is cleared, and based on the preset driving range estimation formula, the estimated range of the fuel cell vehicle is obtained, wherein the preset driving range estimation formula is:

[0121]

[0122] Among them, v t is the current vehicle speed, v t+1is the vehicle speed at the next moment, and Δt is the time difference between the two moments.

[0123] In summary, the method for estimating the cruising range of a fuel cell vehicle based on LSTM and transfer learning in the embodiment of the present application brings the following technical effects:

[0124] (1) Use massive amounts of data to provide more accurate predictions based on real vehicle data.

[0125] (2) Using a recurrent neural network ensures that model training is not affected by the lack of certain types of data.

[0126] (3) Based on the consumption logic training of massive data, the focus is on strengthening the data learning of the current vehicle. Under the model with strong logic and memory, it can highly fit the driving data of the current vehicle to enhance the generalization ability of the model and achieve more accurate predictions on each vehicle.

[0127] According to the method for estimating the cruising range of a fuel cell vehicle of the embodiment of the present application, the continuous driving data of the fuel cell vehicle at the current moment is obtained, input into a preset cruising range prediction model, and the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment are output, and the data is input into the preset cruising range prediction model as new continuous driving data for iterative calculation, thereby obtaining the cruising range estimation result of the fuel cell vehicle. Thus, the estimation method in the related art reduces the accuracy of changes in the vehicle's operating mode, operating conditions, power, etc., lacks data samples of some models, and has weak generalization ability. Through the method for estimating the cruising range of a fuel cell vehicle based on LSTM and transfer learning, the generalization ability of the prediction model is enhanced, thereby improving the accuracy of the prediction of real vehicle data.

[0128] Next, a cruising range estimation device for a fuel cell vehicle according to an embodiment of the present invention will be described with reference to the accompanying drawings.

[0129] Figure 4 It is a block diagram of a device for estimating the cruising range of a fuel cell vehicle according to an embodiment of the present application.

[0130] like Figure 4 As shown, the cruising range estimation device 10 of a fuel cell vehicle includes: an acquisition module 100 , an output module 200 and a calculation module 300 .

[0131] Wherein, the acquisition module 100 is used to acquire the continuous driving data of the fuel cell vehicle at the current moment;

[0132] The output module 200 is used to input the continuous driving data at the current moment into a preset cruising range prediction model, and output the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment;

[0133] The calculation module 300 is used to input the estimated vehicle energy consumption and estimated speed at the next moment as new continuous driving data into a preset cruising range prediction model for iterative calculation until the preset iteration conditions are met to obtain the estimated cruising range of the fuel cell vehicle.

[0134] Furthermore, in some embodiments, before inputting the continuous driving data at the current moment into the preset cruising range prediction model, the output module 200 further includes:

[0135] An acquisition unit, used to acquire historical continuous driving data of a plurality of target fuel cell vehicles, wherein the continuous driving data includes at least one of continuous driving time, hydrogen storage tank temperature, hydrogen storage tank pressure, hydrogen storage tank volume, remaining capacity of power battery, vehicle driving speed and battery pack output power;

[0136] A training unit, used to generate training samples according to historical continuous driving data, and use the training samples to train a basic model of the cruising range to obtain an initial cruising range prediction model;

[0137] The fine-tuning unit is used to fine-tune and train the initial cruising range prediction model based on the continuous driving data to obtain a preset cruising range prediction model.

[0138] Furthermore, in some embodiments, the training unit is specifically configured to:

[0139] Based on a preset hydrogen mass calculation formula, the remaining hydrogen mass at each moment in the historical continuous driving data is calculated to obtain the first driving data;

[0140] Performing nonlinear fitting on the first driving data to obtain a fitting curve, calculating the variance of each data in the historical continuous driving data with respect to the fitting curve, and cleaning out the data that meets a preset cleaning condition based on the variance to obtain the second driving data;

[0141] Based on a preset data sample format, the second driving data is structured to construct data sets of different working conditions with the same data volume, and the data sets are normalized through a preset conversion function to obtain training samples.

[0142] Furthermore, in some embodiments, the cruising range basic model is:

[0143] i t =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i );

[0144] f t=σ(W xf x t +W hf h t-1 +W cf c t-1 +b f );

[0145] c t =f t c t-1 +i t tanh(W xc x t +W hc h t-1 +b c );

[0146] o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o );

[0147] h t =o t tanh(c t );

[0148] Among them, σ represents the sigmoid function, i t represents the input gate, f t represents the forget gate, o t Indicates the output, c t represents the unit activation vector, h t is the hidden layer unit, b is the deviation value of each item, t is the vehicle running time, tanh is the activation function, x t is the data at the current moment, h t-1 is the output data of the previous moment, W xi is the weight matrix of the input gate and the input feature vector, W hi is the weight matrix of the input gate and the hidden layer unit vector, W ci is the weight matrix of the input gate and the activation vector, b i is the input gate deviation value, W xf is the weight matrix of the forget gate and the input feature vector, W hf is the weight matrix of the forget gate and the hidden layer unit vector, W cf is the weight matrix of the forget gate and the activation vector, b f is the forget gate bias value, W xc is the weight matrix of the unit activation vector and the input feature vector, W hcis the weight matrix of the unit activation vector and the hidden layer unit vector, b c is the unit activation vector bias value, W xo is the weight matrix of the output gate and the input feature vector, W ho is the weight matrix of the output gate and the hidden layer unit vector, W co is the weight matrix of the output gate and the hidden layer unit vector, b o is the output gate deviation value.

[0149] Furthermore, in some embodiments, the preset iteration condition is:

[0150]

[0151]

[0152] Among them, M t0 is the remaining hydrogen mass at the current moment, SOC t0 is the remaining power of the power battery at the current moment, δ M is the vehicle energy consumption estimate for the next moment, δ SOC is the vehicle speed estimate for the next moment, t0 is the current moment, and n is the number of iterations.

[0153] Furthermore, in some embodiments, the computing module includes:

[0154] The calculation unit is used to obtain the estimated cruising range of the fuel cell vehicle based on a preset cruising range estimation formula, wherein the preset cruising range estimation formula is:

[0155]

[0156] Among them, v t is the current vehicle speed, v t+1 is the vehicle speed at the next moment, and Δt is the time difference between the two moments.

[0157] According to the cruising range estimation device of the fuel cell vehicle of the embodiment of the present application, by obtaining the continuous driving data of the fuel cell vehicle at the current moment, inputting it into the preset cruising range prediction model, and outputting the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment, it is input as new continuous driving data into the preset cruising range prediction model for iterative calculation, thereby obtaining the cruising range estimation result of the fuel cell vehicle. Thus, the estimation method in the related art reduces the accuracy of changes in the vehicle's operating mode, operating conditions, power, etc., lacks data samples of some models, and has weak generalization ability. Through the estimation method of the cruising range of the fuel cell vehicle based on LSTM and transfer learning, the generalization ability of the prediction model is enhanced, thereby improving the accuracy of the prediction of the actual vehicle data.

[0158] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0159] A memory 501 , a processor 502 , and a computer program stored in the memory 501 and executable on the processor 502 .

[0160] When the processor 502 executes the program, the method for estimating the cruising range of a fuel cell vehicle provided in the above embodiment is implemented.

[0161] Furthermore, the electronic device further comprises:

[0162] The communication interface 505 is used for communication between the memory 501 and the processor 502 .

[0163] The memory 501 is used to store computer programs that can be executed on the processor 502 .

[0164] The memory 501 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0165] If the memory 501, the processor 502 and the communication interface 505 are implemented independently, the communication interface 505, the memory 501 and the processor 502 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0166] Optionally, in a specific implementation, if the memory 501, the processor 502 and the communication interface 505 are integrated on a chip, the memory 501, the processor 502 and the communication interface 505 can communicate with each other through an internal interface.

[0167] The processor 502 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0168] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for estimating the cruising range of a fuel cell vehicle.

[0169] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0170] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0171] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0172] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for estimating the cruising range of a fuel cell vehicle, characterized in that: The following steps are involved: Obtaining continuous driving data of the fuel cell vehicle at the current moment; Inputting the continuous driving data at the current moment into a preset cruising range prediction model, and outputting the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment; The estimated vehicle energy consumption and estimated speed at the next moment are used as new continuous driving data, and input into the preset cruising range prediction model for iterative calculation until the preset iteration conditions are met, thereby obtaining the cruising range estimation result of the fuel cell vehicle; The expressions of the estimated vehicle energy consumption and the estimated speed at the next moment are respectively: d M =M t+1 -M t0 δ SOC =SOC t+1 -SOC t0 v t+1 Among them, M t+1 The remaining hydrogen mass at the next moment, SOC t+1 is the remaining power of the power battery at the next moment, v t+1 is the vehicle speed at the next moment, M t0 is the remaining hydrogen mass at the current moment, SOC t0 The remaining power of the power battery at the current moment; The preset iteration condition is to iterate the prediction of the next time node multiple times until the remaining energy is cleared, that is, to satisfy the expression: Among them, M t0 is the remaining hydrogen mass at the current moment, SOC t0 is the remaining power of the power battery at the current moment, δ M is the vehicle energy consumption estimate for the next moment, δ SOC is the vehicle speed estimate for the next moment, t0 is the current moment, and n is the number of iterations.

2. The method according to claim 1, characterized in that Before inputting the continuous driving data at the current moment into the preset cruising range prediction model, the method further includes: Acquire historical continuous driving data of a plurality of target fuel cell vehicles, wherein the continuous driving data includes at least one of continuous driving time, hydrogen storage tank temperature, hydrogen storage tank pressure, hydrogen storage tank volume, remaining capacity of power battery, vehicle driving speed and battery pack output power; Generating training samples according to the historical continuous driving data, and using the training samples to train a basic model for cruising range to obtain an initial cruising range prediction model; The initial cruising range prediction model is fine-tuned and trained based on the continuous driving data to obtain the preset cruising range prediction model.

3. The method according to claim 2, characterized in that The generating of training samples according to the historical continuous driving data comprises: Based on a preset hydrogen mass calculation formula, the remaining hydrogen mass at each moment in the historical continuous driving data is calculated to obtain first driving data; Performing nonlinear fitting on the first driving data to obtain a fitting curve, calculating the variance of each data in the historical continuous driving data with respect to the fitting curve, and cleaning out the data that meets a preset cleaning condition based on the variance to obtain second driving data; Based on a preset data sample format, the second driving data is structured to construct data sets of different working conditions with the same data volume, and the data sets are normalized by a preset conversion function to obtain the training samples.

4. The method according to claim 2, characterized in that: The cruising range basic model is: i t =σ(W xi x t +W hi h t-1 +W ci c t-1 +b i ); f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f ); c t =f t c t-1 +i t tanh(W xc x t +W hc h t-1 +b c ); o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o ); h t =o t fishy t ); Among them, σ represents the sigmoid function, i t represents the input gate, f t represents the forget gate, o t Indicates the output, c t represents the unit activation vector, h t is the hidden layer unit, b is the deviation value of each item, t is the vehicle running time, tanh is the activation function, x t is the data at the current moment, h t-1 is the output data of the previous moment, W xi is the weight matrix of the input gate and the input feature vector, W hi is the weight matrix of the input gate and the hidden layer unit vector, W ci is the weight matrix of the input gate and the activation vector, b i is the input gate deviation value, W xf is the weight matrix of the forget gate and the input feature vector, W hf is the weight matrix of the forget gate and the hidden layer unit vector, W cf is the weight matrix of the forget gate and the activation vector, b f is the forget gate bias value, W xc is the weight matrix of the unit activation vector and the input feature vector, W hc is the weight matrix of the unit activation vector and the hidden layer unit vector, b c is the unit activation vector bias value, W xo is the weight matrix of the output gate and the input feature vector, W ho is the weight matrix of the output gate and the hidden layer unit vector, W co is the weight matrix of the output gate and the hidden layer unit vector, b o is the output gate deviation value.

5. The method according to claim 1, characterized in that The step of obtaining an estimated result of the cruising range of the fuel cell vehicle includes: Based on a preset range estimation formula, an estimated range result of the fuel cell vehicle is obtained, wherein the preset range estimation formula is: Among them, v t is the current vehicle speed, v t+1 is the vehicle speed at the next moment, and Δt is the time difference between the two moments.

6. A device for estimating the cruising range of a fuel cell vehicle, characterized in that: include: An acquisition module, used to acquire the continuous driving data of the fuel cell vehicle at the current moment; An output module, used to input the continuous driving data at the current moment into a preset cruising range prediction model, and output the estimated vehicle energy consumption and estimated speed of the fuel cell vehicle at the next moment; A calculation module, used to input the estimated vehicle energy consumption and estimated speed at the next moment as new continuous driving data into the preset cruising range prediction model for iterative calculation until a preset iteration condition is met, thereby obtaining a cruising range estimation result of the fuel cell vehicle; The expressions of the estimated vehicle energy consumption and the estimated speed at the next moment are respectively: d M =M t+1 -M t0 δ SOC =SOC t+1 -SOC t0 v t+1 Among them, M t+1 The remaining hydrogen mass at the next moment, SOC t+1 is the remaining power of the power battery at the next moment, v t+1 is the vehicle speed at the next moment, M t0 is the remaining hydrogen mass at the current moment, SOC t0 The remaining power of the power battery at the current moment; The preset iteration condition is to iterate the prediction of the next time node multiple times until the remaining energy is cleared, that is, to satisfy the expression: Among them, M t0 is the remaining hydrogen mass at the current moment, SOC t0 is the remaining power of the power battery at the current moment, δ M is the vehicle energy consumption estimate for the next moment, δ SOC is the vehicle speed estimate for the next moment, t0 is the current moment, and n is the number of iterations.

7. The device according to claim 6, characterized in that Before inputting the continuous driving data at the current moment into the preset cruising range prediction model, the output module further includes: An acquisition unit, used to acquire historical continuous driving data of a plurality of target fuel cell vehicles, wherein the continuous driving data includes at least one of continuous driving time, hydrogen storage tank temperature, hydrogen storage tank pressure, hydrogen storage tank volume, remaining capacity of power battery, vehicle driving speed and battery pack output power; A training unit, configured to generate training samples according to the historical continuous driving data, and use the training samples to train a basic model of the cruising range to obtain an initial cruising range prediction model; A fine-tuning unit is used to fine-tune and train the initial cruising range prediction model based on the continuous driving data to obtain the preset cruising range prediction model.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for estimating the cruising range of a fuel cell vehicle as described in any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method for estimating the cruising range of a fuel cell vehicle as described in any one of claims 1 to 5.