Hybrid off-road vehicle energy management method, apparatus, medium, and product

Abstract

The application discloses a hybrid off-road vehicle energy management method, device, medium and product, and relates to the field of vehicle energy management. The application classifies off-road working conditions based on test data by using a K-means clustering algorithm and relative entropy to obtain a classification result, simplifies variable off-road environments into a transfer probability matrix of electric power demand under different working conditions, greatly reduces the complexity of off-road working condition classification, and realizes an energy management strategy which is adaptively changed in real time according to a driving working condition. Moreover, the application determines an optimal solution corresponding to each type of working condition by using an improved expected SARSA algorithm based on the classification result, reduces online calculation time, and further solves the problems that an energy management strategy based on traditional reinforcement learning is slow in training speed and difficult to realize real-time control of a vehicle.

Description

Technical Field

[0001] This invention relates to the field of vehicle energy management, and in particular to a method, device, medium, and product for energy management of hybrid off-road vehicles. Background Technology

[0002] In recent years, energy management strategies, as a key technology for hybrid vehicles, have experienced rapid development. However, most existing energy management strategies focus on on-road vehicles, with limited research on energy management strategies for hybrid vehicles operating in off-road conditions. The challenge of energy management strategies for hybrid vehicles in off-road conditions lies in the complexity and difficulty of classification and identification based on specific operating conditions due to the large and frequent changes in road surface conditions, making it difficult to achieve effective energy management based on accurate prediction. Furthermore, at the methodological application level, reinforcement learning-based energy management strategies are widely used due to their advantages. However, the increased dimensionality of state and control variables in hybrid vehicles significantly slows down reinforcement learning training. Traditional reinforcement learning-based control strategies struggle to effectively optimize control in real-time for the current driving environment during vehicle operation. These challenges pose significant hurdles to existing technologies. Summary of the Invention

[0003] To address the aforementioned problems in the existing technology, this invention provides a method, device, medium, and product for energy management of hybrid off-road vehicles.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A method for energy management of a hybrid off-road vehicle, the method comprising:

[0006] The K-means clustering algorithm and relative entropy were used to classify off-road conditions based on experimental data to obtain classification results. The classification results include: feature parameters and transition probability matrices corresponding to the category centroids.

[0007] Based on the classification results, the optimal solution for each type of operating condition is determined using the improved expected SARSA algorithm; the optimal solution includes the optimal control law and the transition probability matrix of the power demand corresponding to the optimal control law.

[0008] Based on the optimal solution, an energy management strategy is generated for each type of operating condition.

[0009] Determine the vehicle's operating condition type based on its current driving status;

[0010] Based on the vehicle's operating conditions, the corresponding energy management strategy is selected to complete the online control of vehicle energy.

[0011] Optionally, the off-road conditions are classified based on the experimental data using the K-means clustering algorithm and relative entropy to obtain classification results, specifically including:

[0012] Randomly extract working condition segments from the test data;

[0013] Determine the feature parameters of each working condition segment and the corresponding transition probability matrix;

[0014] Conduct correlation analysis and select characteristic parameters related to power demand patterns under each operating condition;

[0015] The K-means clustering algorithm is used to classify all operating condition segments into multiple operating condition types based on the selected feature parameters related to power demand patterns.

[0016] The k-nearest neighbor algorithm is used to determine the point closest to the Euclidean center for each work condition type, and this point is used as the category center point.

[0017] Determine the relative entropy of the transition probability matrix of the category center point and the corresponding work condition segments of all points in its work condition type;

[0018] When the relative entropy is less than the relative entropy threshold, the corresponding feature parameters and transition probability matrix are output to obtain the classification result;

[0019] When the relative entropy is greater than or equal to the relative entropy threshold, the value of k is increased, and the process of using the K-means clustering algorithm to classify all operating condition segments into multiple operating condition types based on the selected feature parameters related to the power demand law is returned.

[0020] Optionally, based on the classification results, the optimal solution for each type of working condition is determined using an improved expected SARSA algorithm, specifically including:

[0021] Based on the goal of energy management, a multi-objective function optimization problem is constructed;

[0022] The state-value function of the improved expected SARSA algorithm is determined based on the SOC of the vehicle battery, the target electric power of the vehicle battery, and the target power of the vehicle engine generator set.

[0023] The action value function of the improved expected SARSA algorithm is determined based on the actual electrical power of the vehicle battery and the actual power of the vehicle engine generator set.

[0024] An improved expected reward function for the SARSA algorithm is constructed based on the aforementioned multi-objective function optimization problem;

[0025] The optimal action value function is iteratively updated using the transition probability matrix of power demand in each type of driving cycle.

[0026] The optimal solution for each type of working condition is obtained by iterative interaction between the state value function and the action value function.

[0027] Optionally, the multi-objective function optimization problem is:

[0028]

[0029] Where J is a multi-objective function, and SOC(t) is the battery reference charge at time t. This refers to the fuel consumption rate, where β1 is the first weighting coefficient, β2 is the second weighting coefficient, and SOC (State of Charge) is the fuel consumption rate. ref The battery reference state of charge is α, where α is the engine throttle position and n is the reference state of charge. e (t) is the engine speed at time t, n e It is the maximum engine speed, T e (t) is the engine torque at time t, T emin It is the minimum value of engine torque, f(n) e (t) is the engine torque value after power decomposition at time t, where T is the time period. g (t) is the generator torque at time t, T gmin It is the minimum value of the generator torque, T. gmax It is the maximum value of the generator torque, SOC. min It is the minimum reference charge of the battery, SOC. max It is the maximum value of the battery's reference charge.

[0030] Optionally, the vehicle's operating condition type can be determined based on the vehicle's current driving status, specifically including:

[0031] Record the vehicle's current power demand.

[0032] The current characteristic parameters and transition probability matrix are determined in real time based on the recorded power demand;

[0033] Based on the feature parameters and the transition probability matrix, the nearest working condition cluster center is selected using the standardized Euclidean distance.

[0034] The operating condition type corresponding to the current moment of a vehicle's journey is determined based on the operating condition cluster center.

[0035] Optionally, the formula for determining the transition probability matrix is:

[0036]

[0037] Where, p ij From power demand P i req To power demand The transition probabilities, i,j (i∈N, j∈N), are the transition probabilities from power demand P, respectively. i req To power demand index, P req (k) represents the power requirement for the kth stage, and N is the number of data points that meet the requirements.

[0038] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to implement the steps of the hybrid off-road vehicle energy management method described in any of the preceding claims.

[0039] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the hybrid off-road vehicle energy management method described above.

[0040] A computer program product includes a computer program that, when executed by a processor, implements the steps of the hybrid off-road vehicle energy management method described in any of the preceding claims.

[0041] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0042] This invention classifies off-road conditions based on experimental data using K-means clustering and relative entropy, simplifying the variable off-road environment into a transition probability matrix of electric power demand under different conditions. This significantly reduces the complexity of off-road condition classification, enabling energy management strategies that adapt to real-time changes in driving conditions. Furthermore, based on the classification results, this invention utilizes an improved expectation SARSA algorithm to determine the optimal solution for each type of condition, reducing online computation time and thus solving the problems of slow training speed and difficulty in achieving real-time vehicle control in traditional reinforcement learning-based energy management strategies. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart of the hybrid off-road vehicle energy management method provided in Embodiment 1 of the present invention;

[0045] Figure 2 This is an overall flowchart of the hybrid off-road vehicle energy management method provided in Embodiment 1 of the present invention;

[0046] Figure 3 This is a flowchart illustrating the working condition division provided in Embodiment 1 of the present invention;

[0047] Figure 4 This is a flowchart of the online working condition identification process provided in Embodiment 1 of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] The purpose of this invention is to provide a method, device, medium, and product for energy management of hybrid off-road vehicles, which aims to reduce the complexity of off-road condition classification so as to realize an energy management strategy that adapts to driving conditions in real time while reducing online calculation time.

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] Example 1

[0052] This embodiment provides an energy management method for a hybrid off-road vehicle, which mainly comprises two parts:

[0053] Part 1: The off-road conditions are classified by combining the K-means clustering algorithm and relative entropy, and the optimal solution for each type of condition is calculated using the improved expectation-based SARSA algorithm.

[0054] Part Two: Online identification of operating conditions and implementation of online control laws.

[0055] Based on this, such as Figure 1 and Figure 2 As shown, the hybrid off-road vehicle energy management method provided in this embodiment includes:

[0056] Step 100: Based on the experimental data, the K-means clustering algorithm and relative entropy are used to classify the off-road conditions to obtain the classification results. The classification results include: the feature parameters corresponding to the category centroids and the transition probability matrix.

[0057] In practical applications, this step is mainly used to classify off-road conditions, and its implementation process is as follows:

[0058] Correlation calculations are used to select characteristic parameters related to demand patterns under different operating conditions. The characteristic parameters used in this invention include, but are not limited to, average power, power standard deviation, power increase rate, power decrease rate, power stabilization rate, average power increase within 0.1s, average power decrease within 0.1s, average power increase within 0.5s, average power decrease within 0.5s, average power increase within 1s, average power decrease within 1s, percentage of power in the [0-200] interval, percentage of power in the (200-400] interval, percentage of power in the (400-600] interval, percentage of power in the (600-800] interval, and percentage of power greater than 800. The correlation coefficient R0 is... cor The formula for determining it is:

[0059]

[0060] Where, x i y i This represents two different characteristic parameters in different operating condition segments. This represents the average value of the characteristic parameter. n is the number of operating condition segments.

[0061] For example, 200 segments of driving conditions can be selected by random segmentation.

[0062] Through correlation analysis, eight parameters were selected as characteristic parameters: average power, power standard deviation, power flatness ratio, average power increase within 1 second, power

[0200] interval percentage, power (200400] interval percentage, power (400600] interval percentage, and power greater than 600 interval percentage.

[0063] To facilitate the calculation and execution of energy management strategies, the power demand is treated as a Markov stochastic process, and a first-order Markov chain is established (i.e., a state transition matrix is ​​constructed). Based on the time dimension, the data collected during driving is divided into multiple stages, and the power demand is further divided into discrete intervals from minimum to maximum. The resulting state transition probability matrix is ​​as follows:

[0064]

[0065] Where, p ij From power demand P i req To power demand The transition probabilities constitute the transition probability matrix. i,j (1≤i≤ζ, 1≤j≤ζ, i∈N, j∈N) represent the transition probabilities from the power demand P. i req To power demand The index of P. req(k) represents the power requirement for the kth stage, and N represents the number of data points that meet the requirements.

[0066] Specifically, after collecting the data, the following statistics were compiled:

[0067] 1) The power demand during stage K of the driving process is P. i req The number of data points.

[0068] 2) The power demand during stage K+1 of the driving process is The number of data points.

[0069] 3) Dividing the number of two data points gives the transition probability.

[0070] The preceding stages 1) and 2) and the discrete process are for better statistical data points.

[0071] Since the optimal solution to the optimal control problem is related to the power demand, and different transition probability matrices correspond to different optimal solutions, the operating conditions can be classified based on the relative entropy of the state transition probabilities.

[0072]

[0073] Where Φ and Φ′ are two transition probability matrices of size ζ×ζ. D KL (Φ * ||Φ) is the relative entropy (also known as KL divergence) of the transition probability matrices Φ and Φ′, used to measure the difference between two probability distributions. It has two properties: 1) D KL (Φ * ||Φ)≥0, i.e., nonnegativity, 2)D KL (Φ * ||Φ)≠D KL (Φ||Φ * This refers to asymmetry.

[0074] Based on the characteristic parameters of the operating conditions, the operating conditions are classified using the K-means clustering method and relative entropy based on state transition probabilities as classification criteria. The process is as follows: Figure 3As shown. For each category in the clustering results, the relative entropy of the working condition corresponding to the feature parameter of the current cluster center and the transition probability matrix corresponding to other working conditions in the current category is calculated. If it is within the set threshold range, it means that the clustering result meets expectations; otherwise, the value of k is increased, and the clustering is re-organized until the relative entropy is within the set threshold range. Since there are no pre-defined working conditions for the cluster center feature parameters, their corresponding transition probability matrices cannot be known. In this case, the k-nearest neighbor algorithm is used to select the working condition corresponding to the feature parameter closest to the current cluster center according to the standardized Euclidean distance, replacing the working condition corresponding to the cluster center for calculation.

[0075]

[0076] Where χ is the total number of feature parameters, x is the feature parameter matrix of the working condition block, y is the cluster center feature parameter matrix, s is the standard deviation matrix, i is the feature parameter index, M is the total number of working condition segments in this working condition category, m is the working condition segment index in this working condition category, and min indicates taking the minimum value. im It refers to the matrix of the i-th characteristic parameter in the m-th working condition block, s min The minimum value of the standard deviation matrix.

[0077] Step 101: Based on the classification results, determine the optimal solution for each operating condition using the improved expected SARSA algorithm. The optimal solution includes the optimal control law and the transition probability matrix of the power demand corresponding to the optimal control law.

[0078] In practical applications, this step is mainly for offline training, and its specific implementation process can be as follows:

[0079] An improved expected SARSA algorithm is used for offline training under off-road conditions. For example, the energy management problem in this embodiment aims to maintain the battery's stable SOC so that it is always in a working state and to improve the vehicle's fuel economy. The established multi-objective function optimization problem is as follows:

[0080]

[0081] in, This refers to fuel consumption rate, where β1 and β2 are weighting coefficients, with β1 being the first weighting coefficient and β2 being the second weighting coefficient. (SOC) ref The battery reference state of charge is α, where α is the engine throttle position and n is the reference state of charge. e (t) is the engine speed at time t, n e It is the maximum engine speed, T e (t) is the engine torque at time t, T emin It is the minimum value of engine torque, f(n) e(t) is the engine torque value after power decomposition at time t, where T is the time period. g (t) is the generator torque at time t, T gmin It is the minimum value of the generator torque, T. gmax It is the maximum value of the generator torque.

[0082] In the improved expectation-based SARSA algorithm, the state variable S t ∈S={SOC(t),P EG (t)|0.2≤SOC≤0.8,0≤P EG ≤800}, a set of action variables A t ∈A{P Bref (t),P EGref (t)|-150≤P Bref ≤150, 0≤P EGref ≤800} and a reward function R t ∈R. Where, P Bref and P EGref These represent the target electrical power of the battery and the generator set, respectively, in kW. B and P EG These represent the actual electrical power of the battery and the generator set, respectively, in kW. SOC refers to the battery's state of charge. The reward function R is defined as:

[0083]

[0084] Among them, R t Let be the reward function at time t.

[0085] Based on the above description, the improved SARSA algorithm has two main advantages in its implementation:

[0086] 1) During the training process, the power demand transition probability matrix in the driving cycle is used to iteratively update the optimal action value function. It is not necessary to know the entire driving cycle, which speeds up the training process and avoids the bias caused by overtraining.

[0087]

[0088] In the formula, Q(S) t A t ,P t The power requirement is P. t At that time, state S t Next, execute action A t The corresponding Q value. R t+1 It's an instant reward. When the power demand P t+1 When following the distribution Π, state S under all power demandst+1 The expected value of the minimum Q-value, where γ is the discount factor and a is the action. The power demand at the next moment is P. t+1 The probability, Q(S) t+1 Let (a, p) represent the probability that, after taking action a, the required power at the next moment will shift to P with probability p. t+1 The state transitions to S t+1 The Q value is given by A, where A is the set of action variables.

[0089] 2) By using the interactive iteration of the state value function and the action value function, the optimal control law corresponding to the optimal value function is obtained, which improves the convergence speed of the action value function during the training process.

[0090] V(S t ,P t )←V(S t ,P t )+α[R t+1 +γE[V(S t+1 ,P t+1 )|S t+1 ]-V(S t ,P t (8)Q(S) t A t ,P) t ←Q(S t A t ,P t )+α[R t+1 +γE[V(S t+1 ,P t+1 )|S t+1 ]-Q(S t A t ,P t (9)

[0091]

[0092] The steps of the improved expected SARSA algorithm are shown in Table 1. Where S, A, P, and S... - A - ,P - These represent the current state, action, and power demand at the previous moment, respectively.

[0093] Table 1. Steps based on the improved expected SARSA algorithm

[0094]

[0095] Step 102: Generate energy management strategies for each type of operating condition based on the optimal solution. This involves recording the optimal control law and the corresponding power demand transition probability matrix based on the training results, and outputting them as the online control law to the real-time control strategy.

[0096] Step 103: Determine the vehicle's operating condition type based on the vehicle's current driving status.

[0097] In practical applications, this step is mainly for identifying vehicle operating conditions. Specifically, after classifying the vehicle's operating condition segments, it is necessary to determine the operating condition type (i.e., driving mode) based on the vehicle's current driving status. The online operating condition identification process is as follows: Figure 4 As shown in the diagram, the process begins by recording historical data of the vehicle's current driving state, specifically its electric power demand. Then, based on this historical data (recorded electric power demand), the current feature parameters and transition probability matrix are calculated in real-time. Next, standardized Euclidean distance is used to select the nearest operating condition cluster center, thus identifying the current driving mode of the vehicle and applying the corresponding energy management strategy. Each driving condition corresponds to a specific energy management strategy. Because driving conditions are categorized, the current operating condition data is used to determine which condition applies before invoking the appropriate energy management strategy. Finally, the relative entropy between the current transition probability matrix and the transition probability matrix corresponding to the selected driving condition category center is calculated in real-time. If the entropy exceeds a threshold, the process returns to the previous step to reselect the driving condition.

[0098] Step 104: Select the corresponding energy management strategy based on the vehicle's operating condition to complete the online control of vehicle energy.

[0099] This step primarily aims to employ the optimal control law in real time. During implementation, based on the identification results of the aforementioned operating condition type (i.e., driving mode), the corresponding control law is used for online control, with its transition probability matrix serving as a benchmark. The optimal control command corresponding to the current electric power demand and SOC is obtained from the online control law, and then output to the coordinated control strategy to control the hybrid vehicle.

[0100] Based on the above description, the method provided in this embodiment uses an offline calculation method and a K-means clustering method based on relative entropy, taking experimental (operating condition) data as input, to obtain the feature parameters and their transition probability matrices corresponding to the class centroids. Based on the improved expectation SARSA method, and using all state variables of the system (State of Charge (SOC), engine-generator power P...),... EGS ) and their value ranges, and all control variables and their value ranges (engine-generator set target power P) EGSref Battery target power P brefUsing the input as input, the optimal control law is obtained. In practical applications, the current state transition probability matrix is ​​obtained through online collected driving data, enabling online identification of operating conditions and selection of a matching control law to achieve online real-time energy management.

[0101] In summary, compared with the prior art, the present invention has the following advantages:

[0102] 1) This invention establishes an off-road working condition identification model using the K-means clustering method based on relative entropy, which simplifies the variable off-road environment into a transition probability matrix of electric power demand under different working conditions, greatly reducing the complexity of off-road working condition classification, so as to realize an energy management strategy that can be adaptively changed in real time according to driving conditions.

[0103] 2) Because traditional reinforcement learning-based energy management strategies are slow to train and difficult to implement in real-time vehicle control, this invention proposes an online optimization control strategy based on an improved expectation SARSA algorithm. This strategy calculates the optimal control law for different operating conditions offline using the expectation SARSA algorithm, i.e., it uses the state-value function to interactively update the action-value function, and identifies the operating conditions in real time, applying the corresponding optimal control law to the hybrid vehicle. This reduces online computation time.

[0104] Example 2

[0105] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the hybrid off-road vehicle energy management method of Embodiment 1.

[0106] Example 3

[0107] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the hybrid off-road vehicle energy management method of Embodiment 1.

[0108] Example 4

[0109] A computer program product includes a computer program that, when executed by a processor, implements the steps of the hybrid off-road vehicle energy management method of Embodiment 1.

[0110] Example 5

[0111] A computer device, which may be a database, includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The database stores pending transactions. The I / O interfaces facilitate information exchange between the processor and external devices. The communication interface enables communication with external terminals via a network connection. When executed by the processor, the computer program implements the hybrid off-road vehicle energy management method of Embodiment 1.

[0112] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this invention are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0113] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided by this invention may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided by this invention may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0114] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Similar or identical parts between the various embodiments can be referred to mutually. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for energy management of a hybrid off-road vehicle, characterized in that, The method includes: The K-means clustering algorithm and relative entropy were used to classify off-road conditions based on experimental data to obtain classification results. The classification results include: feature parameters and transition probability matrices corresponding to the category centroids. Based on the classification results, the optimal solution for each type of operating condition is determined using the improved expected SARSA algorithm; the optimal solution includes the optimal control law and the transition probability matrix of the power demand corresponding to the optimal control law. Based on the optimal solution, an energy management strategy is generated for each type of operating condition. Determine the vehicle's operating condition type based on its current driving status; The appropriate energy management strategy is selected based on the vehicle's operating condition to complete the online control of vehicle energy; Based on the classification results, the optimal solution for each type of working condition is determined using the improved expected SARSA algorithm, including: Based on the goal of energy management, a multi-objective function optimization problem is constructed; the multi-objective function optimization problem is as follows: ； in, It is a multi-objective function. yes t Battery reference charge at any given time, It is the fuel consumption rate. It is the first weighting coefficient. It is the second weighting coefficient. For the battery reference state of charge, It's the engine throttle. It is the engine speed at time t. It is the maximum engine speed. It is the engine torque at time t. It is the minimum engine torque. It is the engine torque value after power decomposition at time t, where T is the time period. It is the generator torque at time t. It is the minimum value of the generator torque. It is the maximum value of the generator torque. SOC min It is the minimum reference charge of the battery. SOC max It is the maximum value of the battery's reference charge; The state-value function of the improved expected SARSA algorithm is determined based on the SOC of the vehicle battery, the target electric power of the vehicle battery, and the target power of the vehicle engine generator set. The action value function of the improved expected SARSA algorithm is determined based on the actual electrical power of the vehicle battery and the actual power of the vehicle engine generator set. An improved expected reward function for the SARSA algorithm is constructed based on the aforementioned multi-objective function optimization problem; The optimal action value function is updated iteratively using the transition probability matrix of power demand in each type of driving cycle. The optimal solution for each type of working condition is obtained by iterative interaction between the state value function and the action value function.

2. The energy management method for hybrid off-road vehicles according to claim 1, characterized in that, Based on experimental data, the K-means clustering algorithm and relative entropy were used to classify off-road conditions, resulting in the following classification results: Randomly extract working condition segments from the test data; Determine the feature parameters of each working condition segment and the corresponding transition probability matrix; Conduct correlation analysis and select characteristic parameters related to power demand patterns under each operating condition; The K-means clustering algorithm is used to classify all operating condition segments into multiple operating condition types based on the selected feature parameters related to power demand patterns. The k-nearest neighbor algorithm is used to determine the point closest to the Euclidean center for each work condition type, and this point is used as the category center point. Determine the relative entropy of the transition probability matrix of the category center point and the corresponding work condition segments of all points in its work condition type; When the relative entropy is less than the relative entropy threshold, the corresponding feature parameters and transition probability matrix are output to obtain the classification result; When the relative entropy is greater than or equal to the relative entropy threshold, the value of k is increased, and the process of using the K-means clustering algorithm to classify all operating condition segments into multiple operating condition types based on the selected feature parameters related to the power demand law is returned.

3. The energy management method for hybrid off-road vehicles according to claim 1, characterized in that, The vehicle's operating condition is determined based on its current driving status, specifically including: Record the vehicle's current power demand. The current characteristic parameters and transition probability matrix are determined in real time based on the recorded power demand; Based on the feature parameters and the transition probability matrix, the nearest working condition cluster center is selected using the standardized Euclidean distance. The operating condition type corresponding to the current moment of a vehicle's journey is determined based on the operating condition cluster center.

4. The energy management method for hybrid off-road vehicles according to claim 1, characterized in that, The formula for determining the transition probability matrix is: ； in, From power demand To power demand The transition probability, These are respectively based on power demand To power demand index, This is the power demand for the k-th stage. N It is the number of data points that meet the requirements.

5. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the hybrid off-road vehicle energy management method according to any one of claims 1-4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the hybrid off-road vehicle energy management method according to any one of claims 1-4.

7. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the hybrid off-road vehicle energy management method according to any one of claims 1-4.

Images