An end-to-end energy management control method fusing lidar and lstm-transformer speed prediction

Abstract

The present application relates to a kind of fusion laser radar and end-to-end energy management control method of LSTM-transformer speed prediction, belong to the field of hybrid electric vehicle energy management.The method, through laser radar, the speed prediction of future vehicle speed and demand power is obtained by combining with end-to-end LSTM-transformer model to the front obstacle is identified.Equivalent consumption minimization strategy (ECMS) is used to optimize the power distribution of fuel cell and battery, and a fuel cell power change penalty factor is introduced to suppress the performance degradation of fuel cell.In addition, a multi-objective optimization method is used to consider battery energy and fuel cell life, and the comprehensive optimization of energy management strategy is realized.The method effectively improves the efficiency of fuel cell system, prolongs the service life of the system, further reduces the total cost of fuel cell and battery, and significantly improves the energy-saving and emission-reducing performance of hybrid electric vehicle.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of hybrid electric vehicle energy management, and particularly relates to an end-to-end energy management control method combining laser radar and LSTM-transformer speed prediction. BACKGROUND

[0002] With the global concern for energy saving and emission reduction, fuel cell hybrid electric vehicles (FCHEVs) are receiving more and more attention due to their high efficiency and zero emission characteristics. FCHEVs use a combination of fuel cells and batteries, which can provide power while achieving efficient energy management to meet power demands under different working conditions. Effective energy management strategies (EMS) are crucial to improving the overall performance of FCHEVs. Existing energy management strategies can generally be divided into rule-based strategies, optimization strategies, and learning-based strategies. Among them, the control method based on real-time speed prediction is widely concerned due to its strong adaptability and good real-time performance.

[0003] Current speed prediction methods mainly focus on short-term prediction and can be divided into model-based methods and data-driven methods. Model-based methods usually rely on historical motion states of the vehicle for prediction, while data-driven methods use big data analysis to improve prediction accuracy. However, these methods often face several challenges. First, existing speed prediction algorithms mostly rely on the motion state of the ego vehicle, and when the external environment changes significantly, the prediction accuracy will decrease significantly. In addition, traditional methods often fail to effectively integrate surrounding environmental information such as the state of the preceding vehicle, pedestrian dynamics, etc., resulting in limited accuracy of speed prediction.

[0004] Laser radar (LiDAR) as a new sensing technology can provide high-precision environmental perception information. By integrating laser radar data with LSTM-transformer models, the surrounding environmental changes can be more accurately captured, thereby improving the accuracy of speed prediction. At the same time, LSTM-transformer-based models have the advantage of handling time-series data and can quickly respond to dynamically changing traffic scenarios.

[0005] Although existing research has attempted to combine different information sources for speed prediction, in dynamic traffic environments, existing methods still lack real-time performance and accuracy. In addition, current energy management strategies are often based on static power distribution models, lacking the ability to flexibly respond to instantaneous changes in environmental factors. Therefore, how to use laser radar data to improve the accuracy of speed prediction and effectively apply it to the energy management control of FCHEVs has become a problem to be solved. SUMMARY

[0006] The present application aims at the defects of the prior art, and provides an end-to-end energy management control method fusing laser radar and LSTM-transformer speed prediction, which can improve the collaborative working efficiency of fuel cells and batteries by acquiring and analyzing surrounding environment information in real time and dynamically adjusting energy management strategies.

[0007] To achieve the above-mentioned purpose, the technical scheme of the present application is as follows: an end-to-end energy management control method fusing laser radar and LSTM-transformer speed prediction, comprising:

[0008] The point cloud data of the surrounding environment of the vehicle is acquired online by the laser radar, and target tracking is performed after three-dimensional target detection and identification of the surrounding environment information of the vehicle;

[0009] The surrounding environment information of the vehicle after target tracking is input into the LSTM-transformer speed prediction model as real-time state variables and constraint variables, and the current predicted speed sequence of the vehicle is generated through feature extraction and time series prediction;

[0010] The current predicted speed sequence is used to obtain the current power as input, and the objective function of the equivalent consumption minimization strategy (ECMS) is constructed through equivalent hydrogen consumption, fuel cell degradation and power battery degradation for multi-objective optimization, so as to realize the optimal power distribution of the predicted speed sequence to the fuel cell and the lithium ion battery.

[0011] In an embodiment of the present application, the multi-objective optimization refers to considering the minimum battery degradation and equivalent hydrogen consumption, adjusting the equivalent factor, and obtaining the optimal power sequence.

[0012] In an embodiment of the present application, the point cloud data of the surrounding environment of the vehicle is acquired online by the laser radar, and target tracking is performed after three-dimensional target detection and identification of the surrounding environment information of the vehicle, which is implemented as follows:

[0013] Calibration: the external parameter matrix of the laser radar relative to the vehicle coordinate system is measured by using the laser range finder and the level;

[0014] Target vehicle data processing: the point cloud data of the surrounding environment of the vehicle acquired online by the laser radar is subjected to Kalman filtering denoising processing, and then the denoised point cloud data is processed by the PointPillar model for three-dimensional target detection, and the obtained target information is used for environment understanding and real-time tracking by a multi-target tracking algorithm;

[0015] Driving vehicle data acquisition: the current driving speed and acceleration of the vehicle are acquired through the CAN bus.

[0016] In an embodiment of the present application, the laser radar is installed at the center of the vehicle head.

[0017] In an embodiment of the present application, the target information is the target vehicle position P f , size D f , yaw angle and vehicle speed V p in front of the driving vehicle, and the pedestrian position P p and obstacle Po.

[0018] In an embodiment of the present application, the prediction method of the LSTM-transformer speed prediction model is: the sequence data is processed by LSTM, the input data X = [x1, x2,..., x T ], the LSTM generates the hidden state H = [h1, h2,..., h T ], and the LSTM updates and maintains the hidden state and memory cell state in turn through the gating mechanism of the input gate i, the forgetting gate f, the candidate memory C and the output gate o, and the core formula is:

[0019]

[0020] In the formula, 1≤t≤T, x t is the input of the current time step, h t-1 is the hidden state of the previous time step, C t-1 is the cell state of the previous time step, is the current cell state, σ is the activation function Sigmoid, tanh is the activation function, i t , f t and o t are the outputs of the corresponding gates, W i , W f and W o are the weights of the corresponding gates, b i , b f and b o are the biases of the corresponding gates, C t , W C and b C are the memory cell output, weight and bias, h t is the hidden state of the current time step.

[0021] The hidden state H generated by the LSTM is taken as the input of the transformer, and the current predicted speed sequence v pre = Transformer(H) is obtained.

[0022] In an embodiment of the present application, the LSTM-transformer speed prediction model is composed of three layers of LSTM and transformer. First, the target vehicle position P f , size D f , yaw angle and vehicle speed V p in front of the driving vehicle are input to lstm1, the input of lstm1 and the speed and acceleration of the target vehicle are input to lstm2, the output of lstm1, the output of lstm2 and static obstacles are input to lstm3, and the output of lstm1, lstm2 and lstm3 are input to the transformer, and finally the current predicted speed V pre is obtained.

[0023] In an embodiment of the present application, the functional relationship between the current predicted speed sequence and the current power is:

[0024]

[0025] where P pre is the current power, v pre is the current predicted speed; m is the mass of the driving vehicle; F f is the rolling resistance; g = 9.81 m·s -2 is the acceleration of gravity; f = 0.009 is the rolling resistance coefficient; α is the road slope angle (°); F a is the air resistance; C d = 0.312 is the air resistance coefficient; A = 2.06 m 2 is the windward area; ρ a = 1.2 N·s 2 ·m -4 is the air density; F i is the slope resistance; F j is the acceleration resistance; δ = 1.05 is the rotational mass conversion coefficient; dv pre / dt is the driving acceleration (m·s -2 ).

[0026] In an embodiment of the present application, the target function of the equivalent consumption minimization strategy (ECMS) constructed is

[0027]

[0028] where J ECMS represents the instantaneous hydrogen consumption of the fuel cell power system; is the instantaneous hydrogen consumption of the fuel cell; P B (t) is the lithium battery power; P FC (t) is the fuel cell power; KEF is the equivalent factor, and δ is the penalty factor of the fuel cell power variation, is the low heat value of hydrogen (kJ·g -1 );

[0029] The constraint condition is

[0030]

[0031] In the formula, P B (t) is the power on the direct current bus, and the subscripts max and min are the maximum and minimum values of the corresponding variables, respectively.

[0032] In an embodiment of the present application, the SOC variation rate is solved by SQP, and the objective function at the current iteration point u = [P B , P FC ] T is in the form of:

[0033]

[0034] In the formula, P is the gradient of the objective function at the current point, Δu is the update step, and Hk is an approximation of the second-order derivative of the objective function (usually obtained by fitting the Hessian matrix).

[0035] In an embodiment of the present application, based on the generated optimal SOC variation rate curve The target vehicle predicts the vehicle speed, adjusts the vehicle speed reference trajectory V ref (t), and obtains the optimal power distribution of P B (t) and P FC (t).

[0036] The present application also provides a computer readable storage medium having stored thereon computer program instructions capable of being executed by a processor, when the processor executes the computer program instructions, the method steps as described above can be implemented.

[0037] Compared with the prior art, the present application has the following beneficial effects: the end-to-end energy management control method for fusing laser radar and LSTM-transformer speed prediction proposed in the present application identifies the surrounding environment information online through the laser radar, outputs a speed prediction sequence, combines the equivalent fuel minimization strategy, realizes the optimal power distribution of the fuel cell and the lithium ion battery, significantly improves the fuel economy, hydrogen consumption and battery degradation rate of the automobile, prolongs the service life of the fuel cell and the battery, and further improves the economy and environmental protection performance of the hybrid electric vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1A flowchart of an end-to-end energy management control method fusing laser radar and LSTM-transformer speed prediction according to the present application. DETAILED DESCRIPTION

[0039] The technical solutions of the present application will be described in detail below with reference to the accompanying drawings.

[0040] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as would be commonly understood by one of ordinary skill in the art to which the present application belongs.

[0041] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and it should be further understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of the features, steps, operations, devices, components and / or combinations thereof.

[0042] The present application provides an end-to-end energy management control method fusing laser radar and LSTM-transformer speed prediction, comprising:

[0043] The laser radar is used to obtain point cloud data of the vehicle's surrounding environment online, and after three-dimensional target detection, the vehicle's surrounding environment information is identified and tracked;

[0044] The vehicle's surrounding environment information after target tracking is input into the LSTM-transformer speed prediction model as real-time state quantity and constraint quantity, and through feature extraction and time series prediction, the current predicted speed sequence of the vehicle is generated;

[0045] Through the current predicted speed sequence, the current power is obtained as input, and the objective function of the equivalent consumption minimization strategy (ECMS) is constructed through equivalent hydrogen consumption, fuel cell degradation and power battery degradation for multi-objective optimization, so as to realize the optimal power distribution of the predicted speed sequence to the fuel cell and the lithium ion battery.

[0046] The following is the specific implementation process of the present application.

[0047] As shown in Figure 1 An example of the present application provides an end-to-end energy management control method fusing laser radar and LSTM-transformer speed prediction, comprising the following steps:

[0048] Step S1, the point cloud data of the vehicle surrounding environment is obtained online by laser radar, and after three-dimensional target detection, the vehicle environment information is recognized for target tracking.

[0049] Step S2, the environment information after target tracking is input into the Transform+LSTM model as real-time state quantity and constraint quantity, and through feature extraction and time series prediction, the current predicted speed sequence of the vehicle is generated.

[0050] Step S3, the current predicted speed sequence is obtained as input, and the target function of ECMS is constructed through the equivalent factor to perform power distribution;

[0051] Step S4, the optimal power distribution of the speed prediction sequence to the fuel cell and lithium ion battery is realized.

[0052] Step S1 specifically includes the following:

[0053] Calibration: the laser range finder and the level are used to measure the external parameter matrix of the laser radar relative to the vehicle coordinate system, wherein the laser radar is installed at the center of the vehicle head, and is used to obtain the point cloud data.

[0054] Target vehicle data processing: the obtained point cloud data is subjected to Kalman filtering noise reduction processing, and then the noise-reduced point cloud data is processed through the PointPillar model, three-dimensional target detection is performed, the obtained target information is used for environment understanding, and real-time tracking is performed through a multi-target tracking algorithm.

[0055] Driving vehicle data acquisition: the current driving speed and acceleration of the vehicle are obtained through the CAN bus

[0056] The target information is the position P of the target vehicle in front of the driving vehicle f , the size D f , the yaw angle and the speed V p , and the positions P of pedestrians p and obstacles Po.

[0057] In step S2, the LSTM-transformer speed prediction model calculation, first, the sequence data is processed by LSTM, the input data X = [x1, x2,..., x T ], the hidden state H = [h1, h2,..., h T ] is generated by LSTM, and the hidden state and memory cell state are updated and maintained in turn by the gating mechanism of the input gate i, the forgetting gate f, the candidate memory C and the output gate o, and the core formula is:

[0058]

[0059] In the formula, 1≤t≤T, x t is the input of the current time step, h t-1 is the hidden state of the previous time step, C t-1 is the cell state of the previous time step, is the cell state of the current time, σ is the activation function Sigmoid, tanh is the activation function, i t , f t and o t are the outputs corresponding to the gates, W i , W f and W o are the weights corresponding to the gates, b i , b f and b o are the biases corresponding to the gates, C t , W C and b C are the memory cell output, weight and bias, h t is the hidden state of the current time step;

[0060] The hidden state H generated by the LSTM is taken as the input of the transformer to obtain the current predicted speed sequence v pre =Transformer(H).

[0061] In step S2, the LSTM-transformer model is specifically composed of three layers of LSTM and a transformer. First, the target vehicle position P f , size D f , yaw angle and speed V p in front of the driving vehicle are taken as the input of lstm1, the input of lstm1 and the speed and acceleration of the target vehicle are taken as the input of lstm2, the output of lstm1, lstm2 and static obstacles are taken as the input of lstm3, and the output of lstm1, lstm2 and lstm3 are taken as the input of the transformer, and finally the current predicted speed V pre is obtained.

[0062] In step S3, the functional relationship between the current predicted speed sequence and the current power is:

[0063]

[0064] In the formula, P pre is the current power, v pre is the current predicted speed (km·h-1), m is the mass of the driving vehicle, F f is the rolling resistance, and g=9.81m·s -2g = 9.81 m / s2is the gravitational acceleration; f = 0.009 is the rolling resistance coefficient; a is the road slope angle (°); F a air resistance; C d = 0.312 is the air resistance coefficient; A = 2.06 m 2 is the windward area; p a = 1.2 N-s 2 · m -4 is the air density; F i is the slope resistance; F j is the acceleration resistance; d = 1.05 is the rotational mass conversion coefficient; dv pre / dt is the driving acceleration (m-s -2 ).

[0065] In step S3, the objective function of the equivalent consumption minimization strategy (ECMS) constructed is

[0066]

[0067] In the formula, J ECMS represents the instantaneous hydrogen consumption of the fuel cell power system; is the instantaneous hydrogen consumption of the fuel cell; P B (t) is the lithium battery power; P FC (t) is the fuel cell power; K EF is the equivalent factor, d is the penalty factor of the fuel cell power change, is the low heat value of hydrogen (kJ-g -1 ).

[0068] The constraint condition is

[0069]

[0070] In the formula, P B (t) is the power on the direct current bus, and the subscripts max and min respectively represent the maximum and minimum values of the corresponding variable.

[0071] In step S3, the SOC change rate is solved by SQP, and the objective function at the current iteration point u = [P B , P FC ] T is in the form of:

[0072]

[0073] In the formula, is the gradient of the objective function at the current point, Au is the update step, and Hk is an approximation of the second derivative of the objective function (usually obtained by fitting the Hessian matrix).

[0074] In step S4, the optimal SOC rate curve generated in step S3 is used The target vehicle predicts a vehicle speed, adjusts the vehicle speed reference trajectory V ref (t), to obtain the optimal P B (t) and P FC (t) power distribution.

[0075] The application further provides a computer readable storage medium, which has stored computer program instructions capable of being executed by a processor, and when the processor executes the computer program instructions, the method steps as described above can be implemented.

[0076] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0077] The present application is described with reference to 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 flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The functions specified in one or more flows and / or blocks.

[0078] These computer program instructions can also be stored in a computer readable storage medium capable of guiding the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The functions specified in one or more flows and / or blocks.

[0079] These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable device to produce a computer implemented process, so that the instructions executed on the computer or other programmable device provide a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1one or more processes and / or blocks Figure 1 the steps of a function specified in one or more blocks.

[0080] The above descriptions are only preferred embodiments of the present application, not intended to limit the present application to other forms described. Any person skilled in the art can make changes or modifications to the above disclosed technical contents as equivalent embodiments. However, any simple modification, equivalent change and modification made to the above embodiments without departing from the technical solution of the present application, according to the technical essence of the present application, still belongs to the protection scope of the present application.

Claims

1. An end-to-end energy management control method fusing lidar and LSTM-transformer velocity prediction, characterized in that, Comprise: The point cloud data of the vehicle surrounding environment is obtained online by laser radar, and target tracking is performed after three-dimensional target detection and identification of the vehicle surrounding environment information; the specific implementation is as follows: Calibration: use laser range finder and level to measure the external parameter matrix of laser radar relative to vehicle coordinate system; Target vehicle data processing: Kalman filter noise reduction processing is performed on the point cloud data of the vehicle surrounding environment obtained online by laser radar, then the PointPillar model is used to process the noise-reduced point cloud data, three-dimensional target detection is performed, the obtained target information is used for environment understanding, and real-time tracking is performed through multi-target tracking algorithm; Driving vehicle data acquisition: the current driving speed and acceleration of the vehicle are obtained through CAN bus; The vehicle surrounding environment information after target tracking is input into the LSTM-transformer speed prediction model as real-time state quantity and constraint quantity, and through feature extraction and time series prediction, the current predicted speed sequence of the vehicle is generated; The prediction mode of the LSTM-transformer speed prediction model is: the sequence data is processed by LSTM, the input data X=[x1, x2,..., xT], the LSTM generates hidden state H=[h1, h2,..., h T ] and the core formula is: where 1≤t≤T, x t is the input of the current time step, h t-1 is the hidden state of the last time step, C t-1 is the cell state of the last time step, is the cell state of the current time step, σ is the activation function Sigmoid, tanh is the activation function, i t , f t and o t are the outputs of the corresponding gates, W i , W f and W o are the weights of the corresponding gates, b i , b f and b o are the biases of the corresponding gates, C t , W C and b C are the memory cell output, weight and bias, h t is the hidden state of the current time step; The LSTM-generated hidden state H is taken as the input of the transformer to obtain the current predicted speed sequence v pre = Transformer(H); Through the current predicted speed sequence, the current power is obtained as input, and the objective function of the equivalent consumption minimization strategy ECMS is constructed through equivalent hydrogen consumption, fuel cell degradation and power battery degradation for multi-objective optimization, so as to realize the optimal power distribution of the predicted speed sequence to the fuel cell and lithium ion battery.

2. The end-to-end energy management control method of claim 1, wherein Multi-objective optimization refers to considering the minimum of battery degradation and equivalent hydrogen consumption, adjusting the equivalent factor, and obtaining the optimal power sequence.

3. The end-to-end energy management control method of claim 1, wherein, The laser radar is installed at the center of the vehicle head.

4. The end-to-end energy management control method of claim 1, wherein, Target information is a target vehicle position P in front of a traveling vehicle f , a size D f , a yaw angle , and a vehicle speed V p , and a pedestrian position P p , an obstacle Po.

5. The end-to-end energy management control method of claim 1, wherein, The LSTM-transformer speed prediction model is composed of three layers of LSTM and transformer, first taking the target vehicle position P f , size D f , yaw angle and vehicle speed V p in front of the driving vehicle as the input of lstm1, taking the input of lstm1 and the speed and acceleration of the target vehicle as the input of lstm2, taking the output of lstm1, the output of lstm2 and static obstacles as the input of lstm3, and then taking the output of lstm1, lstm2 and lstm3 as the input of transformer, and finally obtaining the current predicted speed V pre .

6. The end-to-end energy management control method of fusing lidar with LSTM- transformer velocity prediction according to claim 1, wherein, The functional relationship between the current predicted speed sequence and the current power is: where P pre is the current time power, v pre is the current predicted speed; m is the mass of the vehicle in motion; F f is the rolling resistance; g is the gravitational acceleration; f is the rolling resistance coefficient; a is the road slope angle; F a is the air resistance; C d is the air resistance coefficient; A is the windward area; p a is the air density; F i is the slope resistance; F j is the acceleration resistance; d is the rotational mass conversion coefficient; dv pre / dt is the travel acceleration.

7. The end-to-end energy management control method of claim 1, wherein, The objective function of the equivalent consumption minimization strategy ECMS constructed is where J ECMS represents the instantaneous hydrogen consumption of the fuel cell power system; is the instantaneous hydrogen consumption of the fuel cell; P B (t) is the lithium battery power; P FC (t) is the fuel cell power; K EF is the equivalent factor, and δ is the penalty factor for the change in fuel cell power; is the low heating value of hydrogen; The constraint condition is where P B (t) is the power on the DC bus, the subscripts max and min are the maximum and minimum values of the corresponding variable, respectively.

8. A computer readable storage medium having stored thereon computer program instructions capable of being executed by a processor, when the processor executes the computer program instructions, the steps of the end-to-end energy management control method fusing laser radar and LSTM-transformer speed prediction according to any one of claims 1-7 can be realized.

Images