阅读说明：本技术 基于神经网络的压缩机控制方法、系统和存储介质 (Compressor control method, system and storage medium neural network based ) 是由 齐冀龙 杨阳 徐世文 于 2019-07-19 设计创作，主要内容包括：本发明提供一种基于神经网络的压缩机控制方法、系统和存储介质,本发明针对滚动转子压缩机无位置传感器情况下,由于负载不均衡,在每个周期的运动中会有周期性的转矩波动,在复杂工况下引入神经网络算法,通过在目标转速与反馈转速偏差后加入神经网络调节器以进行转速自调整,优化永磁同步电机的驱动性能,可以有效降低滚动转子压缩机由于负载变化而导致的转速波动,进而降低滚动转子压缩机的振动噪声,实现对滚动转子压缩机的自适应PI控制。(The present invention provides a kind of compressor control method neural network based, system and storage medium, the present invention is directed in the case of compressor with rolling rotor position-sensor-free, due to load imbalance, periodic torque ripple is had in the movement in each period, neural network algorithm is introduced under complex working condition, by the way that neural network adjuster is added to carry out revolving speed self-adjusting after rotating speed of target and feedback revolving speed deviation, optimize the driveability of permanent magnet synchronous motor, the fluctuation of speed caused by compressor with rolling rotor changes due to load can be effectively reduced, and then reduce the vibration noise of compressor with rolling rotor, it realizes and the adaptive PI of compressor with rolling rotor is controlled.)

1. a kind of compressor control method neural network based, which is characterized in that the described method includes:

Build permanent magnet synchronous motor position-sensor-free driving platform；

Permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to be acquired sample data；

Neural network is trained according to the sample data of acquisition；

After neural network convergence, neural network adjuster is transplanted in driven compressor system, to realize to der Geschwindigkeitkreis PI The adjusting of controller.

2. a kind of compressor control method neural network based according to claim 1, which is characterized in that by described in forever Magnetic-synchro electric machine without position sensor driving platform is placed under various operating conditions, to be acquired sample data, specifically further include:

Under preset various operating conditions, each input magnitude is acquired respectively and respectively exports magnitude, wherein various operating conditions are unloaded shape State, stress state, shock load, anticlimax load, analog compression machine load condition one or more；Each input magnitude is revolving speed Deviation, corner value, d shaft current, q shaft current, the revolving speed deviation are rotating speed of target and the difference for estimating revolving speed, and corner value is forever The rotor-position of magnetic-synchro motor estimation, d shaft current and q shaft current pass through Clark transformation and Park by a, b, c three-phase current Transformation obtains；Each output magnitude is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

3. a kind of compressor control method neural network based according to claim 1, which is characterized in that according to acquisition Sample data neural network is trained, specifically further include:

By it is positive calculate and backward learning process in neural network positive weight and threshold value be modified.

4. a kind of compressor control method neural network based according to claim 1, which is characterized in that nerve net After network convergence, neural network adjuster is transplanted in driven compressor system, specifically further include:

After neural network output convergence, neural network is embedded at der Geschwindigkeitkreis PI controller, the Kp of der Geschwindigkeitkreis PI controller It is determined by neural network according to the output of input quantity with Ki value；

It acquires a little sample datas again to be verified under simulation scenarios, to examine its rotary speed property under varying duty operating condition simultaneously The performance of neural network adjuster is verified, wherein the neural network adjuster uses BP neural network.

5. a kind of compressor control method neural network based according to claim 1, which is characterized in that building forever Magnetic-synchro electric machine without position sensor drives after platform, further includes:

Back-emf observer is constructed using permanent magnet synchronous motor back-emf equation；

It is observed obtaining back-emf observation according to back-emf observer；

Back-emf observation is filtered using preset filtering algorithm, to obtain filtered back-emf observation；

Based on filtered back-emf observation, the revolving speed and corner value of permanent magnet synchronous motor are estimated.

6. a kind of compressor control method neural network based according to claim 1, which is characterized in that the method Further include:

Motor stator biphase current i is measured by Hall current sensor_aAnd i_b, convert to obtain two-phase static coordinate by Clark Under electric current i_αAnd i_β；

Utilize i_α、i_βAnd u_αAnd u_β, estimate to obtain rotor Angle Position and motor speed by observer；

It estimates to obtain motor rotor position angle using observer, so that i_αAnd i_βAfter Park is converted, two cordic phase rotators are obtained Electric current i under system_dAnd i_q；

By motor given rotating speed ω^*Revolving speed is estimated with observerCompare, by speed pi regulator, export quadrature axis current to Determine i^* _q；

If direct-axis current is given as i^* _d=0, cross, straight shaft current respectively with it is actually detected value compared with, then respectively by electricity Pi regulator is flowed, cross, straight shaft voltage value u is exported_qAnd u_d, using Park inverse transformation, obtain the electricity under two-phase stationary coordinate system system Pressure value u_αAnd u_β；

Determine u_αAnd u_βResultant vector be located at some sector in six sectors that space voltage vector is surrounded, selection is appropriate Zero vector and calculate in the sector two adjacent voltage vectors and zero vector respective shared time, phase is set according to calculated result Answer register value, the driving control signal of output inverter.

7. a kind of compressor control method neural network based according to claim 1, which is characterized in that the permanent magnetism Synchronous motor is applied in compressor with rolling rotor, in compressor start and before entering closed loop phase, does not enable based on nerve net The control mode of network；After entering closed loop and stable operation, it is switched to neuron network PI controller director mode, the input of neural network Including revolving speed deviation, corner value, d shaft current and q shaft current, revolving speed deviation at this time is rotating speed of target and the difference for estimating revolving speed Value, corner value are the angular position of estimation, and d shaft current carries out transformation acquisition by estimation angular position with q shaft current；Neural network Output be der Geschwindigkeitkreis PI controller Kp and Ki value.

8. a kind of compressor control system neural network based, which is characterized in that the compressor control neural network based System processed includes: memory and processor, includes a kind of compressor control method journey neural network based in the memory Sequence, the compressor control method program neural network based realize following steps when being executed by the processor:

Build permanent magnet synchronous motor position-sensor-free driving platform；

Permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to be acquired sample data；

Neural network is trained according to the sample data of acquisition；

After neural network convergence, neural network adjuster is transplanted in driven compressor system, to realize to der Geschwindigkeitkreis PI The adjusting of controller.

9. a kind of compressor control system neural network based according to claim 8, which is characterized in that by described in forever Magnetic-synchro electric machine without position sensor driving platform is placed under various operating conditions, to be acquired sample data, specifically further include:

Under preset various operating conditions, each input magnitude is acquired respectively and respectively exports magnitude, wherein various operating conditions are unloaded shape State, stress state, shock load, anticlimax load, analog compression machine load condition one or more；Each input magnitude is revolving speed Deviation, corner value, d shaft current, q shaft current, the revolving speed deviation are rotating speed of target and the difference for estimating revolving speed, and corner value is forever The rotor-position of magnetic-synchro motor estimation, d shaft current and q shaft current pass through Clark transformation and Park by a, b, c three-phase current Transformation obtains；Each output magnitude is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

10. a kind of computer readable storage medium, which is characterized in that be based in the computer readable storage medium including one kind The compressor control method program of neural network, the compressor control method program neural network based are executed by processor When, the step of realizing a kind of compressor control method neural network based as described in any one of claims 1 to 7.

Technical field

The present invention relates to motor control technology field more particularly to a kind of compressor control method neural network based, System and storage medium.

Background technique

Due to the advantages that permanent magnetic synchronous motor structure is simple, small in size, power density is high, high-efficient, runnability is excellent, It has been widely used in recent years.Such as: field of traffic, field of household appliances, the drive area of car and boat, and pump class, compressor, The rotating mechanism of lathe, high accuracy servo system etc..

Permanent magnet synchronous motor driving at present generally uses space vector pulse width modulation, needs to obtain at any time in the art Rotor revolving speed and location information, due to being full of high temperature inside design feature inside compressor with rolling rotor, high pressure and strong The refrigerant of corrosion, inside can not installation site sensor, at present generally use position Sensorless Control mode.For cost And actual conditions consider to drive in many application fields using position-sensor-free mode.Such as: it is compressed in air-conditioning, refrigerator etc. Inside machine can not installation site sensor, generally use the vector control scheme of position-sensor-free in these application fields.Nothing The vector control scheme of position sensor mainly passes through back-emf observer and is observed to the back-emf of permanent magnet synchronous motor, The revolving speed and angular position of rotor are estimated with this.

In the case of compressor with rolling rotor position-sensor-free, since compressor with rolling rotor is inhaled during operation It gas side and calms the anger that there are pressure differences for side, causes compressor with rolling rotor often to rotate a circle in this way, periodic change can be presented in load Change, and then causes compressor rotary speed cyclic fluctuation inevitably occur, and compressor is shaken during operation The problems such as dynamic, noise.Fig. 1 is that compressor with rolling rotor loads the change procedure with mechanical corner in the process of running, due to rolling The uneven variation of the sub- torque in compressor of turn necessarily leads to vibration and the noise of compressor.How by suitably to torque side into Row control will improve compressor with rolling rotor runnability to reduce the fluctuation of speed.

With the development in terms of neural network in recent years, since neural network has in terms of non-linear and uncertain system There is greater advantage, for compressor assembly, if it is possible to neural network regulative mode is introduced, for improving rolling-rotor pressure The control performance of contracting machine, reducing its vibration and noise has preferable prospect.

Summary of the invention

In order to solve at least one above-mentioned technical problem, the invention proposes a kind of compressor controls neural network based Method, system and storage medium.

To achieve the goals above, first aspect present invention proposes a kind of compressor control side neural network based Method, which comprises

Build permanent magnet synchronous motor position-sensor-free driving platform；

Permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to be acquired sample number According to；

Neural network is trained according to the sample data of acquisition；

After neural network convergence, neural network adjuster is transplanted in driven compressor system, to realize to revolving speed The adjusting of ring PI controller.

In the present solution, permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to carry out Collecting sample data, specifically further include:

Under preset various operating conditions, each input magnitude is acquired respectively and respectively exports magnitude, wherein various operating conditions are zero load State, stress state, shock load, anticlimax load, analog compression machine load condition one or more；Each input magnitude is to turn Speed deviation, corner value, d shaft current, q shaft current, the revolving speed deviation are rotating speed of target and the difference for estimating revolving speed, and corner value is The rotor-position of permanent magnet synchronous motor estimation, d shaft current and q shaft current by a, b, c three-phase current by Clark transformation and Park converts to obtain；Each output magnitude is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

In the present solution, the sample data according to acquisition is trained neural network, specifically further include:

By it is positive calculate and backward learning process in neural network positive weight and threshold value be modified.

In the present solution, neural network adjuster is transplanted in driven compressor system, specifically after neural network convergence Further include:

After neural network output convergence, neural network is embedded at der Geschwindigkeitkreis PI controller, der Geschwindigkeitkreis PI controller Kp and Ki value determined by neural network according to the output of input quantity；

It acquires a little sample datas again to be verified under simulation scenarios, to examine its revolving speed under varying duty operating condition special Property and verify the performance of neural network adjuster, wherein the neural network adjuster uses BP neural network.

In the present solution, after building permanent magnet synchronous motor position-sensor-free driving platform, further includes:

Back-emf observer is constructed using permanent magnet synchronous motor back-emf equation；

It is observed obtaining back-emf observation according to back-emf observer；

Back-emf observation is filtered using preset filtering algorithm, to obtain filtered back-emf observation Value；

Based on filtered back-emf observation, the revolving speed and corner value of permanent magnet synchronous motor are estimated.

Further, the method also includes:

Motor stator biphase current i is measured by Hall current sensor_aAnd i_b, convert to obtain two-phase by Clark static Electric current i under coordinate_αAnd i_β；

Utilize i_α、i_βAnd u_αAnd u_β, estimate to obtain rotor Angle Position and motor speed by observer；

It estimates to obtain motor rotor position angle using observer, so that i_αAnd i_βAfter Park is converted, two-phase rotation is obtained Electric current i under coordinate system_dAnd i_q；

By motor given rotating speed ω^*Revolving speed is estimated with observerIt compares, by speed pi regulator, exports quadrature axis electricity Flow given i^* _q；

If direct-axis current is given as i^* _d=0, cross, straight shaft current respectively with it is actually detected value compared with, then pass through respectively Overcurrent pi regulator exports cross, straight shaft voltage value u_qAnd u_d, using Park inverse transformation, obtain under two-phase stationary coordinate system system Voltage value u_αAnd u_β；

Determine u_αAnd u_βResultant vector be located at some sector in six sectors that space voltage vector is surrounded, select Zero vector appropriate simultaneously calculates in the sector two adjacent voltage vectors and zero vector respective shared time, is set according to calculated result Determine corresponding registers value, the driving control signal of output inverter.

In the present solution, the permanent magnet synchronous motor is applied in compressor with rolling rotor, closed in compressor start and entrance Before loop order section, control mode neural network based is not enabled；After entering closed loop and stable operation, it is switched to neuron network PI controller Director mode, the input of neural network include revolving speed deviation, corner value, d shaft current and q shaft current, revolving speed deviation at this time For the difference of rotating speed of target and estimation revolving speed, corner value is the angular position of estimation, and d shaft current and q shaft current are by estimation corner Position carries out transformation acquisition；The output of neural network is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

Second aspect of the present invention also proposes a kind of compressor control system neural network based, described to be based on neural network Compressor control system include: memory and processor, include a kind of compressor neural network based in the memory Control method program realizes following step when the compressor control method program neural network based is executed by the processor It is rapid:

Build permanent magnet synchronous motor position-sensor-free driving platform；

Permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to be acquired sample number According to；

Neural network is trained according to the sample data of acquisition；

After neural network convergence, neural network adjuster is transplanted in driven compressor system, to realize to revolving speed The adjusting of ring PI controller.

In the present solution, permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to carry out Collecting sample data, specifically further include:

Under preset various operating conditions, each input magnitude is acquired respectively and respectively exports magnitude, wherein various operating conditions are zero load State, stress state, shock load, anticlimax load, analog compression machine load condition one or more；Each input magnitude is to turn Speed deviation, corner value, d shaft current, q shaft current, the revolving speed deviation are rotating speed of target and the difference for estimating revolving speed, and corner value is The rotor-position of permanent magnet synchronous motor estimation, d shaft current and q shaft current by a, b, c three-phase current by Clark transformation and Park converts to obtain；Each output magnitude is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

Third aspect present invention also proposes a kind of computer readable storage medium, wraps in the computer readable storage medium Include a kind of compressor control method program neural network based, the compressor control method program quilt neural network based When processor executes, realize such as the step of a kind of above-mentioned compressor control method neural network based.

The present invention is directed in the case of compressor with rolling rotor position-sensor-free, due to load imbalance, in each period Movement in have periodic torque ripple, neural network algorithm is introduced under complex working condition, by rotating speed of target and anti- It presents and neural network adjuster is added after revolving speed deviation carries out revolving speed self-adjusting, optimize the driveability of permanent magnet synchronous motor, it can be with The fluctuation of speed caused by compressor changes due to load is effectively reduced, and then reduces the vibration noise of compressor, realizes to rolling The adaptive PI of rotor compressor controls.

Additional aspect and advantage of the invention will provide in following description section, will partially become from the following description Obviously, or practice through the invention is recognized.

Detailed description of the invention

Fig. 1 shows the relational graph of compressor with rolling rotor load torque and mechanical corner position in the prior art；

Fig. 2 shows a kind of flow charts of compressor control method neural network based of the invention；

Fig. 3 shows a kind of structure chart of BP neural network of the present invention；

Fig. 4 shows a kind of Sigmoid function curve diagram of the present invention；

Fig. 5 shows a kind of structure chart of table dress formula permanent magnet synchronous motor of the present invention；

Fig. 6 shows permanent magnet synchronous motor position-sensor-free driving block diagram of the present invention；

Fig. 7 shows a kind of realization block diagram of neural network adjuster of the present invention；

Fig. 8 shows a kind of block diagram of compressor control system neural network based of the present invention.

Specific embodiment

To better understand the objects, features and advantages of the present invention, with reference to the accompanying drawing and specific real Applying mode, the present invention is further described in detail.It should be noted that in the absence of conflict, the implementation of the application Feature in example and embodiment can be combined with each other.

In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, still, the present invention may be used also To be implemented using other than the one described here other modes, therefore, protection scope of the present invention is not by described below Specific embodiment limitation.

Fig. 2 shows a kind of flow charts of compressor control method neural network based of the invention.

As shown in Fig. 2, first aspect present invention proposes a kind of compressor control method neural network based, the method Include:

S202 builds permanent magnet synchronous motor position-sensor-free driving platform；

Permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to be acquired by S204 Sample data；

S206 is trained neural network according to the sample data of acquisition；

Neural network adjuster is transplanted in driven compressor system by S208 after neural network convergence, with realization pair The adjusting of der Geschwindigkeitkreis PI controller.

It should be noted that position-sensor-free generally uses the model equation of voltage, electric current and magneto, lead to It crosses observer and estimates back-emf information, then release the angular position and rotary speed information of rotor by back-emf observation, as Feedback quantity constitutes closed-loop vector control system.By theory analysis and experiment detection at low speeds to back-emf observation, And estimation tachometer value be filtered can biggish raising system driveability.

According to an embodiment of the invention, permanent magnet synchronous motor position-sensor-free driving platform is placed in various operating conditions Under, to be acquired sample data, specifically further include:

Under preset various operating conditions, each input magnitude is acquired respectively and respectively exports magnitude, wherein various operating conditions are zero load State, stress state, shock load, anticlimax load, analog compression machine load condition one or more；Each input magnitude is to turn Speed deviation, corner value, d shaft current, q shaft current, the revolving speed deviation are rotating speed of target and the difference for estimating revolving speed, and corner value is The rotor-position of permanent magnet synchronous motor estimation, d shaft current and q shaft current by a, b, c three-phase current by Clark transformation and Park converts to obtain；Each output magnitude is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

According to an embodiment of the invention, the sample data according to acquisition is trained neural network, specifically further include:

By it is positive calculate and backward learning process in neural network positive weight and threshold value be modified.

According to an embodiment of the invention, neural network adjuster is transplanted to driven compressor after neural network convergence In system, specifically further include:

After neural network output convergence, neural network is embedded at der Geschwindigkeitkreis PI controller, der Geschwindigkeitkreis PI controller Kp and Ki value determined by neural network according to the output of input quantity；

It acquires a little sample datas again to be verified under simulation scenarios, to examine its revolving speed under varying duty operating condition special Property and verify the performance of neural network adjuster, wherein the neural network adjuster uses BP neural network.

It should be noted that neural network algorithm uses BP neural network, the sample of neural network uses permanent magnet synchronous electric Machine position-sensor-free drives platform acquisition.The operating condition of acquisition such as revolving speed increases or decreases suddenly, and torque increases or decreases suddenly, Or the operating condition of simulation compressor with rolling rotor, add periodically variable load appropriate etc..Adjust the PI ginseng of der Geschwindigkeitkreis Number, is optimal state.By the collecting sample under various operating conditions, off-line training is carried out to neural network, etc. neural networks it is defeated After restraining out, then verified by performance of the sample data to neural network.

In the specific implementation process, the driving of permanent magnet synchronous motor is realized using normal PI controller first, at this moment first It is trained by permanent magnet synchronous motor, since the target of the algorithm is compressor with rolling rotor, it should use position sensorless Device control algolithm, input speed deviation, corner value, d shaft current, q shaft current, as the input quantity of BP neural network, transfer Speed deviation is reference rotation velocity and the difference for estimating revolving speed, and corner value is the rotor-position of the estimation of magneto, d shaft current and q Shaft current can be converted by a, b, c three-phase current by Clark and Park converts to obtain, and the Kp of PI controller, Ki is as defeated Output, intermediate hidden layer use 4 units, and BP neural network adjuster is first not involved in control, by the various operating conditions being manually set, For example, light condition, stress state or shock load, the working condition acquirings such as anticlimax load and the load of analog compression machine are each defeated Enter magnitude, and obtains Kp the and Ki value of der Geschwindigkeitkreis of good performance；Neural network is trained by the sample of acquisition, is passed through Forward direction calculate and backward learning process in neural network positive weight and threshold value be modified, until neural network output convergence Neural network is embedded at der Geschwindigkeitkreis PI controller afterwards, the Kp of PI controller, Ki value is by neural network according to the defeated of input quantity It determines out, then acquires a little samples and verified under simulation scenarios, after verifying is good, controller is realized on hardware, to examine Its rotary speed property under varying duty operating condition and the performance for verifying neural network adjuster.It completes nerve net after above-mentioned work Network PI controller is added in compressor with rolling rotor drive system.

Specifically, the structure of BP neural network is as shown in figure 3, include input layer, hidden layer, output layer.

Activation primitive uses the following equation of sigmoid function:As shown in Figure 4.

The derivative of activation primitive are as follows: f ' (x)=f (1-f)；

The input of j-th of neuron node of hidden layer

Hin (j) is the input of j-th of measuring point of hidden layer in above formula, and Wji is that j-th of hiding node layer is directed toward in i-th of input Weight, bj be j-th of hidden layer neuron node bias.

Output layer and hiding interlayer also meet relationship: hout (j)=f (hin (j))；

Further,

Wherein, yin (j) is the input of j-th of output neuron node, and Vji is that the output of i-th of hiding node layer is directed toward The weight of j-th of output node.

Yout (j)=f (yin (j)), yout (j) are the output valve of j-th of node.

The error function of p-th of sample of performance index function are as follows:

Wherein r (j) is the target value of j-th of output, and yout (j) is the output valve of j-th of output.

The calculation formula of reversed adjustment weight are as follows:Wherein η For learning rate.

According to an embodiment of the invention, after building permanent magnet synchronous motor position-sensor-free driving platform, further includes:

Back-emf observer is constructed using permanent magnet synchronous motor back-emf equation；

It is observed obtaining back-emf observation according to back-emf observer；

Back-emf observation is filtered using preset filtering algorithm, to obtain filtered back-emf observation Value；

Based on filtered back-emf observation, the revolving speed and corner value of permanent magnet synchronous motor are estimated.

Preferably, the back-emf observer can be sliding mode observer, but not limited to this.

According to an embodiment of the invention, the formula of the filtering algorithm are as follows: e₀(k)=(1-a) × e₀(k-1)+a×e_i (k), wherein a is filter factor, and 0 < a < 1, e₀For filtered output, e_iFor the back-emf observation of non-filtered processing Or estimate the tachometer value of rotor.

It should be noted that the common RC low-pass filter of above-mentioned filtering algorithm formula simulation, by by the differential equation It turns to difference equation and obtains above-mentioned Filtering Formula.When a level off to 1 when, output with input tends to be equal, at this time without filter effect； When a level off to 0 when, filter effect is more apparent, by actual test, the value that general filter factor takes is smaller just has it is apparent Effect.Preferably, in actual test a < 0.05, it will obtain relatively good control effect.

It should be noted that the stator structure of permanent magnet synchronous motor (PMSM) is three-phase symmetric winding structure, in order to more preferable Analysis and control, need to establish permanent magnet synchronous motor mathematical model herein.

Fig. 5 is the structure chart that table fills formula permanent magnet synchronous motor.

As shown in Figure 5, three armature winding spatial distributions of permanent magnet synchronous motor, 120 electrical angle of axis mutual deviation, with A phase Winding axis is as stator stationary reference axis, and the magnetic direction that stator rotor permanent-magnet pole generates is d-axis (d axis), then along rotation The position of advanced 90 electrical angle of d-axis in direction be quadrature axis (q axis), and using rotor d-axis relative to stator A axis winding axis as Rotor position angle θ.

Obtain PMSM stator voltage equation formula:

In above formula, [u_a u_b u_c]^TFor stator phase voltage vector；[i_a i_b i_c]^TFor stator phase currents vector；[Ψ_a Ψ_b Ψ_c]^TFor stator phase winding flux linkage vector；diag[R_a R_b R_c] it is stator phase winding resistance diagonal matrix；P=d/dt is differential calculation Son.

Threephase stator electric current main function is the magnetic field for generating a rotation, therefore can be with binary system come equivalent, In This introduces rotation two-phase dq coordinate, from stator three-phase to shown in the transformation matrix such as formula (5-2) of rotation two-phase.

Permanent magnet synchronous motor is obtained in the voltage equation of dq shafting:

In above formula, u_d、u_qRespectively direct-axis voltage, quadrature-axis voltage；R is stator resistance；Ψ_d、Ψ_qRespectively d-axis magnetic linkage, Quadrature axis magnetic linkage；ω_eFor angular rate.

Wherein stator magnetic linkage has:

In formula (5-4), L_d、L_qRespectively stator d-axis inductance, stator axis inductor；Ψ_fFor permanent magnet flux linkage.

Formula (5-4) is substituted into formula (5-3), is obtained:

Due to Ψ_fFor constant, then formula (5-5) can simplify are as follows:

The torque equation of PMSM are as follows:

The equation of motion of PMSM are as follows:

In above formula, J is rotary inertia, T_eFor electromagnetic torque, T_LFor load torque, B is viscous friction coefficient, ω_rFor machinery Angular speed.

Formula permanent magnet synchronous motor (SPMSM) is filled for table, d-axis inductance and axis inductor are equal, i.e. L_d=L_q=L_s, then Torque equation described in formula (5-7) can be further simplified are as follows: Wherein, P_nFor motor number of pole-pairs, Ψ_fFor Permanent magnet magnetic potential, i_qFor quadrature axis current.

Fig. 6 shows permanent magnet synchronous motor position-sensor-free driving block diagram of the present invention.

As shown in fig. 6, permanent magnet synchronous motor position-sensor-free drives process following steps.

Motor stator biphase current i is measured by Hall current sensor_aAnd i_b, convert to obtain two-phase by Clark static Electric current i under coordinate_αAnd i_β；

Utilize i_α、i_βAnd u_αAnd u_β, estimate to obtain rotor Angle Position and motor speed by observer；

It estimates to obtain motor rotor position angle using observer, so that i_αAnd i_βAfter Park is converted, two-phase rotation is obtained Electric current i under coordinate system_dAnd i_q；

By motor given rotating speed ω^*Revolving speed is estimated with observerIt compares, by speed pi regulator, exports quadrature axis electricity Flow given i^* _q；

If direct-axis current is given as i^* _d=0, cross, straight shaft current respectively with it is actually detected value compared with, then pass through respectively Overcurrent pi regulator exports cross, straight shaft voltage value u_qAnd u_d, using Park inverse transformation, obtain under two-phase stationary coordinate system system Voltage value u_αAnd u_β；

Determine u_αAnd u_βResultant vector be located at some sector in six sectors that space voltage vector is surrounded, select Zero vector appropriate simultaneously calculates in the sector two adjacent voltage vectors and zero vector respective shared time, is set according to calculated result Determine corresponding registers value, the driving control signal of output inverter.

It should be noted that the level of torque of motor depends on i_dAnd i_qSize, i.e., control i_dAnd i_qMotor can be controlled Torque, since certain revolving speed and torque correspond to certain i^* _dAnd i^* _q, by the control to the two electric currents, make actual i_d And i_qTrace command value i^* _dAnd i^* _q, to just realize the control of motor torque and speed.

For three-phase permanent magnet synchronous motor, by the three-phase alternating current i for detecting armature winding_a、i_bAnd i_c, then pass through Coordinate transform obtains electric current i under rotation two-phase dq coordinate_dAnd i_q, in the process, need to use the location information of rotor, The present invention estimates back-emf using observer by the input voltage and electric current of detection motor, obtains rotor position by calculating Confidence breath.

Fig. 7 shows a kind of realization block diagram of neural network adjuster of the present invention.

As shown in fig. 7, the permanent magnet synchronous motor is applied in compressor with rolling rotor, closed in compressor start and entrance Before loop order section, control mode neural network based is not enabled using traditional approach and；After entering closed loop and stable operation, cut Change to neuron network PI controller director mode, the input of neural network includes revolving speed deviation, corner value, d shaft current and q shaft current, Revolving speed deviation at this time is rotating speed of target and the difference for estimating revolving speed, and corner value is the angular position of estimation, d shaft current and q axis Electric current carries out transformation acquisition by estimation angular position；The output of neural network is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

Fig. 8 shows a kind of block diagram of compressor control system neural network based of the present invention.

As shown in figure 8, second aspect of the present invention also proposes a kind of compressor control system neural network based 8, it is described Compressor control system 8 neural network based includes: memory 81 and processor 82, includes a kind of base in the memory 81 In the compressor control method program of neural network, the compressor control method program neural network based is by the processing Device 82 realizes following steps when executing:

Build permanent magnet synchronous motor position-sensor-free driving platform；

Permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to be acquired sample number According to；

Neural network is trained according to the sample data of acquisition；

After neural network convergence, neural network adjuster is transplanted in driven compressor system, to realize to revolving speed The adjusting of ring PI controller.

It should be noted that system of the invention can be operated in the terminal devices such as PC, mobile phone, PAD.

It should be noted that the processor can be central processing unit (Central Processing Unit, CPU), it can also be other general processors, Digital Signal Processing (Digital Signal Processor, DSP), dedicated collection At circuit (Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field- Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic, Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor Deng.

Further, permanent magnet synchronous motor position-sensor-free driving platform is placed under various operating conditions, to carry out Collecting sample data, specifically further include:

Under preset various operating conditions, each input magnitude is acquired respectively and respectively exports magnitude, wherein various operating conditions are zero load State, stress state, shock load, anticlimax load, analog compression machine load condition one or more；Each input magnitude is to turn Speed deviation, corner value, d shaft current, q shaft current, the revolving speed deviation are rotating speed of target and the difference for estimating revolving speed, and corner value is The rotor-position of permanent magnet synchronous motor estimation, d shaft current and q shaft current by a, b, c three-phase current by Clark transformation and Park converts to obtain；Each output magnitude is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

Further, neural network is trained according to the sample data of acquisition, specifically further include:

By it is positive calculate and backward learning process in neural network positive weight and threshold value be modified.

Further, after neural network convergence, neural network adjuster is transplanted in driven compressor system, specifically Further include:

After neural network output convergence, neural network is embedded at der Geschwindigkeitkreis PI controller, der Geschwindigkeitkreis PI controller Kp and Ki value determined by neural network according to the output of input quantity；

It acquires a little sample datas again to be verified under simulation scenarios, to examine its revolving speed under varying duty operating condition special Property and verify the performance of neural network adjuster, wherein the neural network adjuster uses BP neural network.

Further, after building permanent magnet synchronous motor position-sensor-free driving platform, further includes:

Back-emf observer is constructed using permanent magnet synchronous motor back-emf equation；

It is observed obtaining back-emf observation according to back-emf observer；

Back-emf observation is filtered using preset filtering algorithm, to obtain filtered back-emf observation Value；

Based on filtered back-emf observation, the revolving speed and corner value of permanent magnet synchronous motor are estimated.

Further, it is also realized when the compressor control method program neural network based is executed by the processor Following steps:

Motor stator biphase current i is measured by Hall current sensor_aAnd i_b, convert to obtain two-phase by Clark static Electric current i under coordinate_αAnd i_β；

Utilize i_α、i_βAnd u_αAnd u_β, estimate to obtain rotor Angle Position and motor speed by observer；

It estimates to obtain motor rotor position angle using observer, so that i_αAnd i_βAfter Park is converted, two-phase rotation is obtained Electric current i under coordinate system_dAnd i_q；

By motor given rotating speed ω^*Revolving speed is estimated with observerIt compares, by speed pi regulator, exports quadrature axis electricity Flow given i^* _q；

If direct-axis current is given as i^* _d=0, cross, straight shaft current respectively with it is actually detected value compared with, then pass through respectively Overcurrent pi regulator exports cross, straight shaft voltage value u_qAnd u_d, using Park inverse transformation, obtain under two-phase stationary coordinate system system Voltage value u_αAnd u_β；

Determine u_αAnd u_βResultant vector be located at some sector in six sectors that space voltage vector is surrounded, select Zero vector appropriate simultaneously calculates in the sector two adjacent voltage vectors and zero vector respective shared time, is set according to calculated result Determine corresponding registers value, the driving control signal of output inverter.

Further, the permanent magnet synchronous motor is applied in compressor with rolling rotor, closes in compressor start and entrance Before loop order section, control mode neural network based is not enabled；After entering closed loop and stable operation, it is switched to neuron network PI controller Director mode, the input of neural network include revolving speed deviation, corner value, d shaft current and q shaft current, revolving speed deviation at this time For the difference of rotating speed of target and estimation revolving speed, corner value is the angular position of estimation, and d shaft current and q shaft current are by estimation corner Position carries out transformation acquisition；The output of neural network is Kp the and Ki value of der Geschwindigkeitkreis PI controller.

Third aspect present invention also proposes a kind of computer readable storage medium, wraps in the computer readable storage medium Include a kind of compressor control method program neural network based, the compressor control method program quilt neural network based When processor executes, realize such as the step of a kind of above-mentioned compressor control method neural network based.

The present invention is directed in the case of compressor with rolling rotor position-sensor-free, due to load imbalance, in each period Movement in have periodic torque ripple, neural network algorithm is introduced under complex working condition, by rotating speed of target and anti- It presents and neural network adjuster is added after revolving speed deviation carries out revolving speed self-adjusting, optimize the driveability of permanent magnet synchronous motor, it can be with The fluctuation of speed caused by compressor changes due to load is effectively reduced, and then reduces the vibration noise of compressor.

In addition, the present invention passes through for the training of neural network adjuster first in permanent magnet synchronous motor position-sensor-free Driving platform be acquired sample, then neural network is trained offline according to the sample of acquisition, wait until neural network After convergence, neural network adjuster is added to der Geschwindigkeitkreis PI controller front end.The present invention must start in compressor with rolling rotor It is just switched to neural network adjuster state after into steady operational status, realizes the adaptive PI control to compressor with rolling rotor System.

In several embodiments provided herein, it should be understood that disclosed device and method can pass through it Its mode is realized.Apparatus embodiments described above are merely indicative, for example, the division of the unit, only A kind of logical function partition, there may be another division manner in actual implementation, such as: multiple units or components can combine, or It is desirably integrated into another system, or some features can be ignored or not executed.In addition, shown or discussed each composition portion Mutual coupling or direct-coupling or communication connection is divided to can be through some interfaces, the INDIRECT COUPLING of equipment or unit Or communication connection, it can be electrical, mechanical or other forms.

Above-mentioned unit as illustrated by the separation member, which can be or may not be, to be physically separated, aobvious as unit The component shown can be or may not be physical unit；Both it can be located in one place, and may be distributed over multiple network lists In member；Some or all of units can be selected to achieve the purpose of the solution of this embodiment according to the actual needs.

In addition, each functional unit in various embodiments of the present invention can be fully integrated in one processing unit, it can also To be each unit individually as a unit, can also be integrated in one unit with two or more units；It is above-mentioned Integrated unit both can take the form of hardware realization, can also realize in the form of hardware adds SFU software functional unit.

Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through The relevant hardware of program instruction is completed, and program above-mentioned can store in computer-readable storage medium, which exists When execution, step including the steps of the foregoing method embodiments is executed；And storage medium above-mentioned includes: movable storage device, read-only deposits Reservoir (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or The various media that can store program code such as CD.

If alternatively, the above-mentioned integrated unit of the present invention is realized in the form of software function module and as independent product When selling or using, it also can store in a computer readable storage medium.Based on this understanding, the present invention is implemented Substantially the part that contributes to existing technology can be embodied in the form of software products the technical solution of example in other words, The computer software product is stored in a storage medium, including some instructions are used so that computer equipment (can be with It is personal computer, server or network equipment etc.) execute all or part of each embodiment the method for the present invention. And storage medium above-mentioned includes: that movable storage device, ROM, RAM, magnetic or disk etc. are various can store program code Medium.

The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any Those familiar with the art in the technical scope disclosed by the present invention, can easily think of the change or the replacement, and should all contain Lid is within protection scope of the present invention.Therefore, protection scope of the present invention should be based on the protection scope of the described claims.