Electric drive commercial vehicle braking energy recovery method considering braking intention

Abstract

The invention discloses an electrically-driven commercial vehicle braking energy recovery method considering a braking intention, which is characterized in that a driver braking intention identification process based on a hidden Markov-interval type-2 fuzzy neural network can effectively process uncertain information in a braking process of a driver and a braking system; the driver braking intention recognition precision is obviously improved; according to the braking energy recovery process based on the driver braking intention, the driver braking intention is fully considered for braking torque distribution, meanwhile, the advantage of rapid response of a regenerative braking system is utilized, the problem of slow response in the commercial vehicle friction braking process is solved, the braking safety is fully guaranteed, and the braking comfort is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric drive commercial vehicles, and specifically relates to a braking energy recovery method for an electric drive commercial vehicle taking braking intention into consideration. Background Art

[0002] As the core component of the new energy system, the electric drive axle of commercial vehicles is a key technology to promote the technical upgrading and market development of new energy commercial vehicles. As an effective means to increase the mileage of electric vehicles, the research on the composite braking system of electric commercial vehicles is of great significance to the improvement of the electric drive axle technology of commercial vehicles.

[0003] The research on existing commercial vehicle braking energy recovery technology focuses on the coordinated distribution of braking torque. In Chinese patent application number CN201811570181.6, a braking energy recovery system and control method for new energy commercial vehicles are proposed. Through a novel structural scheme, the energy recovery system can be compatible with the ABS system without changing the braking system; in Chinese patent application number CN201910765142.X, a braking energy recovery system and method for electric commercial vehicles are proposed, and the braking controller formulates a braking strategy according to the vehicle state and brake pedal information. However, the two electric commercial vehicle braking energy recovery methods in the above patent applications do not consider the driver's braking intention and the response differences between the actuators. The vehicle braking process is a dynamic and time-varying nonlinear process. The existing braking intention recognition methods, such as Markov chain and fuzzy theory, are based on the assumptions of fixed state change probability, correct rules, and independent distribution. The above methods are insufficient in recognition accuracy and real-time performance, and the ability to handle uncertain information in the process of driver braking intention recognition needs to be improved.

[0004] The commercial vehicle chassis-by-wire technology has certain advantages in vehicle handling performance and safety, and the use of regenerative braking systems can effectively improve the endurance of commercial vehicles. However, the driver's braking intention has a great influence on the braking torque distribution. How to accurately identify the driver's braking intention and consider the driver's braking intention and the working characteristics of the actuator to distribute the braking torque has a great room for improvement in the braking energy recovery and braking comfort of commercial vehicles. Summary of the invention

[0005] In view of the above problems, the main purpose of the present invention is to design a braking energy recovery method for electric commercial vehicles that takes braking intention into consideration, so as to solve the problem that the existing braking energy recovery method for electric commercial vehicles does not fully consider the difference between the driver's braking intention and the actuator response, resulting in low braking energy recovery efficiency and reduced braking comfort and safety performance.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A braking energy recovery method for an electric-driven commercial vehicle considering braking intention, the method is an energy recovery method for a driver's braking intention based on a hidden Markov interval type II fuzzy neural network, comprising: a braking intention recognition process of a hidden Markov interval type II fuzzy neural network, and a braking energy recovery process based on the driver's braking intention;

[0008] The method comprises the following steps:

[0009] Step 1: Establish a hidden Markov model of the driver's braking intention, optimize the parameters of the hidden Markov model, and obtain the maximum likelihood estimation array based on the collected brake pedal data;

[0010] Step 2: Based on the maximum likelihood estimation array, after being processed as input, an interval type-II fuzzy neural braking intention recognition network is established, the output of the interval type-II fuzzy neural braking intention recognition network is the braking intention, and the input of the interval type-II fuzzy neural braking intention recognition network is fuzzified;

[0011] Step 3: Based on the fuzzification of the input variables in step 2, perform fuzzy reasoning of the interval type-2 fuzzy neural braking intention recognition network;

[0012] Step 4: interval type-2 fuzzy neural braking intention recognition network fuzzy reduction;

[0013] Step 5: Interval type-2 fuzzy neural braking intention recognition network output;

[0014] Step 6: Identify the driver's braking intention based on the hidden Markov interval type-2 fuzzy neural network and judge the driver's braking intention, which includes light braking, moderate braking, and emergency braking;

[0015] Determine whether the driver's braking intention is slight braking, if so, proceed to step 7, if not, further determine whether the driver's braking intention is moderate braking, if so, proceed to step 8, if not, it is emergency braking, proceed to step 9;

[0016] Step 7: When the driver's braking intention is slight braking, the maximum regenerative braking torque of the motor carried by the commercial vehicle electric drive axle is calculated, and after obtaining the regenerative braking torque and the friction braking torque, step 10 is entered;

[0017] Step 8: When the driver's braking intention is moderate braking, the maximum regenerative braking torque and rear wheel braking torque of the motor carried by the commercial vehicle electric drive axle are calculated, and after obtaining the regenerative braking torque and the friction braking torque, step 10 is entered;

[0018] Step 9: When the driver's braking intention is emergency braking, the maximum regenerative braking torque and the anti-lock braking demand braking torque of the motor carried by the commercial vehicle electric drive axle are calculated, and after the regenerative braking torque and the friction braking torque are obtained, step 10 is entered;

[0019] Step 10: Using the obtained friction braking torque to compensate the braking torque by using the regenerative braking system;

[0020] Step 11: The vehicle controller controls the regenerative braking system and the friction braking system to implement corresponding braking torques according to the magnitudes of the regenerative braking torque and the friction braking torque until the vehicle speed is reduced to zero.

[0021] As a further description of the present invention, in step 1, establishing a hidden Markov model of the driver's braking intention includes the following steps:

[0022] S11: According to the driver's braking operation and the vehicle system status, the input of the hidden Markov model of the driver's braking intention is determined, and its observation sequence vector is:

[0023]

[0024] Among them, O t is the hidden Markov model observation sequence vector, d(t) represents the brake pedal displacement, represents the brake pedal speed, t is the time;

[0025] S12: Based on the pedal displacement d(t) and brake pedal speed signal collected by the brake pedal sensor The Baum-Welch algorithm is used to identify the parameters of the hidden Markov model of the driver's braking intention. The parameters of the hidden Markov model of the driver's braking intention are:

[0026] λ=(π,A,B);

[0027] Among them, λ is the hidden Markov model of the driver's braking intention, π is the initial probability vector, A is the transition probability matrix, and B is the observation probability density function;

[0028] The expression of the observation probability density function is:

[0029]

[0030] Among them, N(O,μ ij ,σ ij ) represents the q-dimensional Gaussian density function of state p, ω pq represents the weight coefficient of the qth mixing function under state p, μ pq represents the vector mean of the qth mixed Gaussian function under state p, and the covariance matrix of the σ mixing function;

[0031] S13: If ε t (p,q) is the probability of the qth mixed Gaussian function of the state p observation sequence at time t, then the Markov chain probability of state p at time t and state q at time t+1 is:

[0032]

[0033] Among them, P[O,μ pq ,σ pq ] is the q-th dimension Gaussian density function in state p, α t (p) is the forward probability of state p, β t (p) is the backward probability of state p, σ pq is the covariance matrix of the qth mixing function under state p;

[0034] After optimizing the parameters of the hidden Markov model, based on the collected brake pedal data, the matching degree with the multidimensional hidden Markov model is calculated using a forward-backward algorithm to obtain a maximum likelihood estimation array.

[0035] As a further description of the present invention, in step 2, the process of establishing the interval type II fuzzy neural braking intention recognition network includes the following steps:

[0036] S21: Determine the input and output of the interval type II fuzzy neural braking intention recognition network;

[0037] The maximum likelihood estimation array is normalized to obtain the Markov model state cumulative probability vector under three braking intentions, and it is used as the input x i ;

[0038]

[0039] Where i = 1, 2, 3; 1 represents light braking, 2 represents moderate braking, and 3 represents emergency braking; ω0 is the initial weight, w k , σ k are the center and width of the radial basis function of the kth neuron respectively;

[0040] S22: Interval type-2 fuzzy neural braking intention recognition network input fuzzification;

[0041] The interval type-2 fuzzy set of the input layer neuron x1 is Type II fuzzy sets of input layer neuron x2 Type II fuzzy sets of input layer neuron x3

[0042] Among them, NB means negative big, NS means negative small, ZE means zero, PS means positive small, and PB means positive big;

[0043] Therefore, the number of neurons in the hidden layer is 15, and the activation function of each neuron, that is, the membership function, is:

[0044]

[0045] in, is the upper bound of the output value of the j-th membership function layer neuron, u ij (x i ) is the lower bound of the output value of the j-th membership function layer neuron, c ij is the lower bound of the central value of the membership function of the neuron in the jth membership function layer, is the upper bound of the central value of the membership function of the neuron in the jth membership function layer, σ ij is the membership function width of the neuron in the jth membership function layer, j = 1, 2, ..., 15.

[0046] As a further description of the present invention, in step 3, the interval type II fuzzy neural braking intention recognition network fuzzy reasoning includes the following:

[0047] The input variables of the interval type II fuzzy neural braking intention recognition network are fuzzified by the membership function to obtain the membership interval of the variable. Each node in the incentive layer represents a fuzzy rule, and the activation interval of each rule is F j (x1,x2,x3), the expression is:

[0048]

[0049] Among them, f j is the lower bound of the activation of the jth rule, is the upper bound of the activation degree of the j-th rule.

[0050] As a further description of the present invention, in step 4, the interval type II fuzzy neural braking intention recognition network fuzzy reduction includes the following:

[0051] The reduction method uses the Enhanced Karnic-Mendel (EKM) algorithm, and the output expression of this layer is:

[0052]

[0053] Among them, J is the number of fuzzy rules, y iL The left endpoint of the network drop output, y iR is the right endpoint of the network degradation output, L is the left turning point of the EKM algorithm, and R is the right turning point of the EKM algorithm.

[0054] As a further description of the present invention, in step 5, the interval type II fuzzy neural braking intention recognition network output includes the following:

[0055] The fifth layer of the interval type-2 fuzzy neural braking intention recognition network is the output layer, whose nodes represent the output of the entire network. The output expression of the network is:

[0056]

[0057] The error back propagation algorithm is used to train the neural network using the gradient descent method, and the performance indicators are defined as:

[0058]

[0059] The parameters that need to be learned in the above network include the uncertainty center c of the type II Gaussian membership function j and Rule post-condition parameters w i and w i , the parameter correction formula is:

[0060]

[0061]

[0062] Among them, τ w Rule post-condition parameter w j and The learning rate, τ c is the uncertain center c j and The learning rate.

[0063] As a further description of the present invention, in step 7, the maximum regenerative braking torque T of the motor carried by the commercial vehicle electric drive axle is calculated. r_max , including the following steps:

[0064] S71: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the required braking torque T? b_i ;

[0065] If so, the vehicle braking torque T b for:

[0066] T b =T r ;

[0067] If not, the vehicle braking torque T b for:

[0068]

[0069] Among them, T b_f is the front axle braking torque, T b_r is the rear axle braking torque, T r is the regenerative braking torque, T m is the friction braking torque;

[0070] S72: Obtaining the regenerative braking torque T r With friction braking torque T m Then go to step 10.

[0071] As a further description of the present invention, in step 8, the maximum regenerative braking torque T of the motor carried by the commercial vehicle electric drive axle is calculated. r_max and rear wheel braking force T b_r , including the following steps:

[0072] S81: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the required braking torque T? b_r ;

[0073] If so, the vehicle braking torque T b for:

[0074]

[0075] If not, the vehicle braking torque T b for:

[0076]

[0077] Among them, T b_f is the front axle braking torque, T b_r is the rear axle braking torque, T r is the regenerative braking torque, T m is the friction braking torque;

[0078] S82: Get the regenerative braking torque T r With friction braking torque T m Then go to step 10.

[0079] As a further description of the present invention, in step 9, the maximum regenerative braking torque T of the motor mounted on the electric drive axle of the commercial vehicle r_max and the anti-lock braking demand torque T b_λ , including the following steps:

[0080] S91: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the anti-lock braking torque T? b_λ ;

[0081] If so, the vehicle braking torque Tb for:

[0082] T b =T m +T r ;

[0083] If not, the vehicle braking torque T b for:

[0084] T b =T r ;

[0085] Among them, T r is the regenerative braking torque, T m is the friction braking torque;

[0086] S92: Get the regenerative braking torque T r With friction braking torque T m Then go to step 10.

[0087] As a further description of the present invention, in step 10, the regenerative braking compensation torque is expressed as:

[0088]

[0089] Among them, T r is the regenerative braking torque, T m is the friction braking torque, is the actual friction braking torque.

[0090] Compared with the prior art, the technical effects of the present invention are:

[0091] The present invention provides a braking energy recovery method for an electric-driven commercial vehicle taking braking intention into consideration, wherein a driver braking intention recognition process based on a hidden Markov interval type-2 fuzzy neural network can effectively process uncertain information between the driver and the braking system during the braking process, and significantly improve the recognition accuracy of the driver's braking intention; a braking energy recovery process based on the driver's braking intention fully considers the driver's braking intention to distribute the braking torque, and at the same time utilizes the advantage of the rapid response of the regenerative braking system to compensate for the slow response problem of the friction braking process of the commercial vehicle, thereby fully ensuring braking safety and improving braking comfort. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 It is a schematic diagram of the driver's braking intention recognition process based on the hidden Markov interval type-2 fuzzy neural network of the present invention;

[0093] Figure 2 The figure is a flow chart of the braking energy recovery process based on the driver's braking intention according to the present invention. DETAILED DESCRIPTION

[0094] The present invention is described in detail below in conjunction with the accompanying drawings:

[0095] In one embodiment of the present invention, a braking energy recovery method for an electric-driven commercial vehicle taking braking intention into consideration is disclosed, such as Figure 1-2 As shown, the method is an energy recovery method of the driver's braking intention based on a hidden Markov interval type II fuzzy neural network, including: a braking intention recognition process of a hidden Markov interval type II fuzzy neural network, and a braking energy recovery process based on the driver's braking intention.

[0096] The method comprises the following steps:

[0097] Step 1: Establish a hidden Markov model of the driver's braking intention, optimize the parameters of the hidden Markov model, and obtain the maximum likelihood estimation array based on the collected brake pedal data;

[0098] Step 2: Based on the maximum likelihood estimation array, after being processed as input, an interval type-II fuzzy neural braking intention recognition network is established, the output of the interval type-II fuzzy neural braking intention recognition network is the braking intention, and the input of the interval type-II fuzzy neural braking intention recognition network is fuzzified;

[0099] Step 3: Based on the fuzzification of the input variables in step 2, perform fuzzy reasoning of the interval type-2 fuzzy neural braking intention recognition network;

[0100] Step 4: interval type-2 fuzzy neural braking intention recognition network fuzzy reduction;

[0101] Step 5: Interval type-2 fuzzy neural braking intention recognition network output;

[0102] Step 6: Identify the driver's braking intention based on the hidden Markov interval type-2 fuzzy neural network and judge the driver's braking intention, which includes light braking, moderate braking, and emergency braking;

[0103] Determine whether the driver's braking intention is slight braking, if so, proceed to step 7, if not, further determine whether the driver's braking intention is moderate braking, if so, proceed to step 8, if not, it is emergency braking, proceed to step 9;

[0104] Step 7: When the driver's braking intention is slight braking, the maximum regenerative braking torque of the motor carried by the commercial vehicle electric drive axle is calculated, and after obtaining the regenerative braking torque and the friction braking torque, step 10 is entered;

[0105] Step 8: When the driver's braking intention is moderate braking, the maximum regenerative braking torque and rear wheel braking torque of the motor carried by the commercial vehicle electric drive axle are calculated, and after obtaining the regenerative braking torque and the friction braking torque, step 10 is entered;

[0106] Step 9: When the driver's braking intention is emergency braking, the maximum regenerative braking torque and the anti-lock braking demand braking torque of the motor carried by the commercial vehicle electric drive axle are calculated, and after the regenerative braking torque and the friction braking torque are obtained, step 10 is entered;

[0107] Step 10: Using the obtained friction braking torque to compensate the braking torque by using the regenerative braking system;

[0108] Step 11: The vehicle controller controls the regenerative braking system and the friction braking system to implement corresponding braking torques according to the magnitudes of the regenerative braking torque and the friction braking torque until the vehicle speed is reduced to zero.

[0109] More specifically, this embodiment describes the above steps in detail as follows:

[0110] In this embodiment, in the above step 1, establishing a hidden Markov model of the driver's braking intention includes the following steps:

[0111] S11: According to the driver's braking operation and the vehicle system status, the input of the hidden Markov model of the driver's braking intention is determined, and its observation sequence vector is:

[0112]

[0113] Among them, O t is the hidden Markov model observation sequence vector, d(t) represents the brake pedal displacement, represents the brake pedal speed, t is the time;

[0114] S12: Based on the pedal displacement d(t) and brake pedal speed signal collected by the brake pedal sensor The Baum-Welch algorithm is used to identify the parameters of the hidden Markov model of the driver's braking intention. The parameters of the hidden Markov model of the driver's braking intention are:

[0115] λ=(π,A,B);

[0116] Among them, λ is the hidden Markov model of the driver's braking intention, π is the initial probability vector, A is the transition probability matrix, and B is the observation probability density function;

[0117] The expression of the observation probability density function is:

[0118]

[0119] Among them, N(O,μ ij ,σ ij ) represents the q-dimensional Gaussian density function of state p, ω pq represents the weight coefficient of the qth mixing function under state p, μ pq represents the vector mean of the qth mixed Gaussian function under state p, and the covariance matrix of the σ mixing function;

[0120] S13: If ε t (p,q) is the probability of the qth mixed Gaussian function of the state p observation sequence at time t, then the Markov chain probability of state p at time t and state q at time t+1 is:

[0121]

[0122] Among them, P[O,μ pq ,σ pq ] is the q-th dimension Gaussian density function in state p, α t (p) is the forward probability of state p, β t (p) is the backward probability of state p, σ pq is the covariance matrix of the qth mixing function under state p;

[0123] After optimizing the parameters of the hidden Markov model, based on the collected brake pedal data, the matching degree with the multidimensional hidden Markov model is calculated using a forward-backward algorithm to obtain a maximum likelihood estimation array.

[0124] In this embodiment, in the above step 2, the process of establishing the interval type II fuzzy neural braking intention recognition network includes the following steps:

[0125] S21: Determine the input and output of the interval type II fuzzy neural braking intention recognition network;

[0126] The maximum likelihood estimation array is normalized to obtain the Markov model state cumulative probability vector under three braking intentions, and it is used as the input x i ;

[0127]

[0128] Where i = 1, 2, 3; 1 represents light braking, 2 represents moderate braking, and 3 represents emergency braking; ω0 is the initial weight, w k , σ k are the center and width of the radial basis function of the kth neuron respectively;

[0129] S22: Interval type-2 fuzzy neural braking intention recognition network input fuzzification;

[0130] The interval type-2 fuzzy set of the input layer neuron x1 is Type II fuzzy set of input layer neuron x2 Type II fuzzy sets of input layer neuron x3

[0131] Among them, NB means negative big, NS means negative small, ZE means zero, PS means positive small, and PB means positive big; therefore, the number of neurons in the hidden layer is 15, and the activation function of each neuron, that is, the membership function, is:

[0132]

[0133] in, is the upper bound of the output value of the j-th membership function layer neuron, u ij (x i ) is the lower bound of the output value of the j-th membership function layer neuron, c ij is the lower bound of the central value of the membership function of the neuron in the jth membership function layer, is the upper bound of the central value of the membership function of the neuron in the jth membership function layer, σ ij is the membership function width of the neuron in the jth membership function layer, j = 1, 2, ..., 15.

[0134] In this embodiment, in the above step 3, the interval type II fuzzy neural braking intention recognition network fuzzy reasoning includes the following:

[0135] The input variables of the interval type II fuzzy neural braking intention recognition network are fuzzified by the membership function to obtain the membership interval of the variable. Each node in the incentive layer represents a fuzzy rule, and the activation interval of each rule is F j (x1,x2,x3), the expression is:

[0136]

[0137] Among them, f j is the lower bound of the activation of the jth rule, is the upper bound of the activation degree of the j-th rule.

[0138] In this embodiment, in the above step 4, the interval type II fuzzy neural braking intention recognition network fuzzy reduction includes the following:

[0139] The reduction method uses the Enhanced Karnic-Mendel (EKM) algorithm, and the output expression of this layer is:

[0140]

[0141] Among them, J is the number of fuzzy rules, y iL The left endpoint of the network drop output, yiR is the right endpoint of the network degradation output, L is the left turning point of the EKM algorithm, and R is the right turning point of the EKM algorithm.

[0142] In this embodiment, in the above step 5, the output of the interval type II fuzzy neural braking intention recognition network includes the following:

[0143] The fifth layer of the interval type-2 fuzzy neural braking intention recognition network is the output layer, whose nodes represent the output of the entire network. The output expression of the network is:

[0144]

[0145] The Error Back Propagation Algorithm is used to train the neural network using the gradient descent method, and the performance indicators are defined as:

[0146]

[0147] The parameters that need to be learned in the above network include the uncertainty center c of the type II Gaussian membership function j and Rule consequent parameter w i and The parameter correction formula is:

[0148]

[0149]

[0150] Among them, τ w Rule post-condition parameter w j and The learning rate, τ c is the uncertain center c j and The learning rate.

[0151] Among them, due to the problem of left and right turning points in the EKM reduction algorithm, the parameters should be learned and adjusted according to the turning points in the function derivation process, and the adjustment formula (where the uncertain center c of the second-type Gaussian membership function is j and The parameter update formula only lists the first neuron, and the parameter update formulas for the remaining neurons are similar) as follows:

[0152]

[0153] Traditional type I fuzzy logic has poor ability to resist external interference. In the process of driver braking intention recognition, interval type II fuzzy neural network is used to give full play to the ability of type II fuzzy system to process uncertain information and resist interference. By enhancing the description of inter-individual and intra-individual uncertainty, the anti-interference ability of fuzzy logic algorithm is enhanced, and the accuracy of driver braking intention recognition is improved.

[0154] In this embodiment, in the above step 6, different steps are entered based on different braking intentions of the driver.

[0155] Specifically, in this embodiment, in the above step 7, when the braking intention is slight braking, the maximum regenerative braking torque T of the motor carried by the electric drive axle of the commercial vehicle is calculated. r_max , including the following steps:

[0156] S71: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the required braking torque T? b_i ;

[0157] If so, the vehicle braking torque T b for:

[0158] T b =T r ;

[0159] If not, the vehicle braking torque T b for:

[0160]

[0161] Among them, T b_f is the front axle braking torque, T b_r is the rear axle braking torque, T r is the regenerative braking torque, T m is the friction braking torque;

[0162] S72: Obtaining the regenerative braking torque T r With friction braking torque T m Then go to step 10.

[0163] In this embodiment, in the above step 8, when the braking intention is moderate braking, the maximum regenerative braking torque T of the motor carried by the commercial vehicle electric drive axle is calculated. r_max and rear wheel braking force T b_r , including the following steps:

[0164] S81: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the required braking torque T? b_r ;

[0165] If so, the vehicle braking torque T b for:

[0166]

[0167] If not, the vehicle braking torque T b for:

[0168]

[0169] Among them, T b_f is the front axle braking torque, T b_r is the rear axle braking torque, T r is the regenerative braking torque, T m is the friction braking torque;

[0170] S82: Get the regenerative braking torque T r With friction braking torque T m Then go to step 10.

[0171] In this embodiment, in the above step 9, when the braking intention is emergency braking, the maximum regenerative braking torque T of the motor mounted on the electric drive axle of the commercial vehicle is r_max and the anti-lock braking demand torque T b_λ , including the following steps:

[0172] S91: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the anti-lock braking torque T? b_λ ;

[0173] If so, the vehicle braking torque T b for:

[0174] T b =T m +T r ;

[0175] If not, the vehicle braking torque T b for:

[0176] T b =T r ;

[0177] Among them, T r is the regenerative braking torque, T m is the friction braking torque;

[0178] S92: Get the regenerative braking torque T r With friction braking torque T m Then go to step 10.

[0179] In this embodiment, in the above step 10, the friction braking torque obtained is used to compensate the braking torque by utilizing the rapid response characteristic of the regenerative braking system. The regenerative braking compensation torque is expressed as:

[0180]

[0181] Among them, T r is the regenerative braking torque, T m is the friction braking torque, is the actual friction braking torque.

[0182] Through the above embodiments, the technical solution of the present invention is disclosed. Compared with the prior art, the present invention has the following advantages:

[0183] 1. The driver braking intention recognition method based on hidden Markov interval type-2 fuzzy neural network of the present invention can effectively process the uncertain information of the driver and the braking system during the braking process, and significantly improve the driver braking intention recognition accuracy;

[0184] 2. The electric commercial vehicle braking energy recovery method based on the driver's braking intention of the present invention fully considers the driver's braking intention to distribute the braking torque, and at the same time utilizes the advantage of the rapid response of the regenerative braking system to compensate for the slow response problem of the friction braking process of the commercial vehicle, thereby fully ensuring braking safety and improving braking comfort.

[0185] The above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Other modifications or equivalent substitutions made to the technical solution of the present invention by ordinary technicians in this field should be included in the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solution of the present invention.

Claims

1. A braking energy recovery method for an electric-driven commercial vehicle taking braking intention into consideration, characterized in that: The method is an energy recovery method of a driver's braking intention based on a hidden Markov interval type II fuzzy neural network, comprising: a braking intention recognition process of a hidden Markov interval type II fuzzy neural network, and a braking energy recovery process based on the driver's braking intention; The method comprises the following steps: Step 1: Establish a hidden Markov model of the driver's braking intention, optimize the parameters of the hidden Markov model, and obtain the maximum likelihood estimation array based on the collected brake pedal data; Step 2: Based on the maximum likelihood estimation array, after being processed as input, an interval type-II fuzzy neural braking intention recognition network is established, the output of the interval type-II fuzzy neural braking intention recognition network is the braking intention, and the input of the interval type-II fuzzy neural braking intention recognition network is fuzzified; Step 3: Based on the fuzzification of the input variables in step 2, perform fuzzy reasoning of the interval type-2 fuzzy neural braking intention recognition network; Step 4: interval type-2 fuzzy neural braking intention recognition network fuzzy reduction; Step 5: Interval type-2 fuzzy neural braking intention recognition network output; Step 6: Identify the driver's braking intention based on the hidden Markov interval type-2 fuzzy neural network and judge the driver's braking intention, which includes light braking, moderate braking, and emergency braking; Determine whether the driver's braking intention is slight braking, if so, proceed to step 7, if not, further determine whether the driver's braking intention is moderate braking, if so, proceed to step 8, if not, it is emergency braking, proceed to step 9; Step 7: When the driver's braking intention is slight braking, the maximum regenerative braking torque of the motor carried by the commercial vehicle electric drive axle is calculated, and after obtaining the regenerative braking torque and the friction braking torque, step 10 is entered; Step 8: When the driver's braking intention is moderate braking, the maximum regenerative braking torque and rear wheel braking torque of the motor carried by the commercial vehicle electric drive axle are calculated, and after obtaining the regenerative braking torque and the friction braking torque, step 10 is entered; Step 9: When the driver's braking intention is emergency braking, the maximum regenerative braking torque and the anti-lock braking demand braking torque of the motor carried by the commercial vehicle electric drive axle are calculated, and after the regenerative braking torque and the friction braking torque are obtained, step 10 is entered; Step 10: Using the obtained friction braking torque to compensate the braking torque by using the regenerative braking system; Step 11: The vehicle controller controls the regenerative braking system and the friction braking system to implement corresponding braking torques according to the magnitudes of the regenerative braking torque and the friction braking torque until the vehicle speed is reduced to zero.

2. The method for braking energy recovery of an electric commercial vehicle taking braking intention into consideration according to claim 1, characterized in that: In step 1, establishing the driver's braking intention hidden Markov model includes the following steps: S11: According to the driver's braking operation and the vehicle system status, the input of the hidden Markov model of the driver's braking intention is determined, and its observation sequence vector is: Among them, O t is the hidden Markov model observation sequence vector, d(t) represents the brake pedal displacement, represents the brake pedal speed, t is the time; S12: Based on the pedal displacement d(t) and brake pedal speed signal collected by the brake pedal sensor The Baum-Welch algorithm is used to identify the parameters of the hidden Markov model of the driver's braking intention. The parameters of the hidden Markov model of the driver's braking intention are: λ=(π,A,B); Among them, λ is the hidden Markov model of the driver's braking intention, π is the initial probability vector, A is the transition probability matrix, and B is the observation probability density function; The expression of the observation probability density function is: Among them, N(O,μ ij ,σ ij ) represents the q-dimensional Gaussian density function of state p, ω pq represents the weight coefficient of the qth mixing function under state p, μ pq represents the vector mean of the qth mixed Gaussian function under state p, and the covariance matrix of the σ mixing function; S13: If ε t (p,q) is the probability of the qth mixed Gaussian function of the state p observation sequence at time t, then the Markov chain probability of state p at time t and state q at time t+1 is: Among them, P[O,μ pq ,σ pq ] is the q-th dimension Gaussian density function in state p, α t (p) is the forward probability of state p, β t (p) is the backward probability of state p, σ pq is the covariance matrix of the qth mixing function under state p; After optimizing the parameters of the hidden Markov model, based on the collected brake pedal data, the matching degree with the multidimensional hidden Markov model is calculated using a forward-backward algorithm to obtain a maximum likelihood estimation array.

3. The braking energy recovery method for an electric-driven commercial vehicle taking braking intention into consideration according to claim 1, characterized in that: In step 2, the process of establishing an interval type II fuzzy neural braking intention recognition network includes the following steps: S21: Determine the input and output of the interval type II fuzzy neural braking intention recognition network; The maximum likelihood estimation array is normalized to obtain the Markov model state cumulative probability vector under three braking intentions, and it is used as the input x i ; Where i = 1, 2, 3; 1 represents light braking, 2 represents moderate braking, and 3 represents emergency braking; ω0 is the initial weight, w k , σ k are the center and width of the radial basis function of the kth neuron respectively; S22: Interval type-2 fuzzy neural braking intention recognition network input fuzzification; The interval type-2 fuzzy set of the input layer neuron x1 is Type II fuzzy sets of input layer neuron x2 Type II fuzzy sets of input layer neuron x3 Among them, NB means negative big, NS means negative small, ZE means zero, PS means positive small, and PB means positive big; Therefore, the number of neurons in the hidden layer is 15, and the activation function of each neuron, that is, the membership function, is: in, is the upper bound of the output value of the j-th membership function layer neuron, u ij (x i ) is the lower bound of the output value of the j-th membership function layer neuron, c ij is the lower bound of the central value of the membership function of the neuron in the jth membership function layer, is the upper bound of the central value of the membership function of the neuron in the jth membership function layer, σ ij is the membership function width of the neuron in the jth membership function layer, j = 1, 2, ..., 15.

4. The method for braking energy recovery of an electric-driven commercial vehicle taking braking intention into consideration according to claim 3, characterized in that: In step 3, the interval type II fuzzy neural braking intention recognition network fuzzy reasoning includes the following: The input variables of the interval type II fuzzy neural braking intention recognition network are fuzzified by the membership function to obtain the membership interval of the variable. Each node in the incentive layer represents a fuzzy rule, and the activation interval of each rule is F j (x1,x2,x3), the expression is: Among them, f j is the lower bound of the activation of the jth rule, is the upper bound of the activation degree of the j-th rule.

5. The method for braking energy recovery of an electric-driven commercial vehicle taking braking intention into consideration according to claim 1, characterized in that: In step 4, the interval type II fuzzy neural braking intention recognition network fuzzy reduction includes the following: The reduction method uses the Enhanced Karnic-Mendel (EKM) algorithm, and the output expression of this layer is: Among them, J is the number of fuzzy rules, y iL The left endpoint of the network drop output, y iR is the right endpoint of the network degradation output, L is the left turning point of the EKM algorithm, and R is the right turning point of the EKM algorithm.

6. The braking energy recovery method for an electric-driven commercial vehicle taking braking intention into consideration according to claim 5, characterized in that: In step 5, the output of the interval type II fuzzy neural braking intention recognition network includes the following: The fifth layer of the interval type-2 fuzzy neural braking intention recognition network is the output layer, whose nodes represent the output of the entire network. The output expression of the network is: The error back propagation algorithm is used to train the neural network using the gradient descent method, and the performance indicators are defined as: The parameters that need to be learned in the above network include the uncertainty center c of the type II Gaussian membership function j and Rule consequent parameter w i and The parameter correction formula is: Among them, τ w Rule post-condition parameter w j and The learning rate, τ c is the uncertain center c j and c j The learning rate.

7. The method for braking energy recovery of an electric-driven commercial vehicle taking braking intention into consideration according to claim 1, characterized in that: In step 7, the maximum regenerative braking torque T of the motor carried by the commercial vehicle electric drive axle is calculated. r_max , including the following steps: S71: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the required braking torque T? b_i ; If so, the vehicle braking torque T b for: T b =T r ; If not, the vehicle braking torque T b for: Among them, T b_f is the front axle braking torque, T b_r is the rear axle braking torque, T r is the regenerative braking torque, T m is the friction braking torque; S72: Obtaining the regenerative braking torque T r With friction braking torque T m Then go to step 10.

8. The method for braking energy recovery of an electric-driven commercial vehicle taking braking intention into consideration according to claim 1, characterized in that: In step 8, the maximum regenerative braking torque T of the motor carried by the commercial vehicle electric drive axle is calculated. r_max and rear wheel braking force T b_r , including the following steps: S81: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the required braking torque T? b_r ; If so, the vehicle braking torque T b for: If not, the vehicle braking torque T b for: Among them, T b_f is the front axle braking torque, T b_r is the rear axle braking torque, T r is the regenerative braking torque, T m is the friction braking torque; S82: Get the regenerative braking torque T r With friction braking torque T m Then go to step 10.

9. The method for braking energy recovery of an electric-driven commercial vehicle taking braking intention into consideration according to claim 1, characterized in that: In step 9, the maximum regenerative braking torque T of the motor mounted on the commercial vehicle electric drive axle r_max and the anti-lock braking demand torque T b_λ , including the following steps: S91: Determine the maximum regenerative braking torque T of the motor r_max Is it greater than the anti-lock braking torque T? b_λ ; If so, the vehicle braking torque T b for: T b =T m +T r ; If not, the vehicle braking torque T b for: T b =T r ; Among them, T r is the regenerative braking torque, T m is the friction braking torque; S92: Get the regenerative braking torque T r With friction braking torque T m Then go to step 10.

10. The method for braking energy recovery of an electric-driven commercial vehicle taking braking intention into consideration according to claim 1, characterized in that: In step 10, the regenerative braking compensation torque is expressed as: Among them, T r is the regenerative braking torque, T m is the friction braking torque, is the actual friction braking torque.

Images