Energy consumption estimation method for hydrogen fuel cell vehicles based on energy consumption characteristics and operating condition segmentation

Abstract

The present invention provides a method for estimating the energy consumption of hydrogen fuel cell vehicles based on energy consumption characteristics and operating condition segmentation, taking into account the characteristics of the two power sources of such vehicles, the power battery and the fuel cell, and giving full play to the advantages of real vehicle big data to mine the energy consumption characteristics of the hydrogen fuel system. The process is more targeted than the existing technology. The method comprehensively considers the complexity of the operating conditions during driving, can identify the typical operating conditions during the driving of hydrogen fuel vehicles, and calculate the corresponding energy consumption for different typical operating conditions, which has higher accuracy than measuring energy consumption based on cyclic operating conditions or average energy consumption over a period of time. The execution of the method is based on data-driven, and there is no need to know the mechanism of vehicle driving energy consumption. According to the correlation features extracted by feature engineering combined with machine learning algorithms, faster and larger typical operating condition energy consumption estimates and future actual driving energy consumption predictions can be achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of hydrogen fuel cell vehicle energy management, and specifically relates to a method for estimating energy consumption of a hydrogen fuel cell vehicle based on energy consumption characteristics and operating condition segmentation. Background Art

[0002] At present, hydrogen fuel cell vehicles are generally equipped with two power sources, lithium-ion power batteries and hydrogen fuel cells. The power batteries can be charged by the excess energy output by the fuel cell. Therefore, the energy consumption calculation of the vehicle needs to be performed separately for the two batteries. If the calculation is combined, it may lead to inaccurate results or repeated calculations of the hydrogen consumption part. In the prior art, the energy consumption calculation of hydrogen fuel cell vehicles is mostly based on multiple test tests of the cycle working conditions for calibration, and there is often a large gap between the test environment and the actual driving process. Even in the same test environment, the energy consumption of different single vehicles of the same model will be different. Therefore, the calibration test needs to be repeatedly performed on a large number of vehicles, and the process is obviously time-consuming, labor-intensive and inefficient. There are also some methods in the prior art that calculate the energy consumption of hydrogen vehicles based on the actual driving data of the vehicle, and comprehensively estimate the energy consumption in combination with certain specific typical working conditions. However, the accuracy of this method is still limited by the shortcomings of insufficient coverage and applicability of typical working conditions to the actual situation, and thus cannot meet the requirements. Summary of the invention

[0003] In view of this, in order to solve the technical problems existing in the art, the present invention provides a method for estimating energy consumption of a hydrogen fuel cell vehicle based on energy consumption characteristics and operating condition segmentation, which specifically includes the following steps:

[0004] S1. The big data platform obtains the vehicle driving data collected and reported in real time by the intelligent vehicle terminals of multiple hydrogen fuel cell vehicles of the target model; the vehicle driving data is screened and preprocessed to obtain the original data containing multiple parameter fields related to vehicle energy consumption;

[0005] S2. Extract the output current and output voltage of the hydrogen fuel cell, the on / off state of the DCDC converter, and the hydrogen pressure from each frame of raw data, and combine these fields to determine whether the hydrogen fuel cell is consuming energy. If the judgment is "yes", calculate the energy consumption of the hydrogen fuel cell corresponding to each frame of raw data, and further extract the parameter fields reflecting the vehicle's driving conditions, such as the vehicle's factory year, vehicle operating status, charging status, time, instantaneous speed, cumulative mileage, SOC, instantaneous total current / voltage, and brake / accelerator pedal travel value, from the raw data as feature items; if the judgment is "no", do not perform the extraction of the above parameter fields and subsequent steps;

[0006] S3. Use the continuous parameter fields extracted in S2 to divide the vehicle into multiple complete motion segments from idling to starting driving and then to parking, and calculate the motion segment characteristics of each motion segment, including: mileage, driving time, average speed, SOC change, hydrogen pressure change, acceleration segment ratio, idling segment ratio, braking segment ratio, average acceleration / brake pedal stroke value, acceleration / brake pedal standard deviation, and total fuel system energy consumption; use the mileage, driving time, average speed, SOC change, and hydrogen pressure change in the motion segment characteristics to characterize the working condition characteristics corresponding to the motion segment, and use the acceleration segment ratio, idling segment ratio, braking segment ratio, average acceleration / brake pedal stroke value, and acceleration / brake pedal standard deviation to characterize the driving behavior characteristics of the driver in the motion segment;

[0007] S4, performing a K-Means clustering algorithm on the operating condition characteristics and driving behavior characteristics obtained in S3 with the total energy consumption of the fuel system as the target, and obtaining different typical operating condition types and driving behavior types corresponding to the motion segments; adding corresponding typical operating condition and driving behavior clustering labels to the motion segments, and calculating the average total fuel system energy consumption of different motion segments under the same operating condition and driving behavior type;

[0008] S5. Use the cluster labels obtained in S4, the average total fuel system energy consumption of different motion segments, and the characteristics of each motion segment in S3 to construct a test set and a training set, and train the random forest algorithm model to obtain a hydrogen fuel vehicle energy consumption estimation model; by executing the above steps on different target vehicle models, an energy consumption estimation model library containing multiple vehicle models is established; the corresponding data of the motion segment characteristics collected from a certain model of real vehicle during actual driving is input into the energy consumption estimation model library to estimate the real-time total fuel system energy consumption of vehicles of the same model.

[0009] Furthermore, the screening and preprocessing of vehicle driving data in S1 includes capturing vehicle driving data for a certain period of time, filling in missing value data, and correcting or deleting abnormal outlier data. For continuous data variables, the average value of the two previous and subsequent valid values ​​is used for filling, and for discrete and irregularly changing data, the data with the same valid value as the previous frame is used for filling.

[0010] Furthermore, S2 combines the hydrogen fuel cell output current and output voltage, the DCDC converter on / off state, and the hydrogen pressure to determine whether the hydrogen fuel cell is consuming energy. Specifically, it is determined that energy is being consumed when the hydrogen fuel cell output current is higher than the rated value, the DCDC converter is in the on state, and the hydrogen fuel pressure shows a downward trend. The hydrogen fuel cell energy consumption W corresponding to each frame of raw data is specifically calculated in the following way: W = ∫U f ×I f dt, where Uf Indicates the instantaneous output voltage of the hydrogen fuel cell, I f represents the instantaneous output current of the hydrogen fuel cell, and t represents the sampling interval between two frames of data.

[0011] Furthermore, each motion segment divided in S3 includes three randomly arranged stages of idling, acceleration and deceleration, and a uniform speed stage is added when the sampling frequency reaches a predetermined value; wherein, the idling stage is defined as a data segment in which the vehicle is in operation, the speed is 0 but the total current is not 0; the total fuel system energy consumption of the motion segment is determined by the sum of the fuel system energy consumption corresponding to each stage.

[0012] Furthermore, in S4, the motion segments are divided into three categories: urban conditions, suburban conditions, and high-speed conditions by executing a clustering algorithm, and the cluster center of each condition is determined by the average value of the characteristics of each motion segment. The driver behavior characteristics can be clustered according to a variety of different classification methods known to those skilled in the art, such as being divided into categories such as slow driving, stable driving, and aggressive driving.

[0013] Furthermore, when constructing the test set and training set in S5, the features of each motion segment are first standardized, and then the principal component analysis (PCA) method is used to reduce the data dimension. According to the principle that the contribution rate of motion segment features to energy consumption is >95%, the data dimension is further reduced to 8 dimensions;

[0014] The effect of training the random forest algorithm model is based on the model's estimated energy consumption for n motion clips. The actual energy consumption i The root mean square error (RMSE), mean square error (MSE), mean absolute percentage error (MAPE), and mean absolute error (MAE) are used to measure the difference. The specific calculation formulas are as follows:

[0015]

[0016]

[0017]

[0018]

[0019] Furthermore, in S5, several algorithm models such as SVR, BP neural network, linear fitting, GBDT and XGBoost are used to replace the random forest algorithm model, and the hydrogen fuel vehicle energy consumption estimation model is obtained by training with the test set and the training set.

[0020] The above-mentioned hydrogen fuel cell vehicle energy consumption estimation method based on energy consumption characteristics and operating condition segmentation provided by the present invention takes into account the characteristics of the two power sources of such vehicles, the power battery and the fuel cell, and gives full play to the advantages of real vehicle big data to mine the energy consumption characteristics of the hydrogen fuel system. The process is more targeted than the existing technology. The method comprehensively considers the complexity of the operating conditions during driving, can identify the typical operating conditions during the driving of hydrogen fuel vehicles, and calculate the corresponding energy consumption for different typical operating conditions, which has higher accuracy than measuring energy consumption based on cyclic operating conditions or average energy consumption over a period of time. The execution of the method is based on data-driven, and there is no need to know the mechanism of vehicle driving energy consumption. According to the correlation features extracted by feature engineering combined with machine learning algorithms, faster and larger typical operating condition energy consumption estimates and future actual driving energy consumption predictions can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of the process provided by the present invention;

[0022] Figure 2 It is a schematic diagram of clustering results of various typical working conditions in an example of the present invention;

[0023] Figure 3 are cluster centers of typical working conditions obtained in examples of the present invention;

[0024] Figure 4 It is a performance error indicator of the hydrogen fuel vehicle energy consumption prediction model trained in the example of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] The method for estimating energy consumption of a hydrogen fuel cell vehicle based on energy consumption characteristics and operating condition segmentation provided by the present invention is as follows: Figure 1 As shown, the specific steps include:

[0027] S1. The big data platform obtains the vehicle driving data collected and reported in real time by the intelligent vehicle terminals of multiple hydrogen fuel cell vehicles of the target model; the vehicle driving data is screened and preprocessed to obtain the original data containing multiple parameter fields related to vehicle energy consumption;

[0028] S2. Extract the output current and output voltage of the hydrogen fuel cell, the on / off state of the DCDC converter, and the hydrogen pressure from each frame of raw data, and combine these fields to determine whether the hydrogen fuel cell is consuming energy. If the judgment is "yes", calculate the energy consumption of the hydrogen fuel cell corresponding to each frame of raw data, and further extract the parameter fields reflecting the vehicle's driving conditions, such as the vehicle's factory year, vehicle operating status, charging status, time, instantaneous speed, cumulative mileage, SOC, instantaneous total current / voltage, and brake / accelerator pedal travel value, from the raw data as feature items; if the judgment is "no", do not perform the extraction of the above parameter fields and subsequent steps;

[0029] S3. Use the continuous parameter fields extracted in S2 to divide the vehicle into multiple complete motion segments from idling to starting driving and then to parking, and calculate the motion segment characteristics of each motion segment, including: mileage, driving time, average speed, SOC change, hydrogen pressure change, acceleration segment ratio, idling segment ratio, braking segment ratio, average acceleration / brake pedal stroke value, acceleration / brake pedal standard deviation, and total fuel system energy consumption; use the mileage, driving time, average speed, SOC change, and hydrogen pressure change in the motion segment characteristics to characterize the working condition characteristics corresponding to the motion segment, and use the acceleration segment ratio, idling segment ratio, braking segment ratio, average acceleration / brake pedal stroke value, and acceleration / brake pedal standard deviation to characterize the driving behavior characteristics of the driver in the motion segment;

[0030] S4, performing a K-Means clustering algorithm on the operating condition characteristics and driving behavior characteristics obtained in S3 with the total energy consumption of the fuel system as the target, and obtaining different typical operating condition types and driving behavior types corresponding to the motion segments; adding corresponding typical operating condition and driving behavior clustering labels to the motion segments, and calculating the average total fuel system energy consumption of different motion segments under the same operating condition and driving behavior type;

[0031] S5. Use the cluster labels obtained in S4, the average total fuel system energy consumption of different motion segments, and the characteristics of each motion segment in S3 to construct a test set and a training set, and train the random forest algorithm model to obtain a hydrogen fuel vehicle energy consumption estimation model; by executing the above steps on different target vehicle models, an energy consumption estimation model library containing multiple vehicle models is established; the corresponding data of the motion segment characteristics collected from a certain model of real vehicle during actual driving is input into the energy consumption estimation model library to estimate the real-time total fuel system energy consumption of vehicles of the same model.

[0032] In a preferred embodiment of the present invention, the screening and preprocessing of vehicle driving data in S1 includes capturing vehicle driving data for a certain period of time, filling missing value data, and correcting or deleting abnormal outlier data, wherein continuous data variables are filled with the average value of the two previous and subsequent valid values, and discrete and irregularly changing data are filled with data that is the same as the valid value of the previous frame.

[0033] In a preferred embodiment of the present invention, S2 combines the hydrogen fuel cell output current and output voltage, the DCDC converter on / off state, and the hydrogen pressure to determine whether the hydrogen fuel cell is consuming energy. Specifically, it is determined that energy is being consumed when the hydrogen fuel cell output current is higher than the rated value, the DCDC converter is in the on state, and the hydrogen fuel pressure shows a downward trend. The hydrogen fuel cell energy consumption W corresponding to each frame of raw data is specifically calculated in the following manner: W = ∫U f ×I f dt, where U f Indicates the instantaneous output voltage of the hydrogen fuel cell, I f represents the instantaneous output current of the hydrogen fuel cell, and t represents the sampling interval between two frames of data.

[0034] In a preferred embodiment of the present invention, each motion segment divided in S3 includes three randomly arranged stages of idling, acceleration and deceleration, and a uniform speed stage is added when the sampling frequency reaches a predetermined value; wherein the idling stage is defined as a data segment in which the vehicle is in a running state, the speed is 0 but the total current is not 0; the total fuel system energy consumption of the motion segment is determined by summing the fuel system energy consumption corresponding to each stage.

[0035] In a preferred embodiment of the present invention, in S4, the motion segments are divided into three categories: urban conditions, suburban conditions and high-speed conditions by executing a clustering algorithm, and the cluster center of each condition is determined by the average value of the characteristics of each motion segment. Figure 2 and Figure 3 2 shows the cluster distribution of three typical driving conditions and the corresponding cluster centers in an example based on the present invention. The driver behavior characteristics can be clustered according to various classification methods known to those skilled in the art, such as being divided into categories such as slow driving, smooth driving, and aggressive driving.

[0036] In a preferred embodiment of the present invention, when constructing the test set and the training set in S5, the features of each motion segment are first standardized, and then the principal component analysis (PCA) method is used to reduce the data dimension, and the data dimension is further reduced to 8 dimensions according to the principle that the contribution rate of the motion segment features to the energy consumption is greater than 95%;

[0037] The effect of training the random forest algorithm model is based on the model's estimated energy consumption for n motion clips. The actual energy consumption i The root mean square error (RMSE), mean square error (MSE), mean absolute percentage error (MAPE), and mean absolute error (MAE) are used to measure the difference. The specific calculation formulas are as follows:

[0038]

[0039]

[0040]

[0041]

[0042] Figure 4 Graph 1 shows various error performance indicators after the model is trained in an example of the present invention. It can be seen that the energy consumption estimation model obtained by the method of the present invention has a high accuracy.

[0043] In a preferred embodiment of the present invention, several algorithm models such as SVR, BP neural network, linear fitting, GBDT and XGBoost are used in S5 to replace the random forest algorithm model, and the hydrogen fuel vehicle energy consumption prediction model is obtained by training with a test set and a training set.

[0044] It should be understood that the size of the serial numbers of the steps in the embodiment of the present invention does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0045] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for estimating energy consumption of hydrogen fuel cell vehicles based on energy consumption characteristics and operating condition segmentation, characterized by: The specific steps include: S1. The big data platform obtains vehicle driving data collected and reported in real time by intelligent vehicle terminals of multiple hydrogen fuel cell vehicles of the target model; Screening and preprocessing the vehicle driving data to obtain raw data containing multiple parameter fields related to vehicle energy consumption; S2. Extract the output current and output voltage of the hydrogen fuel cell, the on / off state of the DCDC converter, and the hydrogen pressure from each frame of raw data, and combine these fields to determine whether the hydrogen fuel cell is consuming energy. If the judgment is "yes", calculate the energy consumption of the hydrogen fuel cell corresponding to each frame of raw data, and further extract the parameter fields reflecting the vehicle driving conditions such as the vehicle's factory year, vehicle operating status, charging status, time, instantaneous speed, cumulative mileage, SOC, instantaneous total current / voltage, and brake / accelerator pedal travel value from the raw data as feature items; If the judgment is "No", the extraction of the above parameter fields and subsequent steps will not be performed; S3. Use the continuous parameter fields extracted in S2 to divide the vehicle into multiple complete motion segments from idling to starting driving and then to parking, and calculate the motion segment characteristics of each motion segment, including: mileage, driving time, average speed, SOC change, hydrogen pressure change, acceleration segment ratio, idling segment ratio, braking segment ratio, average acceleration / brake pedal stroke value, acceleration / brake pedal standard deviation, and total fuel system energy consumption; use the mileage, driving time, average speed, SOC change, and hydrogen pressure change in the motion segment characteristics to characterize the working condition characteristics corresponding to the motion segment, and use the acceleration segment ratio, idling segment ratio, braking segment ratio, average acceleration / brake pedal stroke value, and acceleration / brake pedal standard deviation to characterize the driving behavior characteristics of the driver in the motion segment; S4, performing a K-Means clustering algorithm on the operating condition characteristics and driving behavior characteristics obtained in S3 with the total energy consumption of the fuel system as the target, and obtaining different typical operating condition types and driving behavior types corresponding to the motion segments; adding corresponding typical operating condition and driving behavior clustering labels to the motion segments, and calculating the average total fuel system energy consumption of different motion segments under the same operating condition and driving behavior type; S5. Use the cluster labels obtained in S4, the average total fuel system energy consumption of different motion segments, and the characteristics of each motion segment in S3 to construct a test set and a training set, and train the random forest algorithm model to obtain a hydrogen fuel vehicle energy consumption estimation model; by executing the above steps on different target vehicle models, an energy consumption estimation model library containing multiple vehicle models is established; the corresponding data of the motion segment characteristics collected from a certain model of real vehicle during actual driving is input into the energy consumption estimation model library to estimate the real-time total fuel system energy consumption of vehicles of the same model.

2. The method according to claim 1, characterized in that: The screening and preprocessing of vehicle driving data in S1 includes capturing vehicle driving data for a certain period of time, filling in missing value data, and correcting or deleting abnormal outlier data. For continuous data variables, the average value of the two previous and subsequent valid values ​​is used for filling, and for discrete and irregularly changing data, the data with the same valid value as the previous frame is used for filling.

3. The method according to claim 1, characterized in that: In S2, the hydrogen fuel cell output current and output voltage, the on / off state of the DCDC converter, and the hydrogen pressure are combined to determine whether the hydrogen fuel cell is consuming energy. Specifically, it is determined that the hydrogen fuel cell is consuming energy when the output current of the hydrogen fuel cell is higher than the rated value, the DCDC converter is in the on state, and the hydrogen fuel pressure shows a downward trend. The energy consumption W of the hydrogen fuel cell corresponding to each frame of raw data is calculated in the following way: W = ∫U f ×I f dt, where U f Indicates the instantaneous output voltage of the hydrogen fuel cell, I f represents the instantaneous output current of the hydrogen fuel cell, and t represents the sampling interval between two frames of data.

4. The method according to claim 1, characterized in that: Each motion segment divided in S3 contains three randomly arranged stages of idling, acceleration and deceleration. When the sampling frequency reaches a predetermined value, a uniform speed stage is added. The idling stage is defined as a data segment in which the vehicle is in operation, the speed is 0 but the total current is not 0. The total energy consumption of the fuel system of the motion segment is determined by the sum of the fuel system energy consumption corresponding to each stage.

5. The method according to claim 1, characterized in that: In S4, the motion segments are specifically divided into three categories: urban conditions, suburban conditions, and high-speed conditions by executing a clustering algorithm. The cluster center of each condition is determined by the average value of the characteristics of each motion segment.

6. The method according to claim 1, characterized in that: In S5, when constructing the test set and training set, the features of each motion segment are first standardized, and then the principal component analysis method is used to reduce the data dimension. According to the principle that the contribution rate of motion segment features to energy consumption is greater than 95%, the data dimension is further reduced to 8 dimensions; The effect of training the random forest algorithm model is based on the model's estimated energy consumption for n motion clips. The actual energy consumption i The root mean square error RMSE, mean square error MSE, mean absolute percentage error MAPE, and mean absolute error MAE are used to measure the difference. The specific calculation formulas are as follows:

7. The method according to claim 1, characterized in that: In S5, several algorithm models including SVR, BP neural network, linear fitting, GBDT and XGBoost are used to replace the random forest algorithm model, and the hydrogen fuel vehicle energy consumption estimation model is obtained by training with the test set and the training set.

Images