TWI638512B - Two-vector-based modeless predictive current control method for interior permanent magnet synchronous motor drive systems using prediction error correction technique - Google Patents

Abstract

The invention proposes a dual voltage vector modelless predictive current control rule using a prediction error correction technique, which can be applied to an embedded permanent magnet synchronous motor drive system. In one sampling period, two basic voltage vectors, which are sequentially referred to as the first voltage vector and the second voltage vector, are respectively applied and corresponding to the two-stage application time, which are sequentially referred to as the first application time and The second period of application time can be adjusted separately. The stator current difference produced by the first voltage vector is applied for the first period of time, referred to as the first stage current difference, and the second voltage vector is applied for the second period of time to produce the stator current difference. , called the second segment current difference. The first current difference can be calculated by two current sampling and subtraction operations. Then, using the proposed prediction error correction technique, the predicted value of the second-stage current difference can be effectively improved and the future current difference prediction value can be further obtained. Then, through a simple addition operation, the motor is in the next sampling cycle and The stator current value corresponding to the double voltage vector that can be applied can be predicted. Next, define a cost function to quantify the error between the current predicted value and the command value. By selecting two voltage vectors corresponding to the least cost function, the gate signals corresponding to the two voltage vectors are sequentially applied to the inverter in the next sampling period. In addition, in the present embodiment, the in-line permanent magnet synchronous motor is connected to a three-phase six-switch inverter, which can generate seven basic voltage vectors, namely a zero voltage vector and six non-zero voltage vectors. . Compared with the existing predictive current control law, in which the law applies only one basic voltage vector in one sampling period, the proposed method can effectively reduce the current tracking error of the motor drive system.

Description

Dual voltage vector modelless predictive current control rule using predictive error correction technique for embedded permanent magnet synchronous motor drive system

The invention relates to a dual voltage vector modelless predictive current control rule using a prediction error correction technique, which is a switching switching technology of a three-phase six-switch inverter, which is characterized in that two sampling cycles can be applied. The basic voltage vector and the corresponding application time can be adjusted separately to effectively improve the stator current tracking capability in the embedded permanent magnet synchronous motor drive system.

The conventional technique predicts the amount of stator current that may be generated by the different switching modes of the motor in the next sampling period by measuring the stator current of the motor and calculating the amount of current change to determine the switching state of the inverter in the next sampling period. It is worth mentioning that the limitations of the prior art include: the inverter can only perform one basic voltage vector or one switching state in each sampling period, and at the same time, each selected basic voltage vector or switching state is under The time that is applied during one sampling period is fixed and is typically applied for one sampling period. For more details, please refer to a Republic of China invention patent published on July 21, 105, and invention No. I543521. The current-controlled inverter switching of the embedded permanent magnet synchronous motor and synchronous reluctance motor disclosed is disclosed. In the method, the method can only apply one basic voltage vector in each sampling period , and the application time is maintained for one sampling period . Through the current sensor, the stator current can be detected and the corresponding current change amount can be calculated to update the old value. Next, when predicting the current of the motor, the future value is expected based on the current current change amount recorded in each switch switching state. During the current sampling period, the current stator current value can be detected by the current sensor, and the value is added to the latest current change amount recorded to calculate the motor at the next current sampling time point and different switches. Possible current predictions under mode. Finally, the current predicted values calculated in all switching modes are substituted into a designed cost function for quantization and comparison, and the switching state corresponding to the minimum cost function value will be in the next sampling period. It is applied to further control the power switch within the frequency converter. In summary, the prior art is not the most perfect technology, and it has many limitations and disadvantages. For example, the prior art cannot apply two basic voltage vectors in one sampling period and its corresponding two applications. The time can also be adjusted separately instead of two identical application times. Therefore, how to determine the switching state of the two basic voltage vectors and their corresponding two application times, the current sampling method for different application times, the current prediction method under irregular application time, how to calculate the application time, and the cost function The design method and the like are still the problems that the invention No. I543521 has not disclosed and needs to be overcome. In view of this, Applicants have proposed a new method, called a dual voltage vector modelless predictive current control rule using predictive error correction techniques, to improve the limitations and disadvantages of the above-described prior art. The proposed method is progressive compared to conventional techniques, and progressive validation is obtained through experimental data and results. In addition, the proposed method breaks through the limitations of the prior art that only one basic voltage vector and a fixed application time can be used in each sampling period. In short, the proposed method enables the frequency converter to use two basic voltage vectors of non-fixed application time in each sampling period. In other words, the proposed method has two different and advanced technical features compared to the prior art: 1. The basic voltage vector that can be used in each sampling period is upgraded from one to two ; 2. In each sampling period. The time during which the internal base voltage vector is applied is upgraded from fixed to non-fixed and adjustable . Further, the novel technology developed by the present invention in order to satisfy the above two important technical features is novel, and the new technology cannot be directly promoted by the technique disclosed in the invention No. I543521. This makes the new predictive current control method proposed by the present invention have more advantages than the prior art, including: smaller current tracking error, smaller current ripple, lower total harmonic distortion, and more The design flexibility and so on. Therefore, the proposed invention is more suitable for the related application of the precision industry than the prior art, and is expected to bring more possible industrial utilization.

From the theory of linear algebra, the resultant voltage vector obtained by linear combination of two basic voltage vectors will be more flexible in some applications than using only a single basic voltage vector, for example, in each sampling period. Two basic voltage vectors with non-fixed application time allow the inverter to output a more detailed equivalent voltage and current required by the load in one sampling period. This will undoubtedly increase the resolution of the inverter on the output voltage. As the resolution of the output voltage increases, the difficulty of current prediction for the corresponding load increases. For example, the four possibilities and the 100 possibilities are different in solving the optimization problem. For another example, the invention No. I543521 proposes a method for predicting the stator current of a future motor under consideration of possible four voltage vectors in the next sampling period, but when considering The resolution of the voltage vector is increased. For example, the linear combination of the two basic voltage vectors can synthesize more than four voltages, which can be more than 100 or more than 200, and the digital signal processor used by the digital signal processor. The performance determines that conventional techniques will not solve this type of problem, let alone calculate the individual application times of the two basic voltage vectors. The main technical limitations and descriptions of the invention No. I543521 are as follows:

(1) Influence of current sampling error: When the digital system reads current, it must pass through the current sensor and the analog/digital converter, and then transmit to the digital signal processor, and the sampling time must be extremely accurate. When the stator current of the motor is read from the detected current to the digital signal processor, it will be limited by the hardware circuit and the sensor to generate sampling errors. These restrictions include: delay of the transmission signal; current sensor There is a fixed level offset; the analog/digital converter has a fixed conversion time; the digital system has a fixed resolution error when parsing the analog information. Furthermore, in order to avoid current surges to measure a more stable current, it is necessary to sample in advance or after a fixed period of time. If the two basic voltage vectors are to be applied in each sampling period, the first period of application time and the second period of application time can be freely adjusted. Considering extreme examples, the first or second application time may be extremely short. The smaller the amount of current change that will be measured during the shorter application time, makes the effect of the above sampling error unnegligible. The model-free predictive current control law is based on current sampling for current prediction, so it is extremely sensitive to the accuracy of current sampling. However, the prior art does not propose a countermeasure strategy for the influence of sampling error. In the case of being freely adjusted, the use of conventional techniques may result in a decrease in the stability of the system.

(2) Time taken by the hardware circuit for current sampling: The conventional model-free predictive current control strategy performs current sampling once after the start and end of the voltage vector (or corresponding switch switching state). In terms of hardware circuits, the analogy is converted to a digital signal and then transmitted to The digital signal processor needs to spend some time. For the hardware circuit used in this embodiment, each current sampling needs to consume 10us of conversion time, and the conversion time required for two samplings accounts for about the sampling period (100us). ) about 20%. Therefore, if two voltage vectors are to be applied in one sampling period, the required sampling signal conversion and transmission time will increase, which will increase the difficulty of the implementation of this method.

(3) Arrangement combination of the number of voltage vectors and the length of application time: If the predictive current control strategy used in the invention No. I543521 is applied to a three-phase six-switch voltage source inverter, only seven basic fixed application times are applied. The voltage vector is available for selection. If all possible permutation combinations considering the two basic voltage vectors and different application times are included in the prediction current control algorithm, the amount of computation required will be very time consuming. Therefore, how to effectively reduce the amount of calculation required for predictive current control is necessary to effectively reduce the computational burden of the digital signal processor. Or, at a very high hardware cost, parallel computing with several extremely advanced digital signal processors can complete the large amount of computation required for all of the possible permutations and combinations in a sufficiently short sampling period. In addition, when the cost of the required hardware cost is too expensive, the new method will be difficult to promote in industrial applications.

(4) Cost function form: The traditional model-free predictive current control strategy mostly adopts the cost function in the form of absolute value. The technical threshold of its implementation is low, but because the absolute value function has a non-differentiable turning point, it cannot be effectively selected. The application time corresponding to the two voltage vectors to be used during the sampling period. To solve this problem, you must use a continuous and differentiated cost function form to make the most detailed and effective considerations when performing predictive current control algorithms.

From the above limitations, conventional techniques cannot use two voltage vectors that can adjust the application time in one sampling period. In view of this, the present invention proposes to use a prediction error repair The positive technology dual voltage vector modelless predictive current control law to solve the above limitations and shortcomings of the prior art, the proposed method can be applied to the embedded permanent magnet synchronous motor drive system. Under the existing drive system architecture, it is only necessary to measure the current signal of the motor once after each switch switch state change, and then the dual vector predictive current control algorithm of the present invention is programmed into the digital signal processor in a programming language. By executing the program, the existing technology can be used to increase the number of electric vectors that can be applied in each sampling period to two and the respective application time can be adjusted, so that the performance of the driving system can be further improved.

In general, the switching frequency of the inverter can reach 10 kHz. At such a high switching frequency, the current variation in each sampling interval seems to be linear because the sampling interval is extremely short and the motor is an inductive load. When predicting the current, it is only necessary to perform a simple linear calculation with the linearly amplified current variable according to the switching state to the time ratio of the sampling period, so that the current value in the switching state of different application time can be predicted.

The present invention retains the advantages of the prior art. The proposed predictive current control method does not require any parameter information of the motor, and each time the switching state changes, only the stator current needs to be sampled once to grasp the switching of the inverter. The trend of current changes caused by the state. Through the designed cost function, the optimal ratio of the switching state of the two inverters to the application time is selected in the future sampling period, and according to the result, the two switches of the inverter in the next sampling period are controlled. State, the purpose of predicting current control can be achieved.

The principle of the basic predictive current of the present invention is that it is not necessary to use a mathematical model of the motor, and only the stored current must be stored, and the stator current of the motor can be predicted through a simple analysis. Then, using a digital signal processor, each time a voltage vector is applied, the three-phase stator current of the embedded permanent magnet synchronous motor is sampled and converted into α-β axis current by coordinates, and the stator current prediction can be started after the setting is completed. With control. In the first step, at the beginning of each sampling period, the amount of current that is linearly amplified during the two sampling intervals in the previous sampling period is calculated. In the second step, according to the method of the present invention, the optimal application time ratio for each dual switching state combination is calculated. In the third step, through the above calculation results, the current change caused by each double switching state combination is calculated to predict the current. In the last step, through the designed cost function, the dual switching state combination that enables the current to be closest to the command value is selected, and the two switching states of the inverter in the next sampling period are controlled according to the optimal application time ratio. Through the above steps, the cycle is repeated, and the purpose of predicting current control can be achieved. Since the application time of the two switching states of the frequency converter can be adjusted rather than fixed in one sampling period, the proportion of the error in the sampling value will become more obvious in the shorter sampling interval. The invention specially designs an error correction technique, which can significantly eliminate the influence of sampling error when calculating the linearly amplified current variation.

To make the basic principles and advantages of the present invention more apparent, the following specific embodiments are described in detail with reference to the accompanying drawings.

1‧‧‧In-line permanent magnet synchronous motor

2‧‧‧ Calculation of three-phase current conversion to α-β axis current

3‧‧‧Parts for calculating the predicted value of linear amplified current change

4‧‧‧ Parts of the best ratio calculation for each pair of switching state combinations

5‧‧‧Means of current prediction calculation

6‧‧‧Parts for calculating cost function and decision switching state

7‧‧‧Three-phase six-switch voltage source inverter

k ‧‧‧ represents the kth sampling period

V _DC ‧‧‧ DC voltage source required for three-phase six-switch voltage source inverter

f _a , f _b ‧‧‧ a phase, b phase stator reference coordinate system voltage or current

f _α , f _β ‧‧‧ α-β axis stator reference coordinate system voltage or current

S _a , S _b , S _c ‧‧‧ three-phase switch gate control signals

S ₀ , S ₁ , S ₂ , S ₃ , S ₄ , S ₅ , S ₆ , S ₇ ‧‧‧Inverter switch switching status

V ₀ , V ₁ , V ₂ , V ₃ , V ₄ , V ₅ , V ₆ , V ₇ ‧‧‧ voltage vector

, ‧‧‧The application time of the switching state is applied in one sampling cycle

T _s ‧‧‧A cycle of sampling cycles

, ‧‧‧The ratio of the application time of the two switching states before and after the kth double vector to the proportion of T _s

, , , , , , , , , , , ‧‧‧Synthesis voltage vector superscript symbol c ‧‧‧ represents synthesis

, ‧‧‧The first two switching states of the kth

, ‧‧‧The first two current changes in the kth time

i _x [ k ,1], i _x [ k ,2]‧‧‧ Current subscript symbol pec ‧‧‧ represents the prediction error correction when sampling twice in the kth sampling period

, ‧‧‧Linearly amplified current change during the k -1th sampling period

, ‧‧‧Predicted value of linearly amplified current change in the k -th sampling period

The superscript symbol P ‧ ‧ represents the predicted value

g ‧‧‧cost function

Superscript *‧‧‧ represents the current command value

Simplified parameters of C _{α 1} , C _{α 2} , C _{β 1} , C _{β 2} ‧‧‧ cost function

‧‧‧The best time to apply time

The superscript symbol OPT ‧‧‧ indicates the best value

e ‧‧‧ α-β axis average current tracking error

J ‧‧‧ α-β axis square root current chopping

N ‧‧‧Sampling ratio

Fig. 1 is a block diagram showing a current control system of an in-line permanent magnet synchronous motor according to an embodiment of the present invention.

Fig. 2 is a structural diagram of a three-phase six-switch voltage source inverter connected to the in-line permanent magnet synchronous motor used in the present invention.

Figure 3 is a basic voltage vector diagram of a three-phase six-switch voltage source inverter.

Figure 4 is a plot of 13 different composite voltage vectors on the α-β plane.

Figure 5 is a schematic diagram of current sampling of the present invention during the kth sampling period.

Fig. 6 is a schematic diagram showing the error of the current variation amount being enlarged.

Fig. 7 is a schematic diagram showing the variation of the four-stage current when the present invention calculates the predicted value of the current variation.

Figure 8 is a schematic diagram of the present invention for predicting current control during a subsampling cycle.

Fig. 9 is a measured waveform diagram of a method of linearly varying the amount of current change with a conventional technique when calculating the amount of current change of linear amplification.

Fig. 10 is a view showing a measured waveform pattern of the prediction error correction method of the present invention when calculating a linearly amplified current variation amount in the embodiment of the present invention.

Figure 11 is a conventional technique for applying conventional model-free predictive current control to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, and setting the α-β axis current command to The measured waveform of the corresponding α-β axis current response in the case where the peak value is 4 A and the frequency is 30 Hz.

Figure 12 is an embodiment of the present invention, the proposed dual vector modelless predictive current control is applied to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, and α-β The axis current command is set to the measured waveform of the corresponding α-β axis current response in the case where the peak value is 4 A and the frequency is 30 Hz.

Figure 13 is a conventional technique for applying conventional modelless predictive current control to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, and setting the α-β axis current command to The measured waveform of the corresponding α-β axis current response in the case where the peak value is 4 A and the frequency is 10 Hz.

In the embodiment of the present invention, the proposed dual-vector model-free predictive current control is applied to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, and α-β is applied. The axis current command is set to the measured waveform of the corresponding α-β axis current response in the case where the peak value is 4A and the frequency is 10 Hz.

Figure 15 is a conventional technique for applying conventional model-free predictive current control to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, setting the α-β axis current command to a peak value. The measured waveform of the corresponding α-β axis current response is 4A and the frequency is 10 Hz, and the α axis current command is instantaneously changed from -4A to 4A at 0.05 seconds.

Figure 16 is a diagram showing the proposed dual-vector model-free predictive current control applied to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, and the α-β axis is used in the embodiment of the present invention. The current command is set to a peak value of 4A and a frequency of 10 Hz, and when the α- axis current command is instantaneously changed from -4A to 4A at 0.05 seconds, the corresponding measured waveform of the α-β axis current response.

Figure 17 is a conventional technique for applying the conventional model-free predictive current control to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, and setting the α-β axis current command to 10 Hz. The measured waveform of the corresponding α-β axis current response when the amplitude of the α-β axis current command is instantaneously changed from 1A to 4A at 0.2 seconds.

Figure 18 is a schematic diagram of the proposed dual-vector model-free predictive current control applied to a three-phase six-switch voltage source inverter connected to an in-line permanent magnet synchronous motor drive system, and the α-β axis is used in the embodiment of the present invention. The current command is set to a frequency of 10 Hz, and when the amplitude of the α-β axis current command is instantaneously changed from 1 A to 4 A at 0.2 seconds, the corresponding measured waveform of the α-β axis current response.

Fig. 19 is a comparison diagram of the performance of the conventional technique and the embodiment of the present invention; that is, a comparison chart of the quantitative results of the equations (22) and (23) in Figs. 11 to 18.

Reference will now be made in detail to the embodiments of the invention, In this embodiment, how to apply the dual voltage vector modelless predictive current control law to the embedded permanent magnet synchronous motor for further explanation and formula derivation.

Fig. 1 is a block diagram showing a current control system of an in-line permanent magnet synchronous motor according to an embodiment of the present invention. Referring to Fig. 1, the three-phase stator current in the in-line permanent magnet synchronous motor 1 is sampled in the kth sampling period, and the α-β axis current is calculated via block 2. The dual vector modelless prediction current control law of the embodiment of the present invention includes blocks 3, 4, 5, and 6. First, the sample-converted α-β axis current value is calculated as a linearly amplified current change amount predicted value at block 3. The optimal ratio of the application time of each of the dual switching state combinations in the dual vector mode is calculated in block 4. Then, according to the calculation result of the block 4, the current prediction value which will be caused by combining the respective double-vector application time optimal ratios in the k- th and second-times of the k + 2th time is predicted in block 5. Finally, all the current prediction values are substituted into the cost function at block 6, and the double-switching state combination of selecting the minimum cost value in block 6 is output according to its dual vector application time optimal ratio, and the corresponding switch control gate signal is output to block 7, A closed loop drive control system is completed by controlling the three-phase stator current of the in-line permanent magnet synchronous machine 1. The above k represents the kth sampling. The V _DC in the figure represents the DC voltage source required for a three-phase six-switch voltage source converter.

The predictive current control method of the embodiment of the invention can be applied to the embedded permanent magnet synchronous motor drive system. Under the existing digital drive system and the three-phase six-switch voltage source inverter structure, only the three-phase stator of the motor is needed. Current sampling, which is written into the digital signal processor by the dual vector modelless predictive current control method of the present invention, and executes a program, can replace the traditional inverter switching control strategy. The following describes a dual vector modelless predictive current control method:

In order to effectively simplify the mathematical model of the embedded permanent magnet synchronous motor and reduce the computational complexity, the embodiment of the invention adopts the α-β coordinate transformation commonly used in the field of motor drive technology, which can reduce the complexity of the control system. Assuming the three-phase balance of the embedded permanent magnet synchronous motor, the motor c- phase information can be included by the two phases a and b , and the abc axis projection to the α-β axis coordinate system conversion matrix can be equivalent to the ab axis projection to α- Β- axis coordinate system conversion matrix: In formula (1), f _a and f _b represent the voltage or current of the a- phase and b- phase stator reference coordinate system in the three phases. In contrast, f _α and f _β represent the voltage of the α-β axis stator reference coordinate system or Current.

The three-phase six-switch voltage source inverter structure used in the present invention is as shown in Fig. 2. Each phase has two switches, and a voltage square wave of V _DC or 0 volts can be applied to the three phases a , b , and c of the motor to control the current. FIG 2 is S _a, S _b, S _c representing a, b, c of the three-phase switching gate control signal, the switch is open if it is 0, when a switch is turned on. Note that the upper and lower arm switch gate control signals of the inverter in Figure 2 must be opposite to each other. If the upper arm is turned on, the lower arm is disconnected. Otherwise, if the upper arm is disconnected, the lower arm is turned on. This prevents the upper and lower arm switches from being turned on at the same time. Short circuit of power supply.

The switch gate signals ( S _a , S _b , S _c ) can be arranged as (0,0,0), (0,0,1), (0,1,0), (0,1,1), ( Eight different inverter switching states, such as 1,0,0), (1,0,1), (1,1,0), (1,1,1), respectively, in the order of S ₀ , S ₁ , S ₂ , S ₃ , S ₄ , S ₅ , S ₆ , S ₇ represent, in the field of voltage vector commonly used in the field of frequency converter, can be expressed as eight voltage vectors in the α-β plane, refer to Figure 3 , denoted as V ₀ , V ₁ , V ₂ , V ₃ , V ₄ , V ₅ , V ₆ , V _{7 , respectively} . The corresponding relationship between the above switching state and the voltage vector is as shown in Table 1, wherein V _DC is a DC voltage source of the three-phase six-switch inverter; v _a , v _b , and v _c are respectively a, b, and c phase voltages of the stator; V _α is the voltage of the α axis; V _β is the voltage of the β axis.

Referring to Fig. 4, using the method of vector synthesis in linear algebra, eight basic voltage vectors are synthesized into a composite voltage vector in 13 regions by linear combination, which are respectively recorded as , , , , , , , , , , , , The superscript symbol c represents synthesis. These 13 areas can be divided into 6 line segments, 6 intervals and the origin. Wherein, the composite voltage vector falling at the origin is ,and To Each falls within a line segment or interval. Since any two basic voltage vectors can be combined, there are 8 ² = 64 possible combinations of switching states to be considered. In order to reduce the computational complexity, the present invention selects only one for the origin and each line segment and interval. Double switching state combinations, so only a total of 13 dual switching state combinations have to be considered, and the composite vector selection range before simplification can still be preserved. The basic switching state and the combined voltage vector corresponding to the dual switching state combination in this embodiment are as shown in Table 2.

In Table 2, Indicates the duration of the first switching state in the kth double switching state combination, and Indicates the duration of the second switching state in the kth double switching state combination, T _s is the sampling period, and the field of the linear combination operation represents the calculation method of the composite voltage vector. Compared to using only 8 basic voltage vectors, using the 13 synthesized voltage vectors described above will have a more free voltage vector selection range, so that the current can be finerly adjusted to achieve better current prediction.

The present invention is based on detecting stator current and calculating current variation to predict the stator current value of the drive in a possible switching mode of the inverter in the future. The stator current is sampled twice during each sampling period to calculate the amount of current change for the respective two switching states. For the convenience of explanation, first introduce the mathematical symbols used. Apply a double voltage vector in each sampling cycle (including: the first switching state and the second switching state), refer to Figure 5, in the first switching mode After being applied, the first stator current sampling is performed in advance or delayed for a fixed period of time to avoid reading the current surge generated by the power switch in the inverter at the switching instant. This embodiment selects a delayed reading scheme, which The α-β axis stator current sampled for the first time is expressed as i _x [ k ,1], where the subscript x { α , β }. Similarly, in the second switching mode After a fixed period of time has elapsed, a second current sampling is performed, denoted as i _x [ k , 2]. As can be seen from Figure 5, the first switching state Was imposed The first current variation caused by the time is calculated by subtracting the value of i _x [ k , 1] from i _x [ k , 2], which is expressed as . And the second switching state Was imposed The second current variation caused by the time is calculated by subtracting the value of i _x [ k , 2] from i _x [ k +1,1], which is expressed as . Match The corresponding double switching state combination is recorded as . The superscript c1 represents the first switching state, and the superscript c2 represents the second switching state. The sampling period is recorded as T _s , where the application time of the two switching states versus Both can be adjusted rather than fixed, but the following relationships must be met: Then, in order to simplify the operation process, in the kth sampling period, the ratio of the first application time is defined as And the second period of application time is recorded as . with The ratio of the time of application of the first segment and the second segment to T _s is: As can be seen from the above formula, the ratio of the application time of the first segment and the second segment versus Between the real numbers 0 to 1, ie , [0,1]. According to formula (2) to formula (4), the following relationship can be easily known:

In the architecture of the dual vector model-free predictive current controller, if it is necessary to predict the future stator current value of the motor, the current change amount needs to be predicted. The concept of current prediction used by the dual-vector model-free predictive current controller assumes that the sampling frequency is fast enough to predict the future current variation based on the current change in the same switching state. After sampling the latest α-β axis current value, in order to predict the amount of two current changes generated by applying different voltage vectors with different application times, the amount of current change that can be calculated must first be linearly amplified to The amount of current change over the entire sampling period time T _s , which is referred to as the amount of current change that is linearly amplified . When calculating the current change amount of each switching state with different durations, it is calculated based on the amount of current change that is linearly amplified . The traditional model-free predictive current control uses the most intuitive way, that is, directly divides the current change amount by the corresponding application time ratio for linear amplification, but the disadvantage is that the error of the current change amount is also amplified, as in the sixth The figure shows.

In order to solve the problem that the above error is linearly amplified, a prediction error correction method is proposed for the first time in the present invention: the measured current change amount plus the past predicted linear amplification current change amount is added in proportion to each time. To estimate the sampling cycle time T _s and the switching states are respectively with The current variation using the prediction error correction is calculated as: In formula (6) and formula (7), , { S ₀ ,..., S ₆ }, the subscript pec represents the prediction error correction , with Represented in the switching state as with And the corresponding application time ratio is versus The estimated time of application is the amount of current change corrected by the prediction error at T _s . In addition, in formula (6) and formula (7) versus Corresponding to the predicted values of the 13 linearly amplified current changes stored in the past in the k -1th sampling period. with The predicted value of the current change amount of the two switching states, the old value will be replaced by the new value calculated by the equations (6) and (7) in the kth sampling period, that is, In fact, the algorithm of predictive control is not possible to be completed in an instant in a digital signal processor. In order to make the wrong result in the calculation process, the timing of the kth update is designed. After getting i _x [ k ,1], the update is performed immediately.

It is worth mentioning that using (6)-(9) does not require any motor parameters. It only needs to measure current, calculate current difference and simple multiplication operation to correct the prediction error of current variation. The advantages of conventional techniques without the need to build mathematical models are preserved. Since the performance of today's microcontrollers or digital signal processors can achieve a sufficiently short sampling period T _s , it is acceptable to approximate the value of the next sampling period with the latest recorded value when calculating the future current value. This is also in line with the basic spirit of predictive control, using current and past data to predict future values. When predicting the amount of current change, it is only necessary to multiply the switching state by the time ratio in T _s , and perform a linear reduction operation, and then combine the two switching states into And the sum of the predicted value of the two-stage current change corresponding to the switch-on application time of T _s (refer to the first-stage and second-stage current change amounts in FIG. 7) can be obtained by the following formula: In the above formula, the superscript symbol P represents the predicted value and is defined by the formula (5). Therefore, the application time of the second segment can be expressed as (1- ). In the actual situation, the calculation time of the digital signal processor can not be ignored, and the resulting switch control delay will greatly affect the accuracy of the predictive control. To solve this problem, the current value of the k + 2th time can be further predicted in the next sampling period. Thus, the ratio of the dual switching state of the frequency converter to its corresponding application time ratio during each sampling period will be determined earlier in the previous two sampling periods, and is not affected by the calculation time delay. The method for predicting the k + 2th current value will be described below: the same as the principle of the formula (10), the k in the formula (10) is further advanced, and k +1 is substituted, and the k +1 second two-stage current is used. The sum of the variation prediction values (ie, the third segment and the fourth segment current variation in FIG. 6) can be obtained by the following formula:

Using Equation (10) and Equation (11), it is possible to predict the amount of current variation for a total of four segments during the kth and kth +1 sampling periods, as is evident from Figure 6, if you want to predict at k + The first sampling current value in the 2 sampling period only needs to be based on the current sampling value i _x [ k , 1] at the beginning of the current sampling period, plus the formula (10) and the following k and the following in equation (11) The current change prediction value of k +1 cycle can be used. Then, the formula for predicting the current used in the k + 2th time of the present invention can be organized as:

Thus, using Equation (12), the current predicted value corresponding to each of the dual switching state combinations can be calculated. It can be seen from the formula (10) to the formula (12) that, in one sampling period, if the ratio of the application time of the two switching states before and after the adjustment is made, the current prediction value also changes. In order to select the most suitable double switching state combination and the corresponding application time ratio, so that the current can be accurately controlled, and as close as possible to the current command value, the present invention designs a cost function in the form of a square to facilitate the subsequent selection of the best. Dual switch state combination: In equation (13), g represents the cost function, and the superscript * represents the current command value. This cost function quantifies the magnitude of the error between the current predicted value and the current command value, and the error value can be used to compare the magnitude. From equations (10) to (12), it can be known that the time ratio of the double vector has an important influence in the cost function. In order to present this characteristic more clearly, the formula (13) is expressed as the following form: Among them, the parameters C _{α 1} , C _{α 2} , C _{β 1} , C _{β 2} are respectively defined as: Since a total of 13 types of inverters have to be considered for the combination of the two switching states, C _{α 1} , C _{α 2} , C _{β 1} , and C _{β 2} are also calculated to correspond to 13 groups. From equation (14) to equation (18), it can be clearly observed that the designed cost function is in the form of the sum of the squares of the current tracking errors. The design concept is because The curve corresponding to the cost value when changing has a differentiating characteristic, and therefore the optimal time ratio is applied. Will occur in the cost function g pair In the case of partial differentiation and then zeroing, ie At this time To apply the best ratio of time and be remembered , its calculation formula can be organized according to formula (19): Look at the formula (20), because C _{α 1} , C _{α 2} , C _{β 1} , C _{β 2} are combined in 13 different double switching states ~ The value will be different, so the formula (20) is combined in 13 different double switching states. ~ Will produce 13 different optimal application time ratios . Ideally, 0 1, that is [0,1], if you consider the limitations of the implementation of the hardware circuit and the digital signal processor, for example, the digital signal processor used has no parallel processing function, plus the need to consider converting the analog current signal into digital The current signal requires a conversion time. Generally, the conversion and transmission time of the external analog-to-digital converter is about 5~15 μs , and the sampling period is 100 μs . Therefore, in practice, The limit of the limit is slightly smaller than the theoretical value. Combining 13 double-switching states into their respective substitution formulas (14), selecting the double-switching state combination corresponding to the minimum cost function value, and sequentially outputting the first voltage vector and the first in the k- th +1 sampling period. Two voltage vectors, and their corresponding application time versus The calculation formula is: In this way, a dual vector modelless predictive current controller can be implemented. According to the present invention, the current predicted to the kth + 2nd time is as shown in FIG. 8 , corresponding to 13 switching states considered, and there are 13 current prediction values of the k + 2 times, and each current prediction value is in a respective switching state. The optimal vector ratio of the double vector time. In the kth sampling period, the first switching control signal is output to the inverter according to the past calculation result, and then the current is read, the current of the k + 2th is predicted, and the k + 2 is selected according to the cost function. The second double switching state combination. At the same time, in the process, according to the past calculation results, the second switch control signal is output to the inverter at the calculated time point. Based on the description of the above embodiments, the implementation method of the embodiment of the present invention in the kth sampling period can be organized into the following six steps:

Step 1: According to the decision in the k -1th sampling period, after outputting the first switch control signal to the inverter, read the current i _x [ k , 1], and subtract the previous current sample value i _x [ k -1, 2] to calculate the amount of current change, ie i _x [ k ,1]- i _x [ k -1,2].

Step 2: Calculate formula (6) and formula (7), and update according to formula (8) and formula (9).

Step 3: According to the decision in the k -1th sampling period, the system is configured to output a second switch control signal to the inverter at a desired time point, and read the current i _x [ k , 2] after the inverter switch is switched. To calculate the current change amount i _x [ k , 2] - i _x [ k , 1].

Step 4: Calculate the formula (15) to formula (18) for the 13 combined switching mode combinations, and calculate the 13 groups of C _{α 1} , C _{α 2} , C _{β 1} , C _{β 2} into the formula (20). And then according to the actual conditions of the conditions to obtain the best ratio of 13 application time .

Step 5: Optimal ratio of 13 application times Substituting equation (14) to calculate 13 cost function values.

Step 6: Select the smallest one of the calculated 13 cost function values, and the corresponding double switching state combination (including the first switching state and the second switching state) will be applied in the next sampling period. And formula (21) and formula (22) can calculate the application time corresponding to the first voltage vector and the second voltage vector versus . However, the above-described embodiments are merely illustrative of the present invention and are not intended to limit the actual application of the present invention.

The following is the result of implementation: In the embodiment of the present invention, a driving system of a three-phase six-switch inverter connected to an in-line permanent magnet synchronous motor is actually constructed to verify that the proposed double-vector model-free predictive current control law is embedded in the embedded system. The feasibility and correctness of the permanent magnet synchronous motor. The three-phase stator current information is captured and stored by the TMS320F28335 digital signal processor produced by Texas Instruments, and the actual waveform is drawn in the form of α-β axis current through a personal computer, and the result is quantified. Several implementation results may prove that the method proposed by the embodiments of the present invention is achievable.

In Fig. 9 to Fig. 21, i _α represents the current on the α axis, i _β represents the current on the β axis, i _α ^* represents the current command on the α axis, and i _β ^* represents the β axis. Current command.

Fig. 9 is a measured waveform diagram of current control using the present invention. However, when calculating the amount of linear amplification current change, the current varies linearly with time using a conventional technique, and the current is significantly deviated from the command value. FIG. 10 is a diagram showing the actual measurement waveform pattern of the current control using the prediction error correction method of the present invention in the embodiment of the present invention. It can be seen that the proposed prediction error correction method is effective.

11th, 13th, 15th, and 17th are conventional techniques, and a conventional model-free predictive current control strategy is adopted in a three-phase six-switch inverter connected to an in-line permanent magnet synchronous motor drive system. , measured waveforms under different alpha axis current commands and beta axis current commands. 12th, 14th, 16th, and 18th are conventional techniques. In the driving system of a three-phase six-switch inverter connected to an in-line permanent magnet synchronous motor, the double vector no model proposed by the present invention is adopted. Predictive current control strategy, measured waveforms under different alpha axis current commands and beta axis current commands. It can be seen from FIG. 11 to FIG. 18 that the dual-vector model-free predictive current control strategy of the embodiment of the present invention has a current control effect under both transient and steady-state responses compared to the conventional model-free predictive current control strategy. Better than conventional technology.

In order to quantify the results, the α-β axis average current tracking error e and the rms current chopping J are defined as: In the present embodiment, the sampling ratio N = 1000. The smaller the current error and the chopping current, the better the current control effect. The measured data is quantified by the formula (22) and the formula (23) and plotted in the 19th figure, and the quantized data of the conventional model-free predictive current control is used. 100% reference. It can be seen from FIG. 11 to FIG. 18 that the dual-vector model-free predictive current control strategy of the embodiment of the present invention has better current error and current ripple than the prior art.

In summary, the proposed predictive current control strategy can greatly improve the current response of the embedded permanent magnet synchronous motor drive system. Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention, and those skilled in the art can make some changes and refinements without departing from the spirit and scope of the present invention. The scope of the invention is defined by the scope of the appended claims.

Claims (2)

A dual voltage vector modelless predictive current control method using prediction error correction technology, which is suitable for an in-line permanent magnet synchronous motor powered by a frequency converter, and a dual voltage vector modelless prediction current using prediction error correction technology The control method includes: reading the secondary motor stator current in the sampling period set by the signal processor of each number of times; using the read stator current, according to the stored current record and the switching interval between the switches, with a prediction error Correction method to calculate the linearly amplified α-β axis current change amount; using a double vector application time optimal ratio calculation method to calculate the optimal ratio of each double switching state combination application time, and calculate the optimal ratio The current predicted value at the next sampling period; in the α-β axis coordinate, the difference between the current command and the current predicted value is taken as a cost function to quantify the value of the cost function under each double switching state combination; And by using the cost function, the cost function value under different double switching state combinations can be calculated, so as to pick Selecting a dual switching state combination corresponding to the minimum cost function, the optimal ratio of the dual switching state combination to its application time will output a signal to the frequency converter to control the switching of the power switch in the next sampling period; wherein the prediction error correction method , limited to calculation using the following formula: In formula (A) and formula (B), the subscript symbol x represents the α axis or the β axis; the subscript symbol pec represents the prediction error correction; , Indicates the first and second switching states of the inverter in the k -1th sampling period, and the corresponding linear amplification current changes are respectively with Representative; i _x [k -1,1] is the value of the current sampling time k -1 of the first sampling period; i _x [k -1,2] of the k -1 for the second sub-sampled period Current sampling value; i _x [ k , 1] is the first current sampling value in the kth sampling period; The time ratio of the first switching state in the k -1th sampling period; Is the time ratio of the second switching state in the k -1th sampling period; versus For the k -1th sampling period, among the 13 linearly amplified current variation predicted values stored in the past, corresponding with The two, which will be updated and replaced by the values calculated by equations (A) and (B) during the kth sampling period, are replaced by the following formula: And the timing of the kth substitution is performed immediately after obtaining i _x [ k , 1]; wherein the calculation method of the optimal ratio of the double switching state combination application time is limited to calculation using the following formula: In formula (E), the superscript symbol OPT represents the optimum value; The time-optimal ratio is applied; wherein the definition limits of the parameters C _{α 1} , C _{α 2} , C _{β 1} , C _{β 2} are: In formula (E), the subscript symbols α and β represent current information under the α axis and the β axis; , The current command on the α- axis and the β- axis in the kth pre-control loop ; respectively, where the current prediction value is limited to the calculation using the following formula: In formula (J), the superscript symbol P represents the predicted value of the current, Indicates that in the k + 2th sampling period, the possible double switching state combinations are The predicted value of the current at the first sampling; versus The predicted values of the current changes in the kth and kth +1th sampling periods, respectively, are calculated using the following formula: Among them, the cost function is limited to using the following formula: In formula (M), g represents a cost function. According to the predictive current control method described in the first paragraph of the patent application scope, wherein the dual switching state combination is composed of two switching states according to a sequential switching sequence, and is limited to , , , , , , , , , , , , And so on 13 combinations: The first switching state and the second switching state are both selected to apply S ₀ ; Applying S ₄ in the first switching state, and applying S _{0 in the} second switching state; The first switching state is the application of S ₆ , and the second switching state is the application of S ₀ ; The first switching state is the application of S ₂ , and the second switching state is the application of S ₀ ; The first switching state is the application of S ₃ , and the second switching state is the application of S ₀ ; The first switching state is the application of S ₁ , and the second switching state is the application of S ₀ ; The first switching state is the application of S ₅ , and the second switching state is the application of S ₀ ; The first switching state is the application of S ₄ , and the second switching state is the application of S ₆ ; The first switching state is the application of S ₆ , and the second switching state is the application of S ₂ ; The first switching state is the application of S ₂ , and the second switching state is the application of S ₃ ; The first switching state is to apply S ₃ , and the second switching state is to apply S ₁ ; The first switching state is the application of S ₁ , and the second switching state is the application of S ₅ ; The first switching state is the application of S ₅ , and the second switching state is the application of S ₄ ; wherein the switching states S ₀ , S ₁ , S ₂ , S ₃ , S ₄ , S ₅ , S ₆ correspond to The switch control signal is recorded as ( S _a , S _b , S _c ) as follows: when the switching state is S ₀ , the switch control signals ( S _a , S _b , S _c ) are (0, 0, 0); When the switching state is S ₁ , the switch control signal ( S _a , S _b , S _c ) is (0, 0, 1); when the switching state is S ₂ , the switch control signal ( S _a , S _b , S _c ) is (0,1,0); when the switching state is S ₃ , the switch control signal ( S _a , S _b , S _c ) is (0, 1, 1); when the switching state is S ₄ , the switch control signal ( S _a , S _b , S _c ) is (1, 0, 0); when the switching state is S ₅ , the switch control signal ( S _a , S _b , S _c ) is (1, 0, 1); the switching state is S ₆ The switch control signals ( S _a , S _b , S _c ) are (1, 1, 0); wherein, S _a , S _b , and S _c are respectively a , b , c phase switches of the three-phase six-switch inverter The control signal, when 0, indicates that the upper arm power switch of the corresponding phase is turned off and the lower arm power switch is turned on, and when it is 1, it indicates that the upper arm power switch of the corresponding phase is turned on. The power switch is turned off.