CN112821826A

CN112821826A - Multi-dimensional integrated vehicle-mounted magnetic suspension flywheel battery control system

Info

Publication number: CN112821826A
Application number: CN202110007208.6A
Authority: CN
Inventors: 张维煜; 高映洁; 韩啸雅; 沈琳烽; 俞珏鑫
Original assignee: Jiangsu University
Current assignee: Jiangsu University
Priority date: 2021-01-05
Filing date: 2021-01-05
Publication date: 2021-05-18
Anticipated expiration: 2041-01-05
Also published as: CN112821826B

Abstract

The invention provides a multidimensional integrated vehicle-mounted magnetic suspension flywheel battery control system, and belongs to the technical field of vehicle-mounted flywheel battery control. The current signal processed by the current normalization module is used as the input signal of the support vector machine displacement prediction module, the output signal of the fuzzy PID cross feedback control module and the output signal of the neural network inverse decoupling control module which are processed by the accumulation and coordinate system conversion module are used as the input signal of the improved zero-power control module, the displacement signal output by the support vector machine displacement prediction module and the output signal of the improved zero-power control module are used as the input signals of the linear closed-loop controller, and the output of the linear closed-loop controller is used as the input signal of the fuzzy PID cross feedback control module and the input signal of the neural network inverse decoupling control module. The invention simplifies the complexity of the whole system, greatly improves the real-time response characteristic, can inhibit the gyro effect, improves the stability and reduces the energy consumption and the cost.

Description

Multi-dimensional integrated vehicle-mounted magnetic suspension flywheel battery control system

Technical Field

The invention relates to the technical field of control over a vehicle-mounted magnetic suspension flywheel battery (also called a flywheel energy storage device) for an electric automobile, in particular to a vehicle-mounted magnetic suspension flywheel battery control system integrating multiple dimensions of low energy consumption, low cost, high stability control and the like.

Background

The vehicle-mounted magnetic suspension flywheel battery is a novel mechanical and electrical integrated device based on the magnetic suspension bearing technology, breaks through the limitation of the traditional chemical battery, and has the advantages of high energy storage density, high energy conversion rate, long service life, no pollution and the like. The vehicle-mounted magnetic suspension flywheel battery is a key influence factor in the aspects of realizing engineering application popularization and stability control, energy consumption and cost of a flywheel rotor.

And (3) stability control aspect: the gyro effect of the flywheel rotor is aggravated by the driving state of the automobile and the complex road conditions, and the stability of the flywheel rotor of the vehicle-mounted magnetic suspension flywheel battery is influenced. The main methods for inhibiting the gyro effect of the flywheel rotor and improving the stability of the flywheel rotor comprise three types, namely a control algorithm based on a modern control theory, an intelligent decoupling algorithm and a cross feedback control algorithm based on a traditional decentralized PD controller. Linear state feedback decoupling control (nutty. active electromagnetic bearing flywheel energy storage system gyro effect inhibition research [ D ]. zhejiang: zhejiang university, 2012) in a control algorithm based on modern control theory relates to a feedback array, wherein parameters are related to the rotor rotation speed, and considering that the rotation speed of a flywheel battery rotor is easy to change, the control algorithm is required to need a speed observer with good performance, and meanwhile, the feedback array needs to be calculated on line when the rotation speed changes every time, so that the calculation amount is large, and the control algorithm is not easy to implement in actual engineering. The intelligent decoupling algorithm comprises a neural decoupling algorithm, a fuzzy decoupling algorithm, a sliding mode decoupling algorithm and the like, the algorithms need larger computer configuration resources, and the design of the controller is very complex. Compared with the two types of decoupling algorithms, the cross feedback control algorithm based on the traditional decentralized PD controller is simple and direct, small in calculated amount and easy to implement in engineering. All the control algorithms above only consider stable control and do not consider low energy consumption and cost issues.

And (3) low energy consumption control aspect: the vehicle-mounted magnetic suspension flywheel battery has high standby loss (high self-discharge rate). In the aspect of reducing energy consumption, low bias current control and zero bias current control are mainly adopted, and since the magnetic suspension bearing works in a nonlinear area, the two control methods need to adopt a proper nonlinear control strategy to be matched with magnetic suspension control, and are not easy to realize in engineering. Wang Xiao just (digital control system and control strategy research [ D ]. Nanjing: Nanjing aerospace university, 2009) proposed a displacement compensation method for keeping low power consumption of magnetic bearings, the axial direction is disturbed by external force, and the permanent magnet in the magnetic bearing generates eccentric tension to offset the external force by changing the axial suspension position of the rotor. The three methods only consider low energy consumption, and do not consider stability and cost problems.

Low cost: the vehicle-mounted magnetic suspension flywheel battery needs a large number of displacement sensors and is expensive, and the popularization and application of the vehicle-mounted flywheel battery in industry are greatly limited. The rotor displacement self-detection technology is adopted to replace a displacement sensor, so that the cost is reduced, and the main methods comprise a parameter estimation method, a high-frequency signal injection method, a non-sensing control method based on a neural network and the like. The parameter estimation method depends on an accurate mathematical model of the system, and has higher design requirements on the controller; the high-frequency signal injection method needs a special signal processing technology to realize the estimation of the displacement; the non-sensing control method based on the neural network utilizes stronger nonlinear mapping capability of the neural network to realize the self-detection of the rotor displacement, although the control method makes up the defects of the two methods, the neural network has the defects of overfitting, easy falling into local extremum, dependence on experience of structural design and the like at present. The three methods only consider the problem of low cost, and do not relate to the problems of low energy consumption and stability.

In addition, if the methods for solving the above problems are all stacked in a control system in the form of discrete modules, the control system is inevitably complex, the real-time response of the system is affected, and the method is not suitable for controlling the running state of the vehicle-mounted magnetic suspension flywheel battery system, which still maintains high stability, low energy consumption and low cost under the environment of complex and changeable external environment.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides a multidimensional integrated vehicle-mounted magnetic suspension flywheel battery control system which can inhibit the gyro effect, improve the stability and reduce the energy consumption and the cost.

The present invention achieves the above-described object by the following technical means.

A multidimensional integrated vehicle-mounted magnetic suspension flywheel battery control system comprises a linear closed-loop controller, a fuzzy PID cross feedback control module, a neural network inverse decoupling control module, an improved zero power control module, a coordinate system conversion module, an accumulation and coordinate system conversion module, a composite controlled object and a support vector machine displacement prediction module, wherein the linear closed-loop controller comprises a control mode switching module; the output of the linear closed-loop controller is respectively used as the input of a fuzzy PID cross feedback control module and the input of a neural network inverse decoupling control module, the output of the fuzzy PID cross feedback control module and the output of the neural network inverse decoupling control module through a coordinate system conversion module are used as the input for controlling a composite controlled object, and the input is used as the input of a support vector machine displacement prediction module through transformation multiplexing of the composite controlled object; the outputs of the fuzzy PID cross feedback control module and the neural network inverse decoupling control module are processed by the accumulation and coordinate system conversion module, and are multiplexed with the output of the support vector machine displacement prediction module to be used as the input of the improved zero power control module; the improved zero power control module generates a reference displacement output which is used as the input of the linear closed-loop controller together with the actual displacement output of the support vector machine displacement prediction module.

The composite controlled object comprises a first Clark inverse transformation module, a second Clark inverse transformation module, a third Clark inverse transformation module, a first switching power amplifier, a second switching power amplifier, a current sensor, a current integration module and an actual controlled object, wherein the actual controlled object comprises an axial magnetic bearing a2, a front radial magnetic bearing a1 and a rear radial magnetic bearing b 1;

the inputs of the first Clark inverse transformation module are: radial control current signal { i) output by fuzzy PID cross feedback control module_tax*、i_tay*}；

The input of the first switching power amplifier is: axial control current signal i output by fuzzy PID cross feedback control module_tz*；

The inputs of the third Clark inverse transformation module are: radial control current signal { i) output by fuzzy PID cross feedback control module_tbx*、i_tby*}；

The input of the second Clark inverse transformation module is: the neural network inverse decoupling control module outputs a radial control current signal { i) through the coordinate system conversion module_kax*、i_kay*、i_kbx*、i_kby*}；

The input of the second switching power amplifier is: axial control current signal i output by neural network inverse decoupling control module_kz*；

The outputs of the first Clark inverse transformation module, the first switching power amplifier, the third Clark inverse transformation module, the second Clark inverse transformation module and the second switching power amplifier pass through a current sensor and then are input into a current normalization module; the current signal { i ] is obtained after the current normalization module processes_au、i_av、i_aw、i_bu、i_bv、i_bw、i_zThe method is used as the input of a displacement prediction module of a support vector machine;

i is described_zAs input to the axial magnetic bearing a2, { i_au、i_av、i_awAs input to a front radial magnetic bearing a1, the i_bu、i_bv、i_bwAs input to a rear radial magnetic bearing b 1.

I is described_au＝i_tau+i_kau、i_av＝i_tav+i_kav、i_aw＝i_taw+i_kaw、i_bu＝i_tbu+i_kbu、i_bv＝i_tbv+i_kbv、i_bw＝i_tbw+i_kbw、i_z＝i_tz+i_kzThe said i_tau、i_tav、i_tawIs the output of the first Clark inverse transform module, said i_kau、i_kav、i_kawIs the output of the second Clark inverse transform module, said i_tbu、i_tbv、i_tbwIs the output of the third Clark inverse transform module, said i_kbu、i_kbv、i_kbwIs the output of the second Clark inverse transform module, said i_tzIs the output of the first switching power amplifier, said i_kzIs the output of the second switching power amplifier.

The accumulation and coordinate system conversion module inversely decouples the current signal { i) output by the control module to the neural network_kx*、i_ky*、i_kθx*、i_kθy*、i_kzAnd (4) outputting a current signal { i } by the fuzzy PID cross feedback control module_tax*、i_tay*、i_tz*、i_tbx*、i_tbyProcessing as follows:

will current signal i_tax*、i_tay*、i_tz*、i_tbx*、i_tbyLinear conversion to control current signal i in centroid coordinate system_tx*、i_ty*、i_tz*、i_tθx*、i_tθy*；

The five-freedom-degree control current signal under the mass center coordinate system is as follows:

i_x*＝i_kx*+i_tx*，i_y*＝i_ky*+i_ty*，i_θx*＝i_kθx*+i_tθx*，i_θy*＝i_kθy*+i_tθy*，i_z*＝i_kz*+i_tz*。

the control mode switching module comprises 5 PD controller switching modules, the PD controller switching module comprises 5 PD controllers, and the 5 PD controllers respectively correspond to five working conditions of turning, climbing, transverse vibration, longitudinal vibration and pitching vibration; the mathematical model of the magnetic bearing rotor system of the vehicle-mounted magnetic suspension flywheel battery under the five working conditions is as follows:

where m is the rotor mass and J is the moment of inertia of the rotor about the x-axis and the y-axis (J ═ J)_x＝J_y)，J_zIs the moment of inertia of the rotor about the z-axis, ω is the rotor rotational angular velocity, F_Ax、F_BxElectromagnetic forces in the x-direction, F, of a front radial magnetic bearing a1 and a rear radial magnetic bearing b1, respectively_Ay、F_ByElectromagnetic forces in the y-direction, F, of a front radial magnetic bearing a1 and a rear radial magnetic bearing b1, respectively_zFor the electromagnetic force of the axial magnetic bearing a2 in the z direction, f_x、f_y、f_zFor disturbing forces,/_aThe distance from the axle center of the front radial magnetic bearing to the mass center O, l_bThe distance from the axle center of the rear radial magnetic bearing to the mass center O,

and

to couple terms, k₁、κ₂、κ₃、κ、₄Kappa is a deviation value caused by multiple coordinate conversion, linear amplification and anti-interference links, and the deviation value is obtained through simulation; the PD controller adopts a self-adaptive fuzzy control algorithm, a displacement deviation signal e and a displacement deviation change rate de/dt are used as the input of the PD controller, and the PD controller outputs a second derivative phi of the displacement signal₁、φ₂、φ₃、φ₄、φ₅On-line adjustment of the proportionality coefficient k under the mathematical model of the magnetic bearing rotor system of the vehicle-mounted magnetic suspension flywheel battery under different working conditions_pAnd a differential coefficient k_d。

The neural network inverse decoupling control module comprises a fuzzy neural network inverse system, and the construction of the fuzzy neural network inverse system specifically comprises the following steps:

for (x, y, z, theta)_x、θ_y) Preprocessing the corresponding first derivative and second derivative, normalizing the preprocessed first derivative and second derivative to construct a fuzzy neural network training sample set;

training a fuzzy neural network offline by adopting a BP algorithm of a self-adaptive learning rate; the first layer of the fuzzy neural network is an input layer, and a fuzzy neural network training sample set is input; the second layer of the fuzzy neural network is a membership function input layer, and a central value alpha, a width sigma and a back-part parameter of the membership function are input; the third layer of the fuzzy neural network is a decision layer, and a fuzzy rule is formed to adjust the weight of the neural network on line; the fourth layer is a de-blurring layer for performing a sharpening operation.

Introducing a fuzzy PID cross feedback control module for three working conditions of stable operation, starting acceleration and braking deceleration, wherein the fuzzy PID cross feedback control module comprises 5 fuzzy controllers and 5 incomplete differential PID controllers; the displacement deviation signal e and the displacement deviation change rate de/dt are used as the input of a fuzzy controller, and the proportional parameter regulating quantity delta K is output through fuzzification processing, fuzzy reasoning and defuzzification_pIntegral parameter adjustment amount delta T_iDifferential parameter adjustment amount Δ K_dProportional coefficient K of incomplete differential PID controller_pIntegral time coefficient T_iDifferential coefficient K_dOnline modification is performed.

Outputting (x) to the support vector machine displacement prediction module actual displacement_a、x_b、y_a、y_b) Corresponding incomplete differential PID controller, cross feedback link, formed PID-cross feedback controller, radial displacement deviation signal (e)_ax、e_ay、e_bx、e_by) And obtaining cross feedback gain as a displacement cross feedback item of the PID-cross feedback controller.

The support vector machine displacement prediction module comprises 4 support vector machine radial displacement prediction modules and 1 support vector machine axial displacement prediction module, and a control current signal i of the front radial magnetic bearing a1_au、i_av、i_awInputting the first support vector machine radial displacement prediction module and outputting the radial displacement x_a(ii) a Control current signal i of front radial magnetic bearing a1_au、i_av、i_awInputting the radial displacement prediction module of the second support vector machine and outputting the radial displacement y_a(ii) a Rear radial directionControl current signal i of magnetic bearing b1_bu、i_bv、i_bwInputting the radial displacement prediction module of the third support vector machine and outputting the radial displacement x_b(ii) a Control current signal i of rear radial magnetic bearing b1_bu、i_bv、i_bwInputting the radial displacement prediction module of the fourth support vector machine and outputting the radial displacement y_b(ii) a Control current signal i of axial magnetic bearing a2_zAnd outputting the axial displacement z as the input of the axial displacement prediction module of the support vector machine.

The input of the improved zero power control module further comprises: multiplexing displacement x, y, z, theta_x、θ_yBy a small displacement p_sX, y, z, theta_x、θ_yBy the amount of displacement x_a、y_a、z、x_b、y_bAnd the coordinate system is obtained by conversion through a coordinate system conversion module A.

The invention has the beneficial effects that:

(1) according to the invention, a neural network inverse decoupling control module, a fuzzy PID cross feedback control module, a support vector machine displacement prediction module and an improved zero-power control module are linked through multiplexing of current signals, so that a comprehensive control system integrating high stability, low energy consumption and low cost is realized, the complexity of a magnetic suspension flywheel battery magnetic bearing rotor system is reduced, the real-time response speed of the magnetic suspension flywheel battery magnetic bearing rotor system is increased, and the vehicle-mounted flywheel battery can still keep high stability, low energy consumption and low cost running state control under the environment corresponding to the outside complex and changeable environment. The current signal processed by a current regulation module in a composite controlled object is taken as the input signal of a support vector machine displacement prediction module, the output signal of the support vector machine displacement prediction module, the output signal of a fuzzy PID cross feedback control module and a neural network inverse decoupling control module, and the signal processed by an accumulation and coordinate system conversion module is taken as the input signal of an improved zero-power control module, the displacement signal output by the support vector machine displacement prediction module and the output signal of the improved zero-power control module are taken as the input signals of a linear closed-loop controller, the output of the linear closed-loop controller is taken as the input signals of the fuzzy PID cross feedback control module and the neural network inverse decoupling control module, the output signal of the fuzzy PID cross feedback control module and the output signal of the neural network inverse decoupling control module are taken as the input signals of a composite controlled object through the coordinate system conversion module, the method avoids the situation that the control system is complex and the real-time response of the system is influenced due to the fact that the modules are piled up in the control system.

(2) The invention separately considers the influence of two factors of the driving state and road condition of the automobile on the vehicle-mounted magnetic suspension flywheel battery system, improves the stability of the system by different control methods, starts the corresponding control method under different working conditions by detecting the control current interval, is beneficial to reducing the system response time under a single control method, improves the control efficiency, reduces the control error rate and further improves the control precision. Aiming at the aspect of complex road conditions, a mathematical model of a magnetic bearing rotor system of the vehicle-mounted magnetic suspension flywheel battery aiming at five working conditions (turning, climbing, longitudinal vibration, transverse vibration and pitching vibration) is established through a prototype machine, a dynamic test and ADAMS simulation analysis, and compared with a dynamic model of the vehicle-mounted magnetic suspension flywheel battery aiming at eight working conditions (stable operation, starting acceleration, braking deceleration, turning, climbing, longitudinal vibration, transverse vibration and pitching vibration) in the prior art, the data volume is reduced; the method comprises the steps that five working conditions of turning, climbing, transverse vibration, longitudinal vibration and pitching vibration correspond to one PD controller respectively, when a control current detects a certain interval, the corresponding controller is started, and the stable operation of a flywheel rotor is realized by adopting neural network inverse decoupling control. Fuzzy PID cross feedback control is adopted for the self running state (stable running, starting acceleration and braking deceleration) of the automobile, the control parameters of the vehicle-mounted flywheel battery are adjusted on line in real time, the gyro effect is inhibited, and the stability of the vehicle-mounted magnetic suspension flywheel battery is greatly improved.

(3) The displacement prediction of the support vector machine is realized without special signal processing, the problems of overfitting of a neural network, easy falling into local extreme values, dependence of structural design on experience and the like are solved, the method is simple to operate, only current signals processed by a current warping module need to be introduced, and the simplicity of the system is improved.

(4) The improved zero-power control module regulates the small displacement through multiplexing the small displacement in the displacement, thereby improving the rigidity of the system; and a threshold value adjusting link is introduced, so that low-power-consumption control can be switched between improved zero-power-consumption control of the traditional zero-power control, low energy consumption is realized, and the anti-interference capability of the rotor is improved.

Drawings

FIG. 1 is a structural block diagram of a multidimensional integrated vehicle-mounted magnetic suspension flywheel battery control system;

FIG. 2 is a schematic structural diagram of an actual controlled object according to the present invention;

fig. 3(a) is a block diagram of a switching module according to the present invention, fig. 3(b) is a schematic diagram of a switching module according to a first PD controller of the present invention, and fig. 3(c) is a schematic diagram of a first PD controller according to the present invention;

FIG. 4 is a schematic diagram of the neural network inverse decoupling control of the present invention;

FIG. 5(a) is a block diagram of a fuzzy PID cross feedback control structure of the present invention, and FIG. 5(b) is a block diagram of a second fuzzy controller structure of the present invention;

fig. 6(a) is a schematic diagram of a displacement prediction module of a support vector machine of the present invention, fig. 6(b) is a schematic diagram of a radial displacement prediction module of a first support vector machine of the present invention, fig. 6(c) is a schematic diagram of a radial displacement prediction module of a third support vector machine of the present invention, fig. 6(d) is a schematic diagram of an axial displacement prediction module of a support vector machine of the present invention, and fig. 6(e) is a flowchart of a simplified particle swarm optimization algorithm of the present invention;

FIG. 7 is a block diagram of an improved zero power control module of the present invention;

in the figure: 1-linear closed-loop controller, 11-control mode switching module, 111-first PD controller switching module, 112-second PD controller switching module, 113-third PD controller switching module, 114-fourth PD controller switching module, 115-fifth PD controller switching module, 1111-first PD controller, 1112-second PD controller, 1113-third PD controller, 1114-fourth PD controller, 1115-fifth PD controller, 12-fuzzy PID cross-feedback control module, 121-first fuzzy controller, 122-second fuzzy controller, 123-third fuzzy controller, 124-fourth fuzzy controller, 125-fifth fuzzy controller, 131-first incomplete differential PID controller, 132-second incomplete differential PID controller, 133-a third incomplete differential PID controller, 134-a fourth incomplete differential PID controller, 135-a fifth incomplete differential PID controller, 14-a neural network inverse decoupling control module, 141-a fuzzy neural network inverse system, 15-an improved zero power control module, 151-an external force identification module, 152-a force-compensation displacement/angle module, 153-a memory module, 154-a low-pass filter, 2-a composite controlled object, 211-a first Clark inverse transformation module, 212-a second Clark inverse transformation module, 213-a third Clark inverse transformation module, 221-a first switch power amplifier, 222-a second switch power amplifier, 23-a current sensor, 241-a coordinate system conversion module, 242-an accumulation and coordinate system conversion module, 243-coordinate system conversion module C, 244-coordinate system conversion module A, 245-coordinate system conversion module B, 25-current warping module, 3-support vector machine displacement prediction module, 31-first support vector machine radial displacement prediction module, 311-first training test sample set module, 312-first data preprocessing module, 313-first determined optimal performance parameter module, 314-first support vector machine training module, 32-second support vector machine radial displacement prediction module, 33-third support vector machine radial displacement prediction module, 331-third training test sample set module, 332-third data preprocessing module, 333-third determined optimal performance parameter module, 334-third support vector machine training module, 34-fourth support vector machine radial displacement prediction module, 35-a support vector machine axial displacement prediction module, 351-a fifth training test sample set module, 352-a fifth data preprocessing module, 353-a fifth determined optimal performance parameter module, 354-a fifth support vector machine training module and 4-a vehicle-mounted magnetic suspension flywheel battery magnetic bearing rotor system mathematical model.

Detailed Description

The invention will be further described with reference to the following figures and specific examples, but the scope of the invention is not limited thereto.

As shown in fig. 1, the multidimensional integrated vehicle-mounted magnetic suspension flywheel battery control system includes a linear closed-loop controller 1, a fuzzy PID cross feedback control module 12, a neural network inverse decoupling control module 14, an improved zero power control module 15, a coordinate system conversion module 241, an accumulation and coordinate system conversion module 242, a composite controlled object 2 and a support vector machine displacement prediction module 3, where the linear closed-loop controller 1 includes a control mode switching module 11.

Displacement deviation signal e under magnetic bearing coordinate system_z、e_ax、e_ay、e_bx、e_byAs input, the second derivative phi of the displacement signal is output by the control mode switching module 11₁、φ₂、φ₃、φ₄、φ₅(i.e. the

)；φ₁、φ₂、φ₃、φ₄、φ₅Input to a neural network inverse decoupling control module 14, and output a control current signal i under a rotor centroid coordinate system_kx*、i_ky*、i_kz*、i_kθx*、i_kθyControl current signal i_kx*、i_ky*、i_kθxA and i_kθyThe radial control current signals { i ] aiming at the complex road condition under the magnetic bearing coordinate system are converted by the coordinate conversion module 241_kax*、i_kay*、i_kbx*、i_kby}; displacement deviation signal e_z、e_ax、e_ay、e_bx、e_byAnd the displacement deviation change rate corresponding to each displacement deviation signal is used as the input of the fuzzy PID cross feedback control module 12, and a control current signal { i ] aiming at the running state of the automobile under the magnetic bearing coordinate system is output_tz*、i_tax*、i_tay*、i_tbx*、i_tby}; control current signal i_kz*、i_kax*、i_kay*、i_kbx*、i_kby*}、{i_tz*、i_tax*、i_tay*、i_tbx*、i_tbyInput composite controlled object 2.

The composite controlled object module 2 comprises a first Clark inverse transformation module 211, a second Clark inverse transformation module 212, a third Clark inverse transformation module 213 and a first switchThe power amplifier 221, the second switching power amplifier 222, the current sensor 23, the current normalization module 25 and the actual controlled object, which is shown in fig. 2, includes an axial magnetic bearing a2, a front radial magnetic bearing a1 and a rear radial magnetic bearing b 1. Radial control current signal i in the composite controlled object 2_tax*、i_tayOutputs three-phase radial control current signals i for the front radial magnetic bearing a1 through a first Clark inverse transformation module 211_tau、i_tav、i_tawAnd detected by a current sensor 23 as { i }_tau、i_tav、i_taw}; radial control current signal i_kax*、i_kay*、i_kbx*、i_kbyOutputs three-phase radial control current signals i for the front radial magnetic bearing a1 through a second Clark inverse transformation module 212_kau、i_kav、i_kawAnd detected by a current sensor 23 as { i }_kau、i_kav、i_kaw}, radial control current signal { i_kax*、i_kay*、i_kbx*、i_kbyOutputs three-phase radial control current signals i for the rear radial magnetic bearing b1 through the second Clark inverse transformation module 212_kbu、i_kbv、i_kbwAnd detected by a current sensor 23 as { i }_kbu、i_kbv、i_kbw}; radial control current signal i_tbx*、i_tbyOutputs three-phase radial control current signals (i) aiming at a rear radial magnetic bearing b1 through a third Clark inverse transformation module 213_tbu、i_tbv、i_tbwAnd detected by a current sensor 23 as { i }_tbu、i_tbv、i_tbw}. In the composite controlled object 2, the axial control current signal i_tzOutputs an axial control current signal i for the axial magnetic bearing a2 through the first switching power amplifying module 221_tz(ii) a Axial control current signal i_kzOutputs an axial control current signal i for the axial magnetic bearing a2 through the second switching power amplifying module 222_kz。{i_tau、i_tav、i_taw}、{i_kau、i_kav、i_kaw}、{i_kbu、i_kbv、i_kbw}、{i_tbu、i_tbv、i_tbw}、i_tzAnd i_kzThe input current rectifying module 25 performs the following operations:

three-phase control current signals for the front radial magnetic bearing a 1:

i_au＝i_tau+i_kau，i_av＝i_tav+i_kav，i_aw＝i_taw+i_kaw

three-phase control current signals for the rear radial magnetic bearing b 1:

i_bu＝i_tbu+i_kbu，i_bv＝i_tbv+i_kbv，i_bw＝i_tbw+i_kbw

axial control current signal for the rear axial magnetic bearing a 2:

i_z＝i_tz+i_kz

will current signal i_au、i_av、i_aw、i_bu、i_bv、i_bw、i_zAnd the input control mode switching module 11 is used for detecting a current interval corresponding to the current and switching the working condition and the mathematical model.

Will current signal i_au、i_av、i_aw、i_bu、i_bv、i_bw、i_zThe displacement signal z and x are output as the input of the support vector machine displacement prediction module 3_a、y_a、x_b、y_b。

Two types of control current signals i_kz*、i_kx*、i_ky*、i_kθx*、i_kθyX and i_tz*、i_tax*、i_tay*、i_tbx*、i_tbyVia the add and coordinate system conversion module 242, the following operations are performed:

(1) coordinate conversion function

The control current signal i of the magnetic bearing coordinate system_tz*、i_tax*、i_tay*、i_tbx*、i_tbyLinear conversion to control current signal i under rotor centroid coordinate system_tz*、i_tx*、i_ty*、i_tθx*、i_tθy*；

(2) Accumulation function

The five-freedom control current signal under the rotor mass center coordinate system is as follows:

control current signal i_x*、i_y*、i_z*、i_θx*、i_θyAnd displacement signals z, x_a、y_a、x_b、y_bThe reference displacement signals z x and x are output together as the input of the improved zero power control module 15_a*、y_a*、x_b*、y_bOutput reference displacement signals z x_a*、y_a*、x_b*、y_bActual displacement signals z and x output by displacement prediction module of sum support vector machine_a、y_a、x_b、y_bAnd the linear closed-loop controller 1 is used as an input to form a comprehensive control system for closing the vehicle-mounted magnetic suspension flywheel battery.

The specific implementation of the invention comprises the following steps:

establishing a mathematical model 4 of a magnetic bearing rotor system of a vehicle-mounted magnetic suspension flywheel battery:

(1) the states and parameter change conditions of the automobile under different rotating speeds under five working conditions of turning, climbing, transverse vibration, longitudinal vibration and pitching vibration under different speeds are analyzed by using a prototype machine, a dynamic test and ADAMS simulation.

As shown in figure 2, the five-degree-of-freedom flywheel battery model is adopted, O is the mass center of the rotor, A is the axis of a front radial bearing a1, B is the axis of a rear radial bearing B1, and an O-xyz three-dimensional coordinate system is established by taking O as an original point. l_aIs the distance from point A to the centroid O, l_bThe distance from point B to centroid O, the distance from point a to point B, and the rotor mass assumed to be m, J (═ J)_x＝J_y) Is the moment of inertia of the rotor about the x-axis and y-axis, J_zFor the moment of inertia of the rotor around the z-axis, at any moment, under the magnetic bearing coordinate system, the displacement of the rotor at the point A and the point B relative to the stable position along the x-axis and the y-axis is set as x_a、x_b、y_a、y_bThe displacement of the rotor in the z-axis direction is z, the displacement of the centroid is x and y, and the angles of the rotor around the x-axis and the y-axis are theta_x、θ_y。

Displacement of rotor centroid O:

where ω is the rotor rotational angular velocity, F_Ax、F_BxElectromagnetic forces in the x-direction, F, of a front radial magnetic bearing a1 and a rear radial magnetic bearing b1, respectively_Ay、F_ByElectromagnetic forces in the y-direction, F, of a front radial magnetic bearing a1 and a rear radial magnetic bearing b1, respectively_zFor the electromagnetic force of the axial magnetic bearing a2 in the z direction, f_x、f_y、f_zIs a disturbing force. When the rotor has the gyro effect, a coupling term is generated in the motion equation (3)))

And

x, y, z, theta_x、θ_yThe second derivative of (a). f. of_x、f_y、f_zIs variable, and different values are taken under different working conditions (turning, ascending and descending, longitudinal vibration, transverse vibration and pitching vibration), so that f needs to be obtained by utilizing a prototype machine, a dynamic test and ADAMS simulation_x、f_y、f_zAnd determining mathematical models of the vehicle-mounted magnetic suspension flywheel battery magnetic bearing rotor system of the automobile under five working conditions of turning, climbing, transverse vibration, longitudinal vibration and pitching vibration according to different values.

(2) The invention relates to links of multiple coordinate transformation, linear amplification and anti-interference, which have certain influence on a mathematical model of a magnetic bearing rotor system of a vehicle-mounted magnetic suspension flywheel battery, and the mathematical model of a composite controlled object is determined by taking the mathematical model into consideration:

in the formula, κ₁、κ₂、κ₃、κ₄、κ₅The deviation value is obtained through simulation, and is a deviation value caused by multiple coordinate transformation, linear amplification and anti-interference links.

Under different working conditions, the control current is different in magnitude, and different working conditions are distinguished according to the magnitude of the control current. Control currents under different working conditions are analyzed by using a prototype machine and ADAMS in a simulation mode, as shown in table 1, when the control currents are detected to be within a certain interval, a corresponding mathematical model is started, and a control method for improving stability is adopted.

TABLE 1 control ammeter for different working conditions

As shown in fig. 3(a), the control mode switching module 11 includes five PD controller switching modules, which are a first PD controller switching module 111, a second PD controller switching module 112, a third PD controller switching module 113, a fourth PD controller switching module 114, and a fifth PD controller switching module 115; displacement deviation signal e under magnetic bearing coordinate system_ax、e_ay、e_bx、e_by、e_zThe displacement deviation signal e is converted into a displacement deviation signal e under a rotor centroid coordinate system through a coordinate system conversion module C243_x、e_y、

e_zStarting the process; each PD controller switching module comprises 5 PD controllers which respectively correspond to five working conditions of turning, climbing, transverse vibration, longitudinal vibration and pitching vibration and convert current signals { i } i_au、i_av、i_aw、i_bu、i_bv、i_bw、i_zIntroducing each PD controller switching module, and starting a corresponding PD controller and a composite controlled object mathematical model (formula (4)) when detecting that the current is in a certain interval (table 1).

As shown in fig. 3(b), taking the first PD controller switching module 111 as an example, the first PD controller switching module 111 includes 5 PD controllers (respectively, the first PD controller 1111, the second PD controller 1112, the third PD controller 1113, the fourth PD controller 1114, and the fifth PD controller 1115) respectively corresponding to five conditions, namely, turning, climbing, lateral vibration, longitudinal vibration, and pitching vibration. Each PD controller adopts a self-adaptive fuzzy control algorithm and a displacement deviation signal e_xAnd the displacement deviation change rate de/dt is used as the input of the PD controller to output the second derivative phi of the displacement signal₁. As shown in fig. 3(c), taking the first PD controller 1111 as an example, the scaling factor k is adjusted on line under mathematical models of different working conditions according to the fuzzy control principle_pAnd a differential coefficient k_dTo adapt to different working conditions, realize better control on the magnetic bearing rotor system and output x second derivative phi₁。

Aiming at complex road conditions (turning, ascending and descending, longitudinal vibration, transverse vibration and pitching vibration), the neural network inverse decoupling control module 14 is introduced, and the neural network inverse decoupling control module 14 comprises a neural network inverse system 141. As shown in fig. 4, the construction of the fuzzy neural network inverse system 141:

to be provided with

As the state variable of the composite controlled object module 2, the relative order a ═ a (a) of the system is obtained₁，a₂，a₃，a₄，a₅)^T＝(2，2，2，2，2)^TCan derive

The composite controlled object 2 is reversible when the reversible condition is satisfied.

Second derivative phi of displacement signal₁、φ₂、φ₃、φ₄、φ₅As the input of the fuzzy neural network inverse system 141, the output is the control current signal i_kx*、i_ky*、i_kz*、i_kθx*、i_kθyThe composite controlled object 2 is input through the coordinate system conversion module 241.

The construction of the fuzzy neural network inverse system 141 comprises the following specific steps:

step (1) constructing a fuzzy neural network training sample set

Fully exciting the composite controlled object 2 by random signals within the practical working allowable range, and receiving, converting and acquiring signals x, y, z and theta of the rotor in real time_x、θ_yThe differential method of five orders is used to obtain x, y, z, theta_x、θ_yCorresponding first derivative

And second derivative

For (x, y, z, theta)_x、θ_y) And the corresponding first derivative and second derivative are preprocessed to remove abnormal data, so that the reliability of the data is ensured, the static and dynamic characteristics of the composite controlled object can be fully reflected, and then the preprocessed data are normalized to finally form a fuzzy neural network training sample set.

Step (2) fuzzy neural network training

And (3) training the fuzzy neural network off line by adopting a BP algorithm of a self-adaptive learning rate, and selecting a four-layer neural network. The first layer of the fuzzy neural network is an input layer, and a fuzzy neural network training sample set is input; the second layer of the fuzzy neural network is a membership function input layer which is used for inputting a central value alpha, a width sigma and a back-piece parameter of the membership function;

the membership function is a gaussian function, and specifically includes:

in the formula, alpha is the central value of the membership function, and sigma is the width of the membership;

the third layer is a decision layer, and a fuzzy rule is formed to adjust the weight of the neural network on line according to the conditions of rotor displacement and interference of other factors (including temperature rise, vortex, rotating speed, dynamic load conditions and the like) under five working conditions of turning, climbing, transverse vibration, longitudinal vibration and pitching vibration of a prototype, a dynamic test and ADAMS simulation; the fourth layer is a de-blurring layer for performing a sharpening operation. After the neural network is constructed, the neural network is trained until the training error precision reaches 10^-3Output a control current i_kx*、i_ky*、i_kz*、i_kθx*、i_kθyA first step of; via a coordinate transformation module 241, i_kx*、i_ky*、i_kθx*、i_kθyConversion to i_kax*、i_kay*、i_kbx*、i_kby*。

A pseudo linear system is formed by the neural network inverse decoupling control module 14, the composite controlled object 2 and the support vector machine displacement prediction module 3, the pseudo linear system is equivalent to five mutually independent second-order linear systems, and linear decoupling of the composite controlled object is achieved. As shown in fig. 4.

For the other three conditions (smooth running, start-up acceleration, brake deceleration), the fuzzy PID cross feedback control module 12 is started when the control current is detected in a certain interval (table 1). Aiming at the self running state (stability) of the automobileRun, start acceleration, brake deceleration), introducing a fuzzy PID cross feedback control module 12, as shown in fig. 5(a), the fuzzy PID cross feedback control module 12 includes 5 fuzzy controllers (respectively, a first fuzzy controller 121, a second fuzzy controller 122, a third fuzzy controller 123, a fourth fuzzy controller 124, and a fifth fuzzy controller 125) and 5 incomplete differential PID controllers (respectively, a first incomplete differential PID controller 131, a second incomplete differential PID controller 132, a third incomplete differential PID controller 133, a fourth incomplete differential PID controller 134, and a fifth incomplete differential PID controller 135), displacement signals z, x of five degrees of freedom_a、y_a、x_b、y_bA fuzzy controller and an incomplete differential PID controller are connected in series in sequence. And (3) taking the displacement deviation signal e and the displacement deviation change rate de/dt as the input of a fuzzy controller, and outputting three control parameter regulating variables of PID (proportion integration differentiation) through fuzzification processing, fuzzy reasoning and defuzzification: proportional parameter adjustment Δ K_pIntegral parameter adjustment amount delta T_iDifferential parameter adjustment amount Δ K_dAnd then for three parameters of PID: coefficient of proportionality K_pIntegral time coefficient T_iDifferential coefficient K_dOnline modification is performed. The fuzzy control process of the displacement signal with five degrees of freedom is similar, and the embodiment uses x_aFor example, the fuzzy PID control is shown in fig. 5 (b):

the transfer function of the second incomplete differential PID controller 132 is:

in the formula: k_pIs a proportionality coefficient, T_iAs integral time coefficient, K_dIs a differential coefficient, T_fAre the differential filter time coefficients.

The fuzzy controller 122 includes fuzzification (input quantity), fuzzy inference and clarification (output quantity).

Input quantity e_ax、de_axDt and output Δ K_p、ΔT_i、ΔK_dDetermining, by obfuscating, a corresponding obfuscated subset:

input quantity fuzzy subset: e.g. of the type_ax、de_ax/dt＝{NB NM NS ZO PS PM PB}

Output quantity fuzzy subset: Δ K_p、ΔT_i、ΔK_d＝{NB NM NS ZO PS PM PB}

The distribution of membership functions for the fuzzy subsets is shown in table 2:

TABLE 2 fuzzy subset membership function distribution

The fuzzy control rule is an important basis of fuzzy reasoning and directly influences the performance of fuzzy control. PID parameter control adjustment test is carried out in a laboratory, and experience is summarized in e_ax、de_axThe regulation rules of three parameters of PID under different combination conditions are established to form fuzzy rule and fuzzy control table, and a fuzzy condition statement (if then structure) e is used_ax、de_axAnd/dt is combined by using and statements.

Determining delta K by combining debugging experience and practical tests_p、ΔT_i、ΔK_dControl rules for three parameters, as shown in table 3:

table 3 output quantity control rule table

The output quantity is clarified by a gravity center method.

The output is input to a second partial differential PID controller 132 to adjust the proportionality coefficient K in real time_pIntegral time coefficient T_iDifferential coefficient K_dOutput a control current signal i_taxAnd the stability of the magnetic suspension flywheel battery system is ensured under the self-running state of the automobile.

The gyro effect of the axial z of the magnetic suspension flywheel battery is analyzed by a prototype machine and simulation, the influence on the stability of the system is small, and the gyro effect can not be considered, so that the invention only aims at displacement signals (x) of the other four degrees of freedom_a、x_b、y_a、y_b) The corresponding incomplete differential PID controller introduces a cross feedback link to establish a relation so as to inhibit a gyro effect, and finally outputs a control current i_tax*、i_tay*、i_tbx*、i_tbyThe degree of freedom z directly outputs i through a first incomplete differential PID controller 131_tz*。

The PID-cross feedback controller formed after the cross feedback link is introduced is established as follows:

radial displacement deviation signal (e)_ax、e_ay、e_bx、e_by) And the displacement cross feedback item is used as a displacement cross feedback item of the PID-cross feedback controller to obtain cross feedback gain, and the coupling item is eliminated to inhibit the gyro effect.

According to the mass matrix, the current matrix and the gyro effect matrix of the rotor, determining the feedback gain k of the gyro effect complete compensation as follows:

in the formula: j. the design is a square_zIs the polar moment of inertia, k, of the rotor_iAnd omega is the rotor rotation angular speed.

Because complete compensation cannot be realized due to the phase lag of the magnetic bearing rotor system, disturbance interference error and the like, the attenuation factor C and the actual feedback gain k are introduced_vcThe appropriate C is selected by actual tuning.

As shown in fig. 6(a), the support vector machine displacement prediction module 3 is configured as follows:

the support vector machine displacement prediction module 3 comprises 4 support vector machine radial displacement prediction modules (a first support vector machine radial displacement prediction module 31, a second support vector machine radial displacement prediction module 32, a third support vector machine radial displacement prediction module 33 and a fourth support vector machine radial displacement prediction module respectivelyA vector machine radial displacement prediction module 34) and 1 support vector machine axial displacement prediction module 35. Control current signal i of front radial magnetic bearing a1_au、i_av、i_awInputting the first support vector machine radial displacement prediction module 31, outputting the radial displacement x_a(ii) a Control current signal i of front radial magnetic bearing a1_au、i_av、i_awInputting the radial displacement prediction module 32 of the second support vector machine and outputting the radial displacement y_a(ii) a Control current signal i of rear radial magnetic bearing b1_bu、i_bv、i_bwInputting the radial displacement prediction module 33 of the third support vector machine, and outputting the radial displacement x_b(ii) a Control current signal i of rear radial magnetic bearing b1_bu、i_bv、i_bwInputting the radial displacement prediction module 34 of the fourth support vector machine and outputting the radial displacement y_b(ii) a Control current signal i of axial magnetic bearing a2_zThe axial displacement z is output as an input to the support vector machine axial displacement prediction module 35.

As shown in fig. 6(b), the first support vector machine radial displacement prediction module 31 is formed by sequentially connecting a first training test sample set module 311, a first data preprocessing module 312, a first optimal performance parameter determining module 313 and a first support vector machine training module 314 in series. Control currents i of front radial magnetic bearing a1_au、i_av、i_awThe first training test sample set module 311 is inputted, the first support vector machine training module 314 outputs the radial displacement x_a(ii) a Control currents i of the front radial magnetic bearing a1 for the second support vector machine radial displacement prediction module 32_au、i_av、i_awInputting the training test sample set module, and outputting the radial displacement y by the training module of the support vector machine_a。

As shown in fig. 6(c), the third support vector machine radial displacement prediction module 33 is formed by sequentially connecting a third training test sample set module 331, a third data preprocessing module 332, a third optimal performance parameter determining module 333, and a third support vector machine training module 334 in series. Control currents i of rear radial magnetic bearing b1_bu、i_bv、i_bwInput a third training test sample set modelAt block 331, the third SVM training module 334 outputs a radial displacement x_b(ii) a For the fourth support vector machine radial displacement prediction module 34, the control current i of the rear radial magnetic bearing a1_bu、i_bv、i_bwInputting the training test sample set module, and outputting the radial displacement y by the training module of the support vector machine_b。

As shown in fig. 6(d), the support vector machine axial displacement prediction module 35 is formed by sequentially connecting a fifth training test sample set module 351, a fifth data preprocessing module 352, a fifth optimal performance parameter determining module 353, and a fifth support vector machine training module 354 in series. Control current i of axial magnetic bearing a2_zInput to fifth training test sample set module 351 and fifth support vector machine training module 354 outputs axial displacement z.

Because the components and detection modes of the first support vector machine radial displacement prediction module 31, the second support vector machine radial displacement prediction module 32, the third support vector machine radial displacement prediction module 33, the fourth support vector machine radial displacement prediction module 34 and the support vector machine axial displacement prediction module 35 with five degrees of freedom are similar, the invention specifically explains by taking the first support vector machine radial displacement prediction module 31 as an example; the method specifically comprises the following steps:

step (1), collecting a sample

Continuously collecting control current i of the front radial magnetic bearing a1_au、i_av、i_awThe data of the initial sample set is inputted into the first training test sample set module 311 to form an initial sample set I_a1＝{i_a1u、i_a1v、i_a1wIn which i_a1u＝{i_a1u1、i_a1u2、i_a1u3、…i_a1uM、i_a1uM+1、i_a1uM+2、…i_a1uM+M}，i_a1v＝{i_a1v1、i_a1v2、i_a1v3、…i_a1vM、i_a1vM+1、i_a1vM+2、…i_a1vM+M}，i_a1w＝{i_a1w1、i_a1w2、i_a1w3、…i_a1wM、i_a1wM+1、i_a1wM+2、…i_a1wM+M}。

Step (2), initial sample set I_a1＝{i_a1u、i_a1v、i_a1wPreprocessing the data by a first data preprocessing module 312, removing abnormal data to ensure the reliability of the data, and normalizing the collected data to limit all values to [ -1, 1 [ -1 [ ]]To avoid orders of magnitude adding difficulty to subsequent data processing calculations. Randomly selecting M groups of data by sampling at equal intervals to form a training sample set I_a2＝{i_a2u、i_a2v、i_a2wIn which i_a2u＝{i_a2u1、i_a2u2、i_a2u3、…i_a2uM}，i_a2v＝{i_a2v1、i_a2v2、i_a2v3、…i_a2vM}，i_a2w＝{i_a2w1、i_a2w2、i_a2w3、…i_a2wM}, training sample set I_a2The method is used for training the LS-SVM displacement prediction model of the least square support vector machine, and the rest M groups of data form a test sample set for testing the precision of the LS-SVM displacement prediction model.

Step (3), establishing an LS-SVM displacement prediction model based on particle swarm optimization

Using non-linear mapping

Will train sample set I_a2＝{i_a2u、i_a2v、i_a2wMapping samples of the LS-SVM displacement prediction model in a specific space into a feature space from an original space, wherein the displacement prediction model of the LS-SVM displacement prediction model in the specific space is as follows:

in the formula: w is a weight vector, b₁Is an offset value, i_a2The control current value of the front radial magnetic bearing a1 detected in real time, and y is the actual output value of the rotor displacement.

Defining optimization problem finding w and b₁：

Where J is the minimization objective function, γ is the regularization parameter, e_nAs fitting error, i_a2nFor training sample set I_a2N sample of (1), y_nAnd the actual output value of the rotor displacement is the nth sample.

Solving the optimization problem by using a Lagrangian function:

in the formula: alpha is alpha_nLagrange multipliers.

According to the Kuntta-k KKT condition, the partial derivative of the formula (10) is calculated and made equal to zero, and w and b are calculated₁：

In order to obtain better fitting and predicting effects, the LS-SVM displacement prediction model adopts a radial basis kernel function, and the expression is as follows:

in the formula:

is the nucleus width.

The module 313 for determining the first optimal performance parameter adopts a particle swarm optimization algorithm, and utilizes the strong parallel computation and global optimization capability of the particle swarm optimization algorithm to perform parameter pair

Performing automatic optimization to find an optimal set of parameters

The simplified particle swarm optimization flowchart as shown in fig. 6(e), wherein each sample represents a particle and the nth sample represents the nth particle.

The following operations are carried out:

(1) during initialization, setting parameters related to a particle swarm optimization algorithm: the final iteration time T of the algorithm is 100, and the acceleration factor c₁＝c ₂2, the optimum space range is [0, 1 ]]. Let t equal to 0, randomly get a set of parameters

And establishing an initial displacement prediction model according to the current parameter value.

(2) Calculating fitness F

And (3) taking the root mean square error between the model output value and the actual value as a fitness function, wherein the fitness of the kth particle is as follows:

in the formula: m is the total number of training samples, y_nAnd

respectively, the actual value and the model predicted output value of the nth particle.

(3) Updating the particle position to obtain the current optimal particle position

Calculating the fitness of each particle according to equation (13), if present, to particle p_nThe current individual optimal value is more optimal, and the individual optimal position p is set_n(t) set the position of the current particle and update the individual optimum value. If the optimal value of the current whole individual is superior to the optimal values of all the individual optimal values, the global optimal position g_nAnd (t) is the position of the current particle, and the global optimal value is updated.

(4) Updating the particle position adopts a particle position updating formula:

h_n(t+1)＝ω₁h_n(t)+c₁ r₁(p_n(t)-h_n(t))+c₂ r₁(g_n(t)-h_n(t)) (14)

in the formula: setting the particle population size as M, h_n＝(h_n1、h_n2、…、h_nd) Is the position of the nth particle in space, p_n＝(p_n1、p_n2、…、p_nd) Global optimum position g for the experienced optimum position_n＝(g_n1、g_n2、…、g_nd) D is more than or equal to 1 and less than or equal to D, and D represents a D-dimensional space of the particle population; the position of the nth particle is h when the iteration number is t_n(t) the individual optimum position is p_n(t) global optimum position is g_n(t)；c₁、c₂Is an acceleration factor; r is₁Is [0, 1 ]]A random number within; when the iteration number is t +1, the particle position is h_n(t + 1); inertial weight ω₁According to the formula

And (4) determining.

(5) And (3) judging whether the iteration is stopped, stopping the iteration output result when the calculated optimal value is smaller than the preset convergence precision or the maximum iteration time T is 100, and returning to the step (2) if the calculated optimal value is not smaller than the preset convergence precision or the maximum iteration time T is 100, so that T is T + 1.

Will train sample set I_a2＝{i_a2u、i_a2v、i_a2wThe input first determined optimal performance parameter module 313 outputs a set of optimal performance parameters

Inputting the set of optimal performance parameters into the first SVM training module 314, and outputting a degree of freedom x of radial displacement_a。

Obtaining an LS-SVM displacement prediction model by Mercer conditions:

in the formula: k (i)_a2n，i_a2) Using the first determined optimumOptimal performance parameters output by the performance parameter module 313

Similarly, the second SVM radial displacement prediction modules output y for 32 respectively_aThe third SVM radial displacement prediction module 33 outputs x_bThe fourth SVM radial displacement prediction module 34 outputs y_bThe support vector machine axial displacement prediction module 35 outputs z, specifically:

in the above formula, i_b2nFor training sample set I_b2The nth sample of (1)_b2The radial control current value of the input rear radial magnetic bearing b1 is detected in real time; i.e. i_z2nFor training sample set I_z2The nth sample of (1)_z2The axial control current value of the input axial magnetic bearing a2 is detected in real time; k (i)_b2n，i_b2) Optimal performance parameters using the third determine optimal performance parameters module 333

K(i_z2n，i_z2) Determining optimality using a fifth determinationOptimal performance parameters for the energy parameter module 353

The configuration of the modified zero power control module 15 is shown in fig. 7:

the support vector machine displacement prediction module 3 outputs the displacement x under the magnetic bearing coordinate system_a、y_a、z、x_b、y_bThe displacement is converted into displacement x, y, z, theta under the centroid coordinate system by a coordinate system conversion module A244_x、θ_y(ii) a The control current output by the neural network inverse decoupling control module 14 and the fuzzy PID cross decoupling feedback control module 12 is processed by the accumulation and coordinate system conversion module 242 to obtain i_x*、i_y*、i_z*、i_θx*、i_θy*；x、y、z、θ_x、θ_yAnd i_x*、i_y*、i_z*、i_θx*、i_θyAs input to the modified zero power control module 15, multiplexing the displacement x, y, z, θ_x、θ_yBy a small displacement p_sThe input modified zero power control module 15 can adjust ps to increase the stiffness of the magnetic bearing rotor system. And a threshold value adjusting link is introduced to realize the optimization of the power consumption and the anti-interference capability of the rotor, the low-pass filter 154 is used for preventing high-frequency signals from interfering the improved zero-power control module 15, and the memory module 153 connected with the low-pass filter 154 in parallel is used for keeping data to the next discrete period. By identifying the external force in real time, the memory module 153 and the low pass filter 154 adjust the output F, which is obtained by the displacement (x, y, z) and the angle (theta) of the rotor core deflection through the force/compensation displacement angle identification module 152_x*、θ_y*)，x*、y*、z*、θ_xAnd theta_yThe displacement x is converted into the displacement x under the magnetic bearing coordinate system by the coordinate conversion module B245_a*、y_a*、z*、x_b*、y_b*。

The displacements x, y, z are similar, taking x as an example, the external force expression:

F₁＝(mp²-k_s)x-k_ii_x*+ξ₁p_s1 (20)

in the formula, F₁Interference force of x, m rotor mass, p differential operator, k_sIs a negative displacement coefficient, k_iIs the current coefficient, x is the rotor displacement, ξ, which varies in the x direction₁p_s1For introducing a small displacement p_s1The latter additional displacement coefficient.

Rotor core offset displacement expression:

in the formula, F₁Is F₁And passes to the output of low pass filter 154.

Compared with the traditional zero-power control, the rotor core offset displacement expression has the advantages that the denominator is increased by k_ip_sTerm, appropriately adjusting p_sThe stiffness of the improved zero power control module 15 is increased, thereby increasing the stiffness of the entire magnetic bearing rotor system.

Angle theta_x、θ_ySimilarly, the present embodiment takes θ_xFor example, the external force expression:

in the formula, F₄Is theta_xInterference force of theta_xIs theta_xAngle, xi, of the rotor varying in direction₄p_s4For introducing a small displacement p_s4Additional coefficient of displacement, mu, after₁Are angle-displacement transform coefficients.

The expression of the rotation angle of the rotor core is as follows:

in the formula, F₄Is F₄Passes to the output, μ, of the low pass filter 154₂Coefficient of displacement-angle transformation, χ is electricFlow-displacement-angle compensation factor.

Compared with the traditional zero-power control, the rotor core offset displacement expression has the advantages that the denominator is increased by x k_ip_s4Term, appropriately adjusting p_s4The stiffness of the improved zero power control module 15 is increased, thereby increasing the stiffness of the entire magnetic bearing rotor system.

On the basis of adjustable rigidity, a threshold value adjusting link is introduced to realize further optimization of anti-interference and power consumption reduction, and a transfer function of a step function is set (five degrees of freedom adjustment is similar to the degree of freedom x as an example):

in the formula, x₀The air gap is where the rotor is in the equilibrium position x-direction.

When the rotor is in the threshold range, the traditional zero power control is adopted, the current in the coil always keeps the current to vibrate in a small amplitude near zero, and the improved zero power control is adopted when the current exceeds the threshold range, so that the rigidity of the improved zero power device and the anti-interference capability of the rotor are improved.

The present invention is not limited to the above-described embodiments, and any obvious improvements, substitutions or modifications can be made by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A multidimensional integrated vehicle-mounted magnetic suspension flywheel battery control system is characterized by comprising a linear closed-loop controller (1), a fuzzy PID cross feedback control module (12), a neural network inverse decoupling control module (14), an improved zero power control module (15), a coordinate system conversion module (241), an accumulation and coordinate system conversion module (242), a composite controlled object (2) and a support vector machine displacement prediction module (3), wherein the linear closed-loop controller (1) comprises a control mode switching module (11); the output of the linear closed-loop controller (1) is respectively used as the input of a fuzzy PID cross feedback control module (12) and a neural network inverse decoupling control module (14), the output of the fuzzy PID cross feedback control module (12) and the output of the neural network inverse decoupling control module (14) through a coordinate system conversion module (241) are used as the input for controlling a composite controlled object (2), and are converted and multiplexed through the composite controlled object (2) and used as the input of a support vector machine displacement prediction module (3); the outputs of the fuzzy PID cross feedback control module (12) and the neural network inverse decoupling control module (14) are processed by an accumulation and coordinate system conversion module (242), and are multiplexed with the output of the support vector machine displacement prediction module (3) to be used as the input of an improved zero-power control module (15); the improved zero power control module (15) generates a reference displacement output which is used as the input of the linear closed-loop controller (1) together with the actual displacement output of the support vector machine displacement prediction module (3).

2. The vehicle-mounted magnetic levitation flywheel battery control system according to claim 1, wherein the composite controlled object (2) comprises a first Clark inverse transformation module (211), a second Clark inverse transformation module (212), a third Clark inverse transformation module (213), a first switching power amplifier (221), a second switching power amplifier (222), a current sensor (23), a current normalization module (25) and an actual controlled object comprising an axial magnetic bearing a2, a front radial magnetic bearing a1 and a rear radial magnetic bearing b 1;

the input of the first Clark inverse transformation module (211) is: radial control current signal { i) output by fuzzy PID cross feedback control module (12)_tax*、i_tay*}；

The inputs of the first switching power amplifier (221) are: radial control current signal i output by fuzzy PID cross feedback control module (12)_tz*；

The inputs of the third Clark inverse transformation module (213) are: radial control current signal { i) output by fuzzy PID cross feedback control module (12)_tbx*、i_tby*}；

The inputs of the second Clark inverse transform module (212) are: radial control current signal { i) output by the neural network inverse decoupling control module (14)_kax*、i_kay*、i_kbx*、i_kby*}；

The second switching power amplifier (222) has inputs: radial control current signal i output by neural network inverse decoupling control module (14)_kz*；

The outputs of the first Clark inverse transformation module (211), the first switching power amplifier (221), the third Clark inverse transformation module (213), the second Clark inverse transformation module (212) and the second switching power amplifier (222) pass through a current sensor (23) and then are input into a current normalization module (25); the current signal { i ] is obtained after the current normalization module (25) processes the current signal_au、i_av、i_aw、i_bu、i_bv、i_bw、i_zThe displacement is used as the input of a support vector machine displacement prediction module (3);

i is described_zAs input to the axial magnetic bearing a2, { i_av、i_aw、i_auAs input to a front radial magnetic bearing a1, the i_bv、i_bw、i_buAs input to a rear radial magnetic bearing b 1.

3. Vehicle mounted magnetic levitation flywheel battery control system according to claim 2, wherein i is_au＝i_tau+i_kau、i_av＝i_tav+i_kav、i_aw＝i_taw+i_kaw、i_bu＝i_tbu+i_kbu、i_bv＝i_tbv+i_kbv、i_bw＝i_tbw+i_kbw、i_z＝i_tz+i_kzThe said i_tau、i_tav、i_tawIs the output of a first Clark inverse transform module (211), said i_kau、i_kav、i_kawIs the output of a second Clark inverse transform module (212), said i_tbu、i_tbv、i_tbwIs the output of a third Clark inverse transform module (213), said i_kbu、i_kbv、i_kbwIs the output of a second Clark inverse transform module (212), said i_tzIs the output of a first switching power amplifying module (221), said i_kzIs the output of the second switching power amplification module (222).

4. The vehicle-mounted magnetic levitation flywheel battery control system as recited in claim 1, wherein the accumulation and coordinate system conversion module (242) inversely decouples the current signal { i ] output by the control module (14) to the neural network_kz*、i_kx*、i_ky*、i_kθx*、i_kθyAnd a current signal { i } output by the fuzzy PID cross feedback control module (12)_tz*、i_tax*、i_tay*、i_tbx*、i_tbyProcessing as follows:

will current signal i_tz*、i_tax*、i_tay*、i_tbx*、i_tbyLinear conversion to control current signal i in centroid coordinate system_tx*、i_ty*、i_tz*、i_tθx*、i_tθy*；

5. the vehicle-mounted magnetic suspension flywheel battery control system according to claim 1, characterized in that the control mode switching module (11) comprises 5 PD controller switching modules, the PD controller switching module comprises 5 PD controllers, and the 5 PD controllers respectively correspond to five working conditions of turning, climbing, transverse vibration, longitudinal vibration and pitching vibration; the mathematical model of the magnetic suspension flywheel battery under the five working conditions is as follows:

where m is the rotor mass, J is the moment of inertia of the rotor about the x-axis and the y-axis, and J ═ J_x＝J_y，J_zIs the moment of inertia of the rotor about the z-axis, ω is the rotor rotational angular velocity, F_Ax、F_BxElectromagnetic forces in the x-direction, F, of a front radial magnetic bearing a1 and a rear radial magnetic bearing b1, respectively_Ay、F_ByElectromagnetic forces in the y-direction, F, of a front radial magnetic bearing a1 and a rear radial magnetic bearing b1, respectively_zFor the electromagnetic force of the axial magnetic bearing a2 in the z direction, f_x、f_y、f_zFor disturbing forces,/_aThe distance from the axle center of the front radial bearing to the mass center O, l_bThe distance from the axle center of the rear radial bearing to the mass center O,

and

to couple terms, k₁、κ₂、κ₃、κ₄、κ₅The method comprises the following steps of obtaining a deviation value caused by multiple coordinate transformation, linear amplification and anti-interference links through simulation; the PD controller adopts a self-adaptive fuzzy control algorithm, a displacement deviation signal e and a displacement deviation change rate de/dt are used as the input of the PD controller, and the PD controller outputs a second derivative phi of the displacement signal₁、φ₂、φ₃、φ₄、φ₅On-line adjustment of the proportionality coefficient k under the mathematical model of the magnetic suspension flywheel battery under different working conditions_pAnd a differential coefficient k_d。

6. The vehicle-mounted magnetic suspension flywheel battery control system according to claim 5, characterized in that the neural network inverse decoupling control module (14) comprises a neural network inverse system (141), and the construction of the fuzzy neural network inverse system (141) is specifically:

7. The vehicle-mounted magnetic suspension flywheel battery control system according to claim 5, characterized in that a fuzzy PID cross feedback control module (12) is introduced for three working conditions of smooth operation, start acceleration and brake deceleration, wherein the fuzzy PID cross feedback control module (12) comprises 5 fuzzy controllers and 5 incomplete differential PID controllers; the displacement deviation signal e and the displacement deviation change rate de/dt are used as the input of a fuzzy controller, and the proportional parameter regulating quantity delta K is output through fuzzification processing, fuzzy reasoning and defuzzification_pIntegral parameter adjustment amount delta T_iDifferential parameter adjustment amount Δ K_dProportional coefficient K of incomplete differential PID controller_pIntegral time coefficient T_iDifferential coefficient K_dOnline modification is performed.

8. Vehicle mounted magnetic levitation flywheel battery control system according to claim 7, characterized in that the actual displacement output (x) to the support vector machine displacement prediction module (3) is_a、x_b、y_a、y_b) Corresponding incomplete differential PID controller, cross feedback link, formed PID-cross feedback controller, radial displacement deviation signal (e)_ax、e_ay、e_bx、e_by) And obtaining cross feedback gain as a displacement cross feedback item of the PID-cross feedback controller.

9. The vehicle mounted magnetic levitation flywheel battery control system according to claim 1, wherein the support vector machine displacement prediction module (3) comprises 4 support vector machine radial displacement prediction modules and 1 support vector machine axial displacement prediction module, the control current signal i of the front radial magnetic bearing a1_au、i_av、i_awThe radial displacement prediction module (31) of the first support vector machine is input, and the radial displacement x is output_a(ii) a Control current signal i of front radial magnetic bearing a1_au、i_av、i_awInputting the radial displacement prediction module (32) of the second support vector machine and outputting the radial displacement y_a(ii) a Control current signal i of rear radial magnetic bearing b1_bu、i_bv、i_bwInputting the radial displacement prediction module (33) of the third support vector machine and outputting the radial displacement x_b(ii) a Control current signal i of rear radial magnetic bearing b1_bu、i_bv、i_bwInputting the radial displacement prediction module (34) of the fourth support vector machine and outputting the radial displacement y_b(ii) a Control current signal i of axial magnetic bearing a2_zThe axial displacement z is output as an input of a support vector machine axial displacement prediction module (35).

10. The on-board magnetic levitation flywheel battery control system as recited in claim 9, wherein the input of the modified zero power control module (15) further comprises: multiplexing displacement x, y, z, theta_x、θ_yBy a small displacement p_sX, y, z, theta_x、θ_yBy the amount of displacement x_a、y_a、z、x_b、y_bThe coordinate system is converted by a coordinate system conversion module A (244).