CN102981408B - Running process modeling and adaptive control method for motor train unit - Google Patents

Abstract

The invention discloses a modeling and self-adaptive control method for the running process of an EMU. In view of the characteristics of EMU operation process complexity, incomplete information, and obvious nonlinear characteristics, the method adopts a data-driven modeling method and proposes a TS bilinear model identification method; Constraints such as speed limit conditions, traction force/braking force saturation nonlinearity and other constraints are used to establish the EMU constraint model; according to the above model, a real-time learning model predictive control algorithm is designed to improve the multi-objective optimization control performance index. The invention provides a set of reliable basis for the crew members of the EMU to optimize the operation, ensure safe and punctual operation, improve operation comfort and reduce energy consumption. The invention is suitable for on-line identification and multi-objective optimal control of the running state of the EMU.

Description

A Modeling and Adaptive Control Method for EMU Running Process

technical field

The invention relates to a modeling and self-adaptive control method for the running process of a multiple train set, and belongs to the technical field of online identification and optimal manipulation of the running state of the multiple train set.

Background technique

China's new generation of high-speed trains can reach a speed of 350 kilometers per hour in continuous operation, and the test speed has exceeded 400 kilometers per hour. It is the EMU with the fastest commercial operation speed, the highest technological content, and the best system matching in the world. Compared with other modes of transportation (such as automobiles, airplanes, and medium- and low-speed passenger and freight trains), EMUs can meet transportation needs such as long distances, large volumes, high density, and short travel times. my country currently has the world's largest high-speed railway network with the highest operating speed. By 2020, 16,000 kilometers of high-speed railway will be built. However, China's high-speed railway network has some important characteristics that are different from European and Japanese high-speed railways, such as large-scale road network, significantly increased energy consumption; complex and changeable geographical, geological, and climatic conditions; and obvious differences in the operating conditions of passenger dedicated lines of different speeds, etc. . It has become a technological development trend to study intelligent EMU control systems with self-test, self-diagnosis, and self-decision-making capabilities to achieve safe and reliable operation and optimal operation.

Aiming at the problem of modeling and control of the EMU operation process, relevant scholars have established a linearized mechanism model for the dynamic characteristics of the EMU and designed a Constant speed controller; a multi-mass unit displacement mechanism model was proposed to describe the dynamic process of the EMU, and a fuzzy controller was used to realize the tracking control of speed and displacement; a nonlinear constraint model of EMU traction and braking conditions was established and An adaptive backstepping controller is designed; the energy consumption model of the EMU at different target speeds is studied. However, these mechanism modeling methods are difficult to solve the nonlinear problems of EMU aerodynamics and handle level changes, which may lead to a decrease in the system's comprehensive evaluation index.

Under the condition of high-speed and high-density operation, it is difficult for the manual operation method to meet the multi-objective optimization requirements. Therefore, it is of great significance to study the intelligent operation optimization algorithm for improving the operation performance of EMUs. For example, relevant scholars have proposed a fuzzy control model based on fuzzy C-means clustering analysis and genetic algorithm optimization for the optimization of EMU running time and energy consumption; proposed an intelligent train control system based on fuzzy logic control to optimize the high-speed and high-density operation of EMUs performance; model switching and optimization strategies are used to study the minimum energy consumption control problem of electric multiple units; some scholars also consider the joint design problem of economic control and punctual operation of EMUs under the influence of uncertain factors, and use genetic algorithm and fuzzy linear programming methods respectively optimize. However, most of these methods focus on off-line optimization, which cannot meet the real-time optimization requirements of EMUs well. As a dynamic optimization method, the nonlinear model predictive control algorithm can better solve the slow time-varying nonlinear optimization problem. However, due to the large amount of online calculation in the nonlinear optimization process, it does not have obvious advantages when directly applied to the operation control of EMUs.

Invention content

The purpose of the present invention is to solve the nonlinear problems of aerodynamics and handle level changes that are difficult to solve in the modeling method for the modeling of the EMU operation process, which may lead to the problem that the comprehensive evaluation index of the system is reduced; in addition, the fuzzy logic control based The intelligent train control system to optimize the high-speed and high-density operation performance of EMUs cannot meet the real-time optimization requirements of EMUs. In view of these problems, the present invention discloses a method for modeling and adaptive control of the running process of EMUs, which establishes a T-S bilinear model and a multi-objective constraint model; adopts a local modeling strategy based on real-time learning to dynamically correct model parameters, and accordingly A bilinear adaptive model predictive controller is designed for rolling optimization and closed-loop control to achieve multi-objective optimal control of EMUs such as safety and stability, energy saving and comfort, and punctuality.

The technical solution for realizing the present invention is that the present invention combines the dynamic equations of the EMU and the characteristics of handling control to establish a TS bilinear fuzzy model, and uses an instant learning algorithm to realize the adaptive adjustment of model parameters; the TS bilinear model identification is proposed method. The present invention establishes the multi-objective constraint model of the EMU based on constraints such as the actual operation diagram of the EMU, the situation of the front line, the speed limit condition, the traction force/braking force saturation nonlinear characteristics, and designs a bilinear model predictive controller based on this to realize the optimization of the EMU run. And based on the above model, a real-time learning model predictive control algorithm is designed to improve the multi-objective constraint optimization control performance index of the EMU. That is, when the model output error When the system is within the allowable range, the parameters of the TS bilinear model do not need to be optimized; when the model output error exceeds the threshold, the model parameters are corrected online by using an instant learning algorithm. Based on this, the parameters of the bilinear model predictive controller are dynamically adjusted, and the simultaneous optimization of model parameters and controller parameters is realized, which reduces the online calculation amount of the bilinear model predictive controller. The whole identification process can not only reduce the influence of the unmodeled part of the TS bilinear model and unknown faults or disturbances, but also reduce the calculation amount of the real-time learning algorithm, and improve the multi-objective constraint optimization control level of the system.

The steps of the EMU operation process modeling and adaptive control method are as follows:

(1) Starting from the basic principle of EMU operation control, based on the characteristics of the handling control and dynamic equations, a bilinear model describing the dynamic characteristics, energy consumption and running time of the EMU is established; FCM based on genetic simulated annealing is adopted The fuzzy clustering algorithm clusters and analyzes the collected sample data, obtains the economic control points of the operation process, and provides the flight attendants with prior manipulation and optimization information; determines the antecedent parameters of each fuzzy rule according to the determined model structure, and minimizes the weight by recursion The square method identifies the subsequent parameters of each fuzzy rule, and accurately describes the dynamic characteristics of each local fuzzy rule; when the model output error exceeds the preset threshold, the local modeling method based on real-time learning is used to correct the model parameters online. In this way, the online estimation of the multi-mode model of the running process of the EMU can be realized.

(2) By establishing the multi-objective constraint optimization model of the EMU, on the basis of the T-S bilinear model established above, combined with the nonlinear model predictive control, the predictive model is converted, and the T-S bilinear adaptive model predictive controller is designed To study the optimal maneuvering problem of EMUs.

EMU handling control mainly includes three operating conditions of traction, coasting and braking, involving various control modes such as starting, acceleration, constant speed, coasting and braking. Among them, there are multiple handle levels under the traction and braking conditions, as shown in Figure 3, the handle operated by the high-speed train attendant. All car control commands are issued from the handle level, and different handle control levels determine different control modes. The traction characteristic curves and braking characteristic curves of the EMU at different handle levels are shown in Figure 4(a) and Figure 4(b) respectively.

EMU mechanism model is determined by the following principles and methods in the present invention:

It can be seen from Figure 4 that the traction/braking force required for the operation of the EMU and running speed and handle level There is a multivariate nonlinear relationship between:

(1)

In addition, because different handle control levels determine different control modes, and and are closely related, whose nonlinear dynamics present control variables with state variable The phenomenon of multiplication can be equivalent to describe the running process of the EMU as a bilinear system.

For the bilinear system in the operation process of the EMU, considering that the relative displacement between each power unit is usually approximately zero, and the speed of each vehicle is approximately equal, the electrical separation phase point, ramp and curvature are functions of distance, and the distance is Independent variables are more appropriate, then The dynamic characteristics, energy consumption and running time of the EMU can be described by the following bilinear model:

(2)

(3)

(4)

In the formula: is the equivalent total mass of the EMU; is the running distance of the EMU, and the system input is the control force acting on the EMU at different handle levels (traction effort/braking effort); system output is speed ;symbol Kronecker operator such that and Satisfy the product relationship; G is the line parameter (electrical separation phase point, slope and curvature); and represent the start and end positions, respectively; is the running time; is the energy consumption during the operation of the EMU; is the coefficient of mechanical resistance, generally in the about; is the air resistance coefficient, generally in the around when , the nonlinear air resistance term The proportion in formula (2) is small.

In order to simplify the train operation process modeling and controller design, many engineering applications and researchers ignore its influence. For example, without considering nonlinear air resistance, relevant researchers designed a fuzzy PID gain controller to adjust the running speed of subway trains; used adaptive optimization control algorithm to solve the problems of subway train operation management and energy-saving optimization; An open-loop heuristic optimization strategy and a heuristic-based closed-loop LQR control algorithm are proposed for speed control of a freight train dynamics model. but when , the nonlinear air resistance term The proportion in formula (2) is getting larger and larger, becoming the main resistance to be overcome during the operation of the EMU, and its energy consumption is also increasing. The linear modeling and control method based on ordinary medium and low-speed trains is difficult to satisfy High-precision tracking control and multi-objective optimization requirements in the running process of EMUs.

Adopt T-S bilinear model self-learning predictive control algorithm in the present invention to research the running process of EMU:

For each fuzzy rule, using product reasoning machine, single value fuzzer and center average defuzzification, the output of T-S fuzzy bilinear model is:

(5)

In the formula: Indicates that for system input and output variables, the first The fuzzy membership degree of the rule.

(1) T-S bilinear model structure identification

How to optimize the structure of the T-S bilinear model has an important impact on the speed and accuracy of model identification. Commonly used structure identification methods include FCM clustering algorithm, etc., but the local searchability of FCM and the sensitivity to the initial value of the cluster center limit its application. The invention adopts the FCM clustering analysis method based on the genetic simulated annealing algorithm to identify the structure of the EMU model. The algorithm inherits the strong parallel and global search ability of the genetic algorithm, and adopts the Metropolis acceptance criterion of the simulated annealing algorithm to maintain the diversity of the population, improve the local search ability, and overcome the premature phenomenon of the genetic algorithm and the low convergence speed of the simulated annealing algorithm. defect.

The above clustering algorithm can be obtained by specifying The global optimal cluster center in each category. However, the operating conditions of EMUs are complex and changeable, and it is difficult to determine the system operating point in advance. The performance of clustering algorithms under different categories can be measured by the effectiveness index. The Davies-Bouldin (DB) index is a class of classical clustering validity evaluation index, which uses intra-class compactness and inter-class separation to evaluate the quality of clustering results.

(2) Parameter identification of EMU T-S bilinear model

Based on the principle of model parameter identification, formula (5) can be transformed into the following form:

(6)

In the formula: is the observation vector, and satisfy:

;

is the parameter to be identified. This is a typical least squares estimation problem, and the parameters can be obtained by the following formula :

(7)

But its objective function is optimized globally, which cannot accurately describe the dynamic characteristics of each local fuzzy rule.

The invention adopts the recursive weighted least square method to iteratively identify model parameters and avoid matrix inversion.

(3) Online correction of model parameters based on real-time learning

In order to improve the efficiency of the real-time learning algorithm, the present invention enables the real-time learning algorithm to correct and update the learning set only when the model output error exceeds the threshold. How to establish the learning set is the main factor affecting the accuracy of the model. In order to improve the accuracy of online modeling, the present invention takes the bilinear dynamic change trend of the EMU into the criterion for selecting samples.

The present invention designs a bilinear adaptive model predictive controller to perform rolling optimization and closed-loop control, and realize multi-objective optimization control such as safety and stability of the EMU, energy saving and comfort, and punctuality. The main calculation steps are as follows:

By minimizing the objective function, the T-S bilinear model predictive control algorithm can accurately describe its dynamic process and give the optimal control sequence, but the bilinear term in the model is nonlinear, and the online calculation of multi-step prediction is large . In order to reduce the calculation amount of online optimization, formula (5) can be transformed into:

(8)

During the operation of the EMU, the design of the controller should make the operation of the EMU stable, comfortable, energy-saving, punctual and ensure precise parking. Then the multi-objective constrained optimization model for EMU operation is:

(9)

In the formula: and are the runtime weight and energy consumption weight respectively, is the target speed, is the speed tracking error range, is the running time of the graph, is the expected energy consumption of interval operation, represents the maximum braking force, is the control increment, Represents maximum traction.

Since the energy-saving index and smooth comfort can be described by the change of the control quantity, and the punctuality rate and accurate parking index can be realized by accurately tracking the optimal speed curve, the objective function can be expressed as:

(10)

In the formula: is the future velocity reference trajectory, are the minimum output length, predicted length, and control length, respectively, is the weighted coefficient sequence, constraining the control quantity; Controls the sequence of increments for the future.

Based on the rolling optimization mechanism, the optimal control law can be obtained:

(11)

in

will be calculated The first value of a control increment is put into effect, enabling scrolling optimization. in the current At sampling positions, the control quantity is expressed as:

(12)

To sum up, for the modeling and real-time optimal control of the EMU operation process, the T-S bilinear model can be established to effectively describe its multi-mode operation based on its traction characteristic curve, coasting resistance curve, braking mode curve and operating data. Process; Designing instant learning model predictive controllers to improve their multi-objective optimal control performance metrics.

The beneficial effect of the present invention compared with the prior art is that the operation of the EMU involves various scenarios such as ramps, tunnels, bridges, electric phase separation, and weather changes. The operation process is complicated, the information is incomplete, and the nonlinear characteristics are obvious. The traditional control method It is difficult to establish an effective description model and implement multi-objective optimization operations such as safety, punctuality, energy saving, and comfort. The present invention firstly proposes that the EMU runs regularly within the specified time period and interval according to the established operation diagram, and the dynamic relationship between the control variable and the state variable follows the change of the traction characteristic curve, the coasting resistance curve and the braking mode curve, which is based on data The driving modeling and optimization control methods provide the possibility; then the T-S bilinear fuzzy model is established by combining its dynamic equations and handling control characteristics, and the real-time learning algorithm is used to realize the adaptive adjustment of model parameters; according to the actual operation diagram of the EMU, Constraint models are established based on constraints such as line conditions ahead, speed limit conditions, traction force/braking force saturation nonlinear characteristics, and a bilinear model predictive controller is designed based on this to realize optimal operation of EMUs and provide prior information for flight attendants to optimize manipulation, thus changing It eliminates the blindness of empirical adjustment, and is an effective auxiliary means for optimal operation. The present invention is more intuitive and rapid, and is not limited by conditions such as site and environment, and has the advantages of being simple and practical, improving the optimal manipulation level of high-speed rail attendants, reducing human resource investment, improving the operating efficiency of railway departments, and reducing costs.

The invention is suitable for on-line identification and multi-objective optimal control of the running state of the EMU.

Description of drawings

Fig. 1 is the structural diagram of the EMU operation process control system;

Fig. 2 is the basic principle of EMU operation control;

Fig. 3 is the EMU main control unit;

Figure 4(a) is the traction characteristic curve of the EMU under different handle levels;

Figure 4(b) is the braking characteristic curve of the EMU under different handle levels;

Figure 5 is a line speed limit diagram;

Fig. 6 is the actual operation process diagram of the EMU;

Fig. 7 is output and error curve thereof of modeling method of the present invention;

Fig. 8 (a) is the velocity tracking and error curve that control method of the present invention obtains;

Fig. 8 (b) is the control force change curve that the inventive method obtains;

Fig. 8(c) is the optimized energy consumption and running time curve obtained by the method of the present invention.

Detailed ways

The present invention is specifically implemented in the Jinan-Xuzhou East downlink section of the Beijing-Shanghai high-speed railway under the jurisdiction of a certain railway bureau, and the operating data is collected on-site on the EMU CRH380AL. It passes through Tai'an Station, Qufu East Station, Tengzhou East Station and Zaozhuang Station, but only stops at Tai'an Station for 2 minutes. The starting mileage is 393.74km, the mileage at Tai'an Station is 465.77km, and the final mileage is 693.74km. The operating conditions of high-speed EMUs are complex and subject to constraints such as line gradient (the maximum gradient has reached 20‰), operating time, and speed limit. There are 9 tunnels in the whole process, 11 electrical separation phase points, and 17 slope values exceeding 12‰. Table 1 shows the operating hours and minutes between stations specified in the operation diagram, and Figure 5 is the line speed limit diagram.

Table 1 Interval operation timetable

interval name Interval running hours and minutes (hour: minute: second) Jinan→Taian 09:38:30→10:03:19 Tai'an → Qufu East 10:05:19→10:21:39 Qufu East → Tengzhou East 10:21:39→10:33:02 Tengzhou East → Zaozhuang 10:33:02→10:40:22 Zaozhuang→Xuzhou East 10:40:22→10:56:30

Figure 6 describes the actual running time and operating speed of this type of EMU on July 13, 2012. Among them, the actual running time is 1 hour, 19 minutes and 51 seconds, which is 1 minute and 51 seconds later than the time specified in the operation diagram of 1 hour and 18 minutes; the braking process adopts the electro-pneumatic combined braking method, in which the regenerative power can be fed back to the grid, and the total energy consumption It is the traction power consumption minus the regenerative braking power, and its value is 14230kwh.

In the embodiment of the present invention, based on the traction/braking characteristic curves of the CRH380AL EMU at different handle levels, 2000 sets of data in field operation are preprocessed to obtain 1800 sets of valid data within the range. Utilize the clustering effectiveness algorithm of the present invention to analyze these data, when the number of working conditions is 6, the DB value is the smallest, that is, the number of optimal fuzzy rules is 6, and the corresponding optimal working points of the model are respectively:

start-up conditions;

Low and medium speed coasting conditions;

Constant power traction working condition;

High speed idling condition;

High-speed braking and speed regulation working conditions;

Low-speed braking and parking conditions;

Figure 7 shows the output of the T-S bilinear model based on real-time learning and its error curve with the actual output.

As can be seen from Fig. 7, in the multi-working mode operation process of the EMU, the model output of the modeling method of the present invention can still track the variation of the actual output better (the root mean square error is ). Especially in the transition stages of different traction handle levels and different brake handle levels, the maximum positive error and minimum negative error output by the model are only and , whose absolute values are all within the line speed limit range, the model identification accuracy and generalization ability are high, which can better meet the CTCS-3 train control system error requirements, that is, 30 the following 2 , 30 The above does not exceed 2% of the speed value.

In order to further verify the validity of the modeling and control method in this paper and adapt to the changes of objects and disturbance characteristics, so that the system working area is located in the most economical operating area, a multi-objective optimal control simulation experiment is carried out according to the field operation data of the EMU.

Using the predictive control algorithm based on the TS bilinear self-learning model, Figure 8(a) shows that the method in this paper can further improve the speed tracking performance index in the complex operating environment of the EMU, and the operating safety index can also be improved. Its maximum control error and minimum negative control The error is and , both meet the target speed error requirements; the root mean square error of the control system ( ) is significantly better than the root mean square error of the modeled system ( ). It can be seen from Fig. 8(b) that the change curve of the control force conforms to the change of working conditions, that is, the control force in the traction condition is greater than zero, the control force in the coasting condition is equal to zero, and the control force in the braking condition is less than zero; the control force in the starting process is based on The time optimal principle changes, the constant speed process control force changes in an orderly manner in accordance with the principle of energy saving, constant power and idling conditions, the braking process uses regenerative braking to feed energy back to the grid for energy saving, reduces the frequency of use of the maximum braking, and improves During the whole interval running, the control force maintains smooth transition and switching, which improves the comfort of passengers. Fig. 8 (c) has described the optimization energy consumption that the inventive method obtains and running time curve, and energy consumption curve is downward trend in braking process; When energy consumption increases obviously, running time growth rate slows down; The energy consumption is 14095kwh, the running time is 1 hour, 17 minutes and 42 seconds, and the operation is on time. Compared with the flight attendant's experience control method, this algorithm saves 265kwh of energy saving, which is 1.88% of relative energy saving. It can better meet the multi-objective optimization of safety, energy saving, punctuality, stability and comfort. Require.

Claims (3)

1. A kind of EMU operation process modeling and self-adaptive control method, it is characterized in that, described method is based on EMU actual operation chart, line situation ahead, speed limit condition, tractive force/braking force saturation non-linear characteristic constraint condition establishes EMU A multi-objective constraint optimization model for group operation, based on which a bilinear model predictive controller is designed to realize the optimal operation of the EMU; and a real-time learning model predictive control algorithm is designed according to the above model to improve the multi-objective constraint optimization control performance index of the EMU; when the model outputs When the error e is within the allowable range of the system, the parameters of the T-S bilinear model do not need to be optimized; when the model output error exceeds the threshold, the model parameters are corrected online using an instant learning algorithm; accordingly, the parameters of the bilinear model predictive controller are dynamically adjusted , realizing the simultaneous optimization of model parameters and controller parameters, reducing the amount of online calculation of the bilinear model predictive controller; The multi-objective constraint optimization model for the operation of the EMU is: $\mathrm{<span\; class="notranslate">\; min</span>}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; |</span>\frac{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; *</span>}}{{\mathrm{<span\; class="notranslate">\; t</span>}}^{<span\; class="notranslate">\; *</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; E.</span>}}<span\; class="notranslate">\; |</span>\frac{\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; E.</span>}}^{<span\; class="notranslate">\; *</span>}}{{\mathrm{<span\; class="notranslate">\; E.</span>}}^{<span\; class="notranslate">\; *</span>}}<span\; class="notranslate">\; |</span>$ (ν _r -δ)≤ν≤(ν _r +δ) t≤t ^* U _min ≤ U ≤ U _max Δu _min ≤ Δu ≤ Δu _max In the formula: F represents the objective function; W _T is the weight of running time; ^W _E is the weight of energy consumption; v _r is the target speed; δ is the error ^range of speed tracking; Expected energy consumption; U _min represents the maximum braking force; Δu is the control increment; U _max represents the maximum traction force; Δu _min is the lower limit of the control increment; Δu _max is the upper limit of the control increment; The bilinear model is: The bilinear model of the dynamic characteristics, energy consumption and running time of the N-car EMU is: $\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; dx</span>}$ $\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\overset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{<span\; class="notranslate">\; \&Integral;</span>}}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; dx</span>}$ In the formula: M is the equivalent total mass of the EMU; x is the running distance of the EMU; the system input is the control force U(x) acting on the EMU at different handle levels; the system output is the speed v(x) ;symbol The operator makes U(x) and v(x) satisfy the product relationship; G is the line parameter; x ₁ represents the position of the starting point; x ₂ represents the position of the end point; t is the running time; energy consumption; C ₁ is the mechanical resistance coefficient, which is 7.4×10 ^-3 , and C ₂ is the air resistance coefficient, which is 1.24×10 ^-4 . 2. A kind of EMU operation process modeling and adaptive control method according to claim 1, is characterized in that, described method step is: (1) Starting from the basic principle of EMU operation control, and based on the characteristics of maneuvering control and dynamic equations, establish a bilinear model describing the dynamic characteristics, energy consumption and running time of EMUs; determine according to the established model structure For each fuzzy rule antecedent parameter, each fuzzy rule subsequent parameter is identified by the recursive weighted least square method, and the dynamic characteristics of each local fuzzy rule are accurately described; when the model output error exceeds the preset threshold, the real-time based The local modeling method of learning corrects the model parameters online; in this way, the online estimation of the multi-mode model in the operation process of the EMU can be realized; (2) On the basis of the T-S bilinear model established above, combined with the nonlinear model predictive control, the predictive model is converted, and the T-S bilinear adaptive model predictive controller is designed to study the optimal handling of EMUs. 3. A kind of EMU operation process modeling and adaptive control method according to claim 1, it is characterized in that, described T-S bilinear model identification and adaptive control method comprise: (1) For parameter identification of the EMU T-S bilinear model, the recursive weighted least squares method is used to iteratively identify model parameters and avoid matrix inversion; (2) Online correction of model parameters based on real-time learning. When the model output error exceeds the threshold, the real-time learning algorithm is used to correct and update the learning set; design a bilinear adaptive model predictive controller for rolling optimization and closed-loop control to realize EMU Safe and stable, energy-saving and comfortable, punctual multi-objective optimal control.