Direct current brushless motor driving method

Abstract

The invention relates to the technical field of direct current motors, and particularly discloses a direct current brushless motor driving method. A central electric signal, a first electric signal, a second electric signal and a third electric signal of a motor stator winding are analyzed by using a signal processing and analysis algorithm based on deep learning, so that implicit features in each electric signal are captured; and after difference calculation is carried out on the center electric signal feature and other electric signal features, query code representation of the rotor rotation position is obtained in a feature query matching mode, so that the rotor rotation position is determined. Therefore, accurate judgment on the rotation position of the rotor can be realized through artificial intelligence and a deep learning algorithm, and the influence of noise interference and external condition change is reduced not only through a hardware component mode, so that the flexibility and generalization ability are improved through a more intelligent direct current brushless motor driving mode.

Description

Technical Field

[0001] The present application relates to the technical field of direct current motors, and more specifically, to a direct current brushless motor driving method. Background Art

[0002] Brushless DC motor (BLDC) is a high-efficiency motor widely used in industrial automation, home appliances, automotive electronics and other fields. Compared with traditional brushed DC motors, brushless DC motors have higher reliability and longer service life because they have no physical contact brushes. In addition, brushless DC motors are becoming more and more popular in modern applications due to their high efficiency and good control performance.

[0003] In the driving process of BLDC motor, accurately detecting and controlling the position of the rotor is the key to achieve efficient and smooth operation. Traditional rotor position detection usually relies on external sensors such as Hall effect sensors or encoders to provide feedback signals. However, this method increases the complexity and cost of the system, and when the distance between the Hall sensor and the rotor is too far, it may cause poor sensing, thus becoming a weak link in system reliability.

[0004] In response to the above problems, the invention with patent announcement number CN115149846A proposes a brushless DC motor and its driving method, which prevents the circuit part from malfunctioning in harsh environments (such as low temperature, humidity, and corrosive environments) by placing the control drive circuit outside the motor body and connecting it to the stator winding of the motor through a wire, thereby extending the service life of the motor. Specifically, the comparator in the control drive circuit obtains and compares the center point electrical signal CT and each phase electrical signal of the three-phase coil of the motor stator winding, and then the induced voltage detector detects the zero crossing point of the coil induced voltage in the non-powered state to determine the specific position of the rotor. After that, the waveform synthesizer will generate a corresponding control signal based on the position information of the rotor, and after adjustment by the level converter, the output module will finally drive the motor to work.

[0005] In the above-mentioned driving method of the brushless DC motor, a signal processing method based on traditional hardware components, such as a comparator, an induction voltage detector, etc., is used to detect the rotation position of the rotor. This method simply compares the electrical signals and may be affected by noise interference, resulting in errors in the detected zero crossing points, thereby affecting the accuracy of the rotor position. This error is particularly evident in high-speed operation or when the signal quality is poor. In addition, if the external conditions change greatly, such as temperature fluctuations or power supply fluctuations, the performance of these hardware components may be affected, resulting in unstable detection results, thereby affecting the normal operation of the motor. In addition, hardware components are usually fixedly configured, lack intelligent processing methods, and are difficult to change once the design is completed. This means that if the detection algorithm needs to be adjusted to adapt to different working conditions, the hardware may need to be redesigned, which is time-consuming and laborious.

[0006] Therefore, an optimized DC brushless motor driving solution is desired. Summary of the invention

[0007] The present application provides a brushless DC motor driving method that can accurately determine the rotation position of the rotor through artificial intelligence and deep learning algorithms, rather than just through hardware components, thereby reducing the impact of noise interference and changes in external conditions, thereby improving flexibility and generalization capabilities with a more intelligent brushless DC motor driving method.

[0008] In a first aspect, a method for driving a brushless DC motor is provided, comprising:

[0009] Obtaining the central electrical signal of the center point of the three-phase coil in the motor stator winding;

[0010] Acquire a first electrical signal, a second electrical signal and a third electrical signal of a three-phase coil in a stator winding of the motor;

[0011] Determine the rotation position of the rotor based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the stator winding of the motor and the central electrical signal of the central point of the three-phase coil;

[0012] The rotor rotation position and the voltage stabilization signal are input into a waveform synthesizer for processing, so as to output a level signal to an output module through the waveform synthesizer to drive the motor to operate;

[0013] Wherein, based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the stator winding of the motor and the central electrical signal of the center point of the three-phase coil, the rotor rotation position is judged, including: after feature extraction of the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal, feature differential comparison of the central electrical signal feature with the first electrical signal feature, the second electrical signal feature and the third electrical signal feature respectively to obtain the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature; using the feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature as a sequence of simulated key features, using the central electrical signal feature as a query feature, calculating the gated query matching feature between the query feature and the sequence of simulated key features to obtain a rotor rotation position query coding representation; determining the rotor rotation position based on the rotor rotation position query coding representation.

[0014] In a possible implementation, after extracting features of the central electrical signal, the first electrical signal, the second electrical signal, and the third electrical signal, the central electrical signal feature is respectively compared with the first electrical signal feature, the second electrical signal feature, and the third electrical signal feature to obtain a first electrical signal differential feature, a second electrical signal differential feature, and a third electrical signal differential feature, including:

[0015] Inputting the central electric signal, the first electric signal, the second electric signal and the third electric signal into an electric signal feature extractor based on a dilated convolutional layer to obtain a central electric signal feature vector as the central electric signal feature, a first electric signal feature vector as the first electric signal differential feature, a second electric signal feature vector as the second electric signal feature and a third electric signal feature vector as the third electric signal feature;

[0016] Calculate the differential features between the central electric signal feature vector and each of the first electric signal feature vector, the second electric signal feature vector and the third electric signal feature vector to obtain the first electric signal differential feature vector as the first electric signal differential feature, the second electric signal differential feature vector as the second electric signal differential feature, and the third electric signal differential feature vector as the third electric signal differential feature.

[0017] In a possible implementation, a feature sequence consisting of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature is used as a sequence of simulated key features, the central electrical signal feature is used as a query feature, and a gated query matching feature between the query feature and the sequence of simulated key features is calculated to obtain a query encoding representation of the rotor rotation position, including:

[0018] Arranging the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector and the third electric signal differential feature vector into a sequence of simulated key feature vectors as the sequence of simulated key features;

[0019] Performing one-dimensional convolution coding on the sequence of the simulated key feature vectors to obtain a sequence of simulated key local neighborhood feature vectors, wherein the scale of the one-dimensional convolution coding is equal to an integer multiple of the length of the simulated key feature vector;

[0020] Processing the central electrical signal feature vector using a value embedding coding matrix to obtain a central electrical signal value feature vector;

[0021] The sequence of the simulated key feature vectors and the sequence of the simulated key local neighborhood feature vectors are respectively used as the sequence of key vectors, and the central electrical signal value feature vector is used as the value vector. Local neighborhood implicit optimization and attention-based gating processing are performed on the query vector, the value vector and the sequence of the key vector to obtain the rotor rotation position query encoding representation.

[0022] In one possible implementation, the central electrical signal feature vector is processed using a value embedding coding matrix to obtain the central electrical signal value feature vector, including: after matrix multiplication of the central electrical signal feature vector and the value embedding coding matrix, the obtained central electrical signal feature value time series embedding vector and the value bias vector are added by position to obtain the central electrical signal value feature vector.

[0023] In a possible implementation, the sequence of the simulated key feature vectors and the sequence of the simulated key local neighborhood feature vectors are respectively used as the sequence of key vectors, and the central electrical signal value feature vector is used as the value vector, and the query vector, the value vector and the sequence of the key vector are subjected to local neighborhood implicit optimization and attention-based gating processing to obtain the rotor rotation position query encoding representation, including:

[0024] Using the central electric signal feature vector as a query vector, the central electric signal value feature vector as a value vector, and using the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector, and the third electric signal differential feature vector as a sequence of key vectors, inputting the sequence of the query vector, the value vector, and the key vector into a first multi-head attention query module based on a converter structure to obtain a sequence of electric signal key granularity query matching implicit vectors;

[0025] Taking the central electric signal feature vector as a query vector, the central electric signal value feature vector as a value vector and the sequence of the key-like local neighborhood feature vectors as a sequence of key vectors, inputting the query vector, the value vector and the sequence of the key vectors into a second multi-head attention query module based on a converter structure to obtain a sequence of electric signal key neighborhood granularity query matching implicit vectors;

[0026] Calculating the positional mean vector of the sequence of the electric signal key granularity query matching implicit vectors and the sequence of the electric signal key neighborhood granularity query matching implicit vectors to obtain the electric signal key granularity query matching feature vector and the electric signal key neighborhood granularity query matching feature vector;

[0027] The electrical signal key granularity query matching feature vector and the electrical signal key neighborhood granularity query matching feature vector are fused to obtain a rotor rotation position query encoding representation vector as the rotor rotation position query encoding representation.

[0028] In a possible implementation, the central electric signal feature vector is used as a query vector, the central electric signal value feature vector is used as a value vector, and the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector and the third electric signal differential feature vector are used as a sequence of key vectors, and the sequence of the query vector, the value vector and the key vector is input into a first multi-head attention query module based on a converter structure to obtain a sequence of electric signal key granularity query matching implicit vectors, including:

[0029] After calculating the product between the central electric signal feature vector and the transposed vector of the key vector to obtain a central electric signal-rest electric signal feature association matrix, the central electric signal-rest electric signal feature association matrix is ​​divided by the square root of the length of the key vector by position to obtain a central electric signal-rest electric signal feature association modulation matrix;

[0030] Input the central electrical signal-other electrical signal feature correlation modulation matrix into a softmax function to obtain a central electrical signal-other electrical signal feature correlation activation matrix;

[0031] The central electric signal-rest electric signal feature association activation matrix is ​​multiplied by the central electric signal value feature vector to obtain the electric signal key granularity query matching implicit vector.

[0032] In a possible implementation, fusing the electrical signal key granularity query matching feature vector and the electrical signal key neighborhood granularity query matching feature vector to obtain a rotor rotation position query encoding representation vector as the rotor rotation position query encoding representation includes:

[0033] Respectively calculating the positional subtraction, positional dot product and positional addition between the electric signal key granularity query matching feature vector and the electric signal key neighborhood granularity query matching feature vector to obtain an electric signal key granularity-key neighborhood granularity difference vector, an electric signal key granularity-key neighborhood granularity dot product vector and an electric signal key granularity-key neighborhood granularity sum vector;

[0034] Cascading the electrical signal key granularity-key neighborhood granularity difference vector, the electrical signal key granularity-key neighborhood granularity dot product vector and the electrical signal key granularity-key neighborhood granularity sum vector to obtain an electrical signal key granularity-key neighborhood granularity cascade vector;

[0035] After performing one-dimensional convolution coding on the electrical signal key granularity-key neighborhood granularity cascade vector, the obtained electrical signal key granularity-key neighborhood granularity convolution coding vector is subjected to local window-based maximum pooling processing to obtain the rotor rotation position query coding representation vector.

[0036] In a possible implementation, determining the rotor rotational position based on the rotor rotational position query coded representation includes: inputting the rotor rotational position query coded representation vector into a decoder to obtain a decoded value of the rotor rotational position.

[0037] Compared with the prior art, the present application has the following beneficial effects: in the process of judging the rotation position of the rotor, the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal of the motor stator winding are analyzed by using a signal processing and analysis algorithm based on deep learning, so as to capture the implicit features in each electrical signal, and after differential calculation of the central electrical signal feature and the other electrical signal features, the query coding representation of the rotor rotation position is obtained by feature query matching, thereby determining the rotor rotation position. In this way, accurate judgment of the rotor rotation position can be achieved through artificial intelligence and deep learning algorithms, rather than just through hardware components, reducing the influence of noise interference and changes in external conditions, thereby improving flexibility and generalization capabilities with a more intelligent DC brushless motor drive method. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0039] Figure 1 A schematic flowchart of a brushless DC motor driving method according to an embodiment of the present application.

[0040] Figure 2 This is a schematic flowchart of determining the rotation position of the rotor based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the stator winding of the motor and the center electrical signal of the center point of the three-phase coil in the DC brushless motor driving method of an embodiment of the present application.

[0041] Figure 3 In the DC brushless motor driving method of the embodiment of the present application, a feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature is used as a sequence of simulated key features, the central electrical signal feature is used as a query feature, and a gated query matching feature between the query feature and the sequence of simulated key features is calculated to obtain a schematic flow chart of the rotor rotation position query coding representation.

[0042] Figure 4 In the DC brushless motor driving method of an embodiment of the present application, the sequence of the simulated key feature vectors and the sequence of the simulated key local neighborhood feature vectors are used as the sequence of key vectors, and the central electrical signal value feature vector is used as the value vector, and the query vector, the value vector and the sequence of the key vector are subjected to local neighborhood implicit optimization and attention-based gating processing to obtain a schematic flowchart of the rotor rotation position query coding representation.

[0043] Figure 5 A schematic block diagram of a brushless DC motor drive system according to an embodiment of the present application. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0045] In the prior art, comparators are used to collect and analyze the center point (CT) signal from the three-phase coil of the motor stator winding and the electrical signals of each phase. Then, an induced voltage detection device is responsible for monitoring the zero-crossing point of the induced voltage generated by the coil in the non-powered state, so as to infer the exact position of the rotor. Then, the waveform synthesis unit in the system generates appropriate control signals based on the obtained rotor position data. After these signals are processed by level conversion, they are sent to the motor by the output device to start working. However, this method relies on traditional hardware components (such as comparators, induced voltage detectors, etc.) to detect the rotor position. This method is limited to directly comparing electrical signals, and may cause errors due to noise interference. When the motor runs at high speed or the signal quality is poor, this error is particularly significant. In addition, changes in the external environment, such as fluctuations in temperature or power supply voltage, will also affect the performance of hardware components, making the detection results unstable, thereby affecting the normal operation of the motor.

[0046] In view of the above technical problems, a DC brushless motor driving method is proposed in the technical solution of the present application. Figure 1 FIG. 1 is a schematic flow chart of a brushless DC motor driving method according to an embodiment of the present application. Figure 1 As shown, the brushless DC motor driving method includes: S1, obtaining a central electrical signal of a central point of a three-phase coil in a motor stator winding; S2, obtaining a first electrical signal, a second electrical signal and a third electrical signal of the three-phase coil in the motor stator winding; S3, judging a rotor rotation position based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the motor stator winding and the central electrical signal of the central point of the three-phase coil; S4, inputting the rotor rotation position and the regulated voltage signal into a waveform synthesizer for processing, so as to output a level signal to an output module through the waveform synthesizer to drive the motor to operate.

[0047] In the above-mentioned DC brushless motor driving method, S1 obtains the central electrical signal of the center point of the three-phase coil in the motor stator winding; S2 obtains the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the motor stator winding. In the present application, different electrical signals of the three-phase coil are obtained from the three-phase stator winding of the motor, and the contrast features with the central point electrical signals are further extracted from these signals, and then the coded representation describing the rotor position is obtained by feature query matching. This method not only simplifies the design of the motor and improves the accuracy and reliability of rotor position detection, but also realizes a more intelligent DC brushless motor driving method.

[0048] In particular, in the process of detecting and judging the rotational position of the rotor, it is crucial to filter out the influence of noise interference in order to extract effective features from the electrical signal of the motor winding. The technical concept of the present application is to analyze the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal of the motor stator winding by using a signal processing and analysis algorithm based on deep learning in the process of judging the rotational position of the rotor, so as to capture the implicit features in each electrical signal, and obtain the query coding representation of the rotor rotational position by feature query matching after differential calculation of the central electrical signal feature and other electrical signal features, thereby determining the rotational position of the rotor. In this way, accurate judgment of the rotor rotational position can be achieved through artificial intelligence and deep learning algorithms, rather than just through hardware components, reducing the influence of noise interference and changes in external conditions, thereby improving flexibility and generalization capabilities in a more intelligent DC brushless motor drive method.

[0049] Optionally, in one embodiment of the present application, Figure 2 This is a schematic flow chart of determining the rotation position of the rotor based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the stator winding of the motor and the central electrical signal of the center point of the three-phase coil in the brushless DC motor driving method of the embodiment of the present application. Figure 2 As shown, based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the stator winding of the motor and the central electrical signal of the center point of the three-phase coil, the rotor rotation position is judged, including: S31, after extracting features from the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal, the central electrical signal feature is subjected to feature differential comparison with the first electrical signal feature, the second electrical signal feature and the third electrical signal feature to obtain the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature; S32, taking the feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature as a sequence of simulated key features, taking the central electrical signal feature as a query feature, and calculating the gated query matching feature between the query feature and the sequence of simulated key features to obtain a rotor rotation position query coding representation; S33, determining the rotor rotation position based on the rotor rotation position query coding representation.

[0050] In the above-mentioned DC brushless motor driving method, the S31, after extracting the features of the central electric signal, the first electric signal, the second electric signal and the third electric signal, respectively compares the features of the central electric signal with the features of the first electric signal, the second electric signal and the third electric signal to obtain the first electric signal differential feature, the second electric signal differential feature and the third electric signal differential feature. It should be understood that in order to obtain the difference features between the electric signals, these difference features can reflect the position change of the rotor relative to the stator. By calculating the difference between these feature vectors, the change law of the electric signal in the motor when the rotor moves can be captured, especially the change of the electric signal feature vector pattern caused by the rotor approaching a certain stator winding position. The central electric signal feature vector represents a reference signal, while the other feature vectors reflect the actual operating state of the motor at different phases. When the central electric signal feature vector is differentially compared with other electric signal feature vectors, the feature changes between different electric signals can be revealed, especially the phase difference between them can reflect the position of the rotor inside the motor. These differential eigenvectors (first electrical signal differential eigenvector, second electrical signal differential eigenvector, third electrical signal eigenvector) provide instant information about the rotor motion, which helps to understand and monitor the dynamic conditions inside the motor.

[0051] Optionally, in one embodiment of the present application, after feature extraction is performed on the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal, the central electrical signal feature is respectively compared with the first electrical signal feature, the second electrical signal feature and the third electrical signal feature to obtain the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature, including: inputting the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal into an electrical signal feature extractor based on a hole convolution layer to obtain a central electrical signal feature vector as the central electrical signal feature, the second electrical signal feature and the third electrical signal feature. An electric signal feature vector is used as the first electric signal differential feature, a second electric signal feature vector is used as the second electric signal feature, and a third electric signal feature vector is used as the third electric signal feature; the differential features between the central electric signal feature vector and each of the first electric signal feature vector, the second electric signal feature vector, and the third electric signal feature vector are calculated to obtain the first electric signal differential feature vector as the first electric signal differential feature, the second electric signal differential feature vector as the second electric signal differential feature, and the third electric signal differential feature vector as the third electric signal differential feature.

[0052] Specifically, in the technical solution of the present application, in the process of judging the rotation position of the rotor, first, the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal are input into an electrical signal feature extractor based on a hole convolution layer for feature capture, so as to respectively extract the electrical signal implicit feature information in the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal, thereby obtaining a central electrical signal feature vector, a first electrical signal feature vector, a second electrical signal feature vector and a third electrical signal feature vector.

[0053] Here, it should be understood that dilated convolution is a convolution operation that inserts a fixed number of zeros between the parameters of the convolution kernel (i.e., increases the sampling rate) so that the model can perceive a larger receptive field without increasing the parameters. This is particularly useful when processing sequence data because it can capture longer-distance dependencies without losing resolution.

[0054] Optionally, in one embodiment of the present application, the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal are input into an electrical signal feature extractor based on a dilated convolution layer to obtain a central electrical signal feature vector as the central electrical signal feature and the first electrical signal feature vector, including: constructing a feature extractor model based on a dilated convolution layer. This model will include one or more dilated convolution layers, which can effectively capture patterns at different scales, which is very important for analyzing complex patterns in electrical signals. The dilated convolution layer can expand its receptive field by adding holes between its kernels without significantly increasing the computational burden. The prepared central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal are input into the feature extractor respectively. This means that each signal will pass through the feature extractor separately and extract features using the same model architecture. When the signal passes through the dilated convolution layer, these layers will automatically identify and capture important features in the signal. Each signal will pass through the same dilated convolution layer, but due to different inputs, the final feature vectors will also be different. These feature vectors contain implicit feature information in the original signal and can be used for further analysis or as input for subsequent tasks. For each input signal, the feature extractor will output a corresponding feature vector. Therefore, four feature vectors will be obtained in the end: the central electric signal feature vector, the first electric signal feature vector, the second electric signal feature vector and the third electric signal feature vector.

[0055] Then, considering that when the rotor of the motor is rotating, the electrical signal in the stator winding will change, these changes can be characterized by the electrical signal feature vector. The differential features between the central electrical signal feature vector and other electrical signal feature vectors provide instant information about the movement of the rotor, which can help understand the dynamic situation inside the motor. For example, if the rotor is approaching the position of a stator winding, the electrical signal feature vector associated with the winding will show a certain pattern of change, and this change can be captured by calculating the differential features. Therefore, in the technical solution of the present application, the differential features between the central electrical signal feature vector and each of the electrical signal feature vectors in the first electrical signal feature vector, the second electrical signal feature vector and the third electrical signal feature vector are further calculated to obtain the first electrical signal differential feature vector, the second electrical signal differential feature vector, and the third electrical signal feature vector. By calculating the difference (i.e., differential features) between the central electrical signal feature vector and each of the other electrical signal feature vectors, this difference actually reflects the characteristic changes between different electrical signals, especially when there is a phase difference between the central electrical signal and other electrical signals, this phase difference can reflect the relative position of the rotor inside the motor.

[0056] Optionally, in one embodiment of the present application, the central electric signal feature vector is subtracted from each of the first electric signal feature vector, the second electric signal feature vector, and the third electric signal feature vector, and the result obtained is the differential feature vector. For example, the first electric signal differential feature vector is obtained by calculating the central electric signal feature vector minus the first electric signal feature vector; similarly, the second electric signal differential feature vector and the third electric signal differential feature vector can be obtained.

[0057] In the above-mentioned DC brushless motor driving method, in S32, the feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature is used as a sequence of imitation key features, and the central electrical signal feature is used as a query feature, and the gated query matching feature between the query feature and the sequence of the imitation key features is calculated to obtain the rotor rotation position query encoding representation. It should be understood that the central electrical signal feature vector represents a reference signal when the motor is running, and other feature vectors, such as the first electrical signal feature vector, the second electrical signal feature vector, the third electrical signal feature vector, the first electrical signal differential feature vector, the second electrical signal differential feature vector and the third electrical signal differential feature vector reflect the actual operating state of the motor at different phases, including different coil electrical signal features of the three-phase coils in the stator winding of the motor and the difference electrical signal difference features between the reference signal. By using a variety of electrical signal feature vectors (original feature vectors and their differential feature vectors), a multi-level data perspective can be provided, so that the model can analyze the relationship between the signal and the central electrical signal from different angles, thereby more comprehensively understanding the rotor position information. Based on this, in the technical solution of the present application, the feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature is further used as a sequence of simulated key features, and the central electrical signal feature is used as a query feature, and the gated query matching feature between the query feature and the sequence of simulated key features is calculated to obtain a rotor rotation position query encoding representation.

[0058] Optionally, in one embodiment of the present application, Figure 3 In the DC brushless motor driving method of the embodiment of the present application, the feature sequence consisting of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature is used as a sequence of simulated key features, the central electrical signal feature is used as a query feature, and the gated query matching feature between the query feature and the sequence of simulated key features is calculated to obtain a schematic flow chart of the query encoding representation of the rotor rotation position. Figure 3As shown, a feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature is used as a sequence of analog key features, the central electrical signal feature is used as a query feature, and a gated query matching feature between the query feature and the sequence of analog key features is calculated to obtain a rotor rotation position query encoding representation, including: S321, arranging the first electrical signal feature vector, the second electrical signal feature vector, the third electrical signal feature vector, the first electrical signal differential feature vector, the second electrical signal differential feature vector and the third electrical signal differential feature vector into a sequence of analog key feature vectors as the S322, performing one-dimensional convolution coding on the sequence of the simulative key feature vectors to obtain a sequence of simulative key local neighborhood feature vectors, wherein the scale of the one-dimensional convolution coding is equal to an integer multiple of the length of the simulative key feature vector; S323, processing the central electric signal feature vector using a value embedding coding matrix to obtain a central electric signal value feature vector; S324, using the sequence of the simulative key feature vectors and the sequence of the simulative key local neighborhood feature vectors as the sequence of key vectors, and using the central electric signal value feature vector as the value vector, respectively, performing local neighborhood implicit optimization and attention-based gating processing on the query vector, the value vector and the sequence of the key vector to obtain the rotor rotation position query encoding representation.

[0059] Optionally, in one embodiment of the present application, the central electrical signal feature vector is processed using a value embedding coding matrix to obtain the central electrical signal value feature vector, including: after matrix multiplication of the central electrical signal feature vector and the value embedding coding matrix, the obtained central electrical signal feature value time series embedding vector and the value bias vector are added by position to obtain the central electrical signal value feature vector.

[0060] Optionally, in one embodiment of the present application, Figure 4 In the brushless DC motor driving method of the embodiment of the present application, the sequence of the simulated key feature vectors and the sequence of the simulated key local neighborhood feature vectors are used as the sequence of key vectors, and the central electrical signal value feature vector is used as the value vector, and the query vector, the value vector and the sequence of the key vector are subjected to local neighborhood implicit optimization and attention-based gating processing to obtain a schematic flow chart of the rotor rotation position query encoding representation. Figure 4As shown, the sequence of the simulated key feature vectors and the sequence of the simulated key local neighborhood feature vectors are respectively used as the sequence of key vectors, and the central electric signal value feature vector is used as the value vector, and the sequence of the query vector, the value vector and the key vector are subjected to local neighborhood implicit optimization and attention-based gating processing to obtain the rotor rotation position query coding representation, including: S3241, using the central electric signal feature vector as the query vector, the central electric signal value feature vector as the value vector and the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector and the third electric signal differential feature vector as the sequence of key vectors, and inputting the sequence of the query vector, the value vector and the key vector into a first multi-head attention query module based on a converter structure to obtain an electric signal key granularity query matching. S3242, using the central electric signal feature vector as the query vector, the central electric signal value feature vector as the value vector and the sequence of the imitation key local neighborhood feature vectors as the sequence of key vectors, inputting the query vector, the value vector and the sequence of the key vector into a second multi-head attention query module based on a converter structure to obtain a sequence of electric signal key neighborhood granularity query matching implicit vectors; S3243, calculating the positional mean vector of the sequence of the electric signal key granularity query matching implicit vectors and the sequence of the electric signal key neighborhood granularity query matching implicit vectors to obtain an electric signal key granularity query matching feature vector and an electric signal key neighborhood granularity query matching feature vector; S3244, fusing the electric signal key granularity query matching feature vector and the electric signal key neighborhood granularity query matching feature vector to obtain a rotor rotation position query encoding representation vector as the rotor rotation position query encoding representation.

[0061] Optionally, in one embodiment of the present application, the central electric signal feature vector is used as the query vector, the central electric signal value feature vector is used as the value vector, and the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector and the third electric signal differential feature vector are used as a sequence of key vectors, and the sequence of the query vector, the value vector and the key vector is input into a first multi-head attention query module based on a converter structure to obtain a sequence of electric signal key granularity query matching implicit vectors, including: calculating the central electric signal feature vector; After obtaining the central electric signal-rest electric signal feature association matrix by multiplying the feature vector and the transposed vector of the key vector, the central electric signal-rest electric signal feature association matrix and the square root of the length of the key vector are divided by position to obtain the central electric signal-rest electric signal feature association modulation matrix; the central electric signal-rest electric signal feature association modulation matrix is ​​input into the softmax function to obtain the central electric signal-rest electric signal feature association activation matrix; the central electric signal-rest electric signal feature association activation matrix is ​​multiplied by the central electric signal value feature vector to obtain the electric signal key granularity query matching implicit vector.

[0062] Optionally, in one embodiment of the present application, the electrical signal key granularity query matching feature vector and the electrical signal key neighborhood granularity query matching feature vector are merged to obtain a rotor rotation position query encoding representation vector as the rotor rotation position query encoding representation, including: respectively calculating the positional subtraction, positional dot multiplication and positional addition between the electrical signal key granularity query matching feature vector and the electrical signal key neighborhood granularity query matching feature vector to obtain an electrical signal key granularity-key neighborhood granularity difference vector, an electrical signal key granularity-key neighborhood granularity dot product vector and An electrical signal key granularity-key neighborhood granularity sum vector; cascading the electrical signal key granularity-key neighborhood granularity difference vector, the electrical signal key granularity-key neighborhood granularity dot product vector and the electrical signal key granularity-key neighborhood granularity sum vector to obtain an electrical signal key granularity-key neighborhood granularity cascade vector; performing one-dimensional convolution encoding on the electrical signal key granularity-key neighborhood granularity cascade vector, and then performing local window-based maximum pooling processing on the obtained electrical signal key granularity-key neighborhood granularity convolution encoding vector to obtain the rotor rotation position query encoding representation vector.

[0063] In summary, in the embodiment of the present application, a feature sequence consisting of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature is used as a sequence of simulated key features, and the central electrical signal feature is used as a query feature. The gated query matching feature between the query feature and the sequence of simulated key features is calculated using the following gated query matching formula to obtain a query encoding representation of the rotor rotation position;

[0064] Wherein, the gated query matching formula is:

[0065] K={k1,k2,...,k i ,...,k n}

[0066] K'=Conv 1×l ({k1,k2,...,k i ,...,k n})

[0067] K'={k1',k2',...,k i ',...,k m '}

[0068] v v =v q W q +b q

[0069]

[0070] Wherein, K represents a sequence of key-like feature vectors consisting of the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector and the third electric signal differential feature vector, k1, k2, ..., k i ,...,k n is the first, second, ..., i-th, ..., n-th analog key feature vector in the sequence of analog key feature vectors, n is the number of feature vectors in the sequence of analog key feature vectors, Conv 1×l (·) is a one-dimensional convolutional code, l is an integer multiple of the length of the key-like feature vector, K' is a sequence of key-like local neighborhood feature vectors, k1', k2', ..., k i ',...,k m 'The first, second, ..., i-th, ..., m-th local neighborhood feature vectors of the sequence of the local neighborhood feature vectors of the simulated key, where m is the number of feature vectors in the sequence of the local neighborhood feature vectors of the simulated key, vq is the central electrical signal eigenvector, W q is the value embedding encoding matrix, b q is the value bias vector, v v is the central electrical signal value eigenvector, k i T and k i ' T k i and k i ', d1 and d2 are the lengths of the key-like feature vector and the key-like local neighborhood feature vector, softmax(·) is the softmax function, v ki is the ith electric signal key granularity query matching implicit vector in the sequence of electric signal key granularity query matching implicit vectors, v ni is the ith electric signal key neighborhood granularity query matching implicit vector in the sequence of the electric signal key neighborhood granularity query matching implicit vector, v k is the electronic signal key granularity query matching feature vector, v n is the electric signal key neighborhood granularity query matching feature vector, ⊙ and are position-wise subtraction, position-wise multiplication, and position-wise addition, respectively. [·,·,·] is a cascade process. conv1D(·) is a one-dimensional convolutional encoding operation. MaxPool(·) is a maximum pooling operation. p is the rotor rotation position query encoding representation vector.

[0071] Specifically, the processing process of the gated query matching feature includes: first, the first electrical signal feature vector, the second electrical signal feature vector, the third electrical signal feature vector, the first electrical signal differential feature vector, the second electrical signal differential feature vector and the third electrical signal differential feature vector are arranged into a sequence of quasi-key feature vectors as the sequence of quasi-key features, and the sequence of quasi-key feature vectors is one-dimensionally convolutionally encoded to obtain a sequence of quasi-key local neighborhood semantic association feature vectors. The main purpose of applying a one-dimensional convolution operation to the sequence of quasi-key feature vectors is to capture the local neighborhood information in the sequence. Through the convolution operation, the relationship between adjacent elements in the feature vector can be extracted, which helps to enhance the understanding of the correlation between different electrical signal features and electrical signal difference features. The size of the convolution kernel is set to an integer multiple of the length of the quasi-key feature vector to ensure that the contextual information within a specific range can be captured, so as to better reflect the semantic association of the local neighborhood. This helps to more accurately perform matching queries of central electrical signal features and other electrical signal features in subsequent steps, thereby detecting and judging the rotation position of the rotor. Next, the central electric signal feature vector as the query feature vector is processed using the value embedding coding matrix to generate a new value feature vector, namely the central electric signal value feature vector, so that the query feature vector is converted into a form that is more suitable as the value vector in the attention mechanism. Through embedding coding, the central electric signal features can be mapped to a higher-dimensional space, so that in the subsequent attention mechanism, these feature vectors can better express the matching and query intention of the central electric signal, and effectively interact with other feature vectors, which is conducive to more accurately capturing information about the rotor position. Further, the sequence of the key-like feature vector and the sequence of the key-like local neighborhood feature vector are used as the sequence of key vectors, and the central electric signal value feature vector is used as the value vector, and the query vector, value vector and key vector are input into the multi-head attention query module based on the converter structure, which is to pay attention to the relationship between the central electric signal features and each key vector from different perspectives. Through the multi-head attention mechanism, the attention distribution can be calculated simultaneously in different representation subspaces, so as to capture the multiple dependencies between the query features of the central electric signal and the features of other electric signals. The first multi-head attention query module generates a key-granularity query matching implicit vector sequence, focusing on the characteristics of the key vector itself, that is, the inherent information of other electrical signal features; while the second multi-head attention query module generates a key-neighborhood granularity query matching implicit vector sequence, focusing on the characteristics of the key vector and its neighborhood. This helps to comprehensively consider local and global information, improve the accuracy of feature query matching, and provide a basis for subsequent rotor rotation position detection.Next, the positional mean vector of the sequence of the electric signal key granularity query matching implicit vectors and the sequence of the electric signal key neighborhood granularity query matching implicit vectors is calculated in order to aggregate the electric signal query matching feature information at different granularity levels. By calculating the mean vector, a representation that integrates the features at different granularity levels can be obtained to ensure that the model not only takes into account the information of the single feature vector itself, but also takes into account their interactions in the local neighborhood. This can improve the accuracy of the rotor position estimation. Finally, the electric signal key granularity query matching feature vector and the electric signal key neighborhood granularity query matching feature vector are fused into the final rotor rotation position query encoding representation vector to integrate the information at two different granularity levels, ensuring that the final rotor rotation position query encoding representation can fully reflect the query information of the central electric signal and the multiple relationships between the key vectors of other electric signals, thereby more accurately inferring the rotation position of the rotor.

[0072] In the above-mentioned DC brushless motor driving method, the S33 determines the rotor rotation position based on the rotor rotation position query code representation. The rotation position of the rotor is determined by decoding regression using the characteristics of the rotor rotation position query code representation. In this way, accurate judgment of the rotor rotation position can be achieved through artificial intelligence and deep learning algorithms, rather than just through hardware components, reducing the influence of noise interference and changes in external conditions, thereby improving flexibility and generalization capabilities in a more intelligent DC brushless motor driving method.

[0073] Optionally, in one embodiment of the present application, determining the rotor rotation position based on the rotor rotation position query coded representation includes: inputting the rotor rotation position query coded representation vector into a decoder to obtain a decoded value of the rotor rotation position.

[0074] In a preferred embodiment of the present application, the rotor rotation position query code representation vector is input into a decoder to obtain a decoded value of the rotor rotation position, including:

[0075] Calculating the sum of the absolute values ​​of each eigenvalue of the rotor rotation position query code representation vector to obtain a first rotor rotation position query code and modulation value, and calculating the square root of the sum of the squares of each eigenvalue of the rotor rotation position query code representation vector to obtain a second rotor rotation position query code and modulation value;

[0076] After performing dot-wise subtraction on the rotor rotation position query code representation vector and the second rotor rotation position query code and modulation value, perform dot-wise multiplication on the number of eigenvalues ​​of the rotor rotation position query code representation vector and the reciprocal of the first rotor rotation position query code and modulation value, respectively, and take the reciprocal of each eigenvalue to obtain a first rotor rotation position query code phase conversion vector;

[0077] After performing dot-wise subtraction on the rotor rotation position query code representation vector and the first rotor rotation position query code and modulation value, perform dot-wise multiplication with the square root of the number of eigenvalues ​​of the rotor rotation position query code representation vector and the reciprocal of the second rotor rotation position query code and modulation value, respectively, and take the reciprocal of each eigenvalue to obtain a second rotor rotation position query code phase conversion vector;

[0078] Subtract the dot product vector of the second rotor rotation position query encoding phase conversion vector and the weighted hyperparameter from the first rotor rotation position query encoding phase conversion vector to obtain an optimized rotor rotation position query encoding representation vector;

[0079] The optimized rotor rotation position query code representation vector is input into a decoder to obtain a decoded value of the rotor rotation position.

[0080] Here, the rotor rotation position query code representation vector, denoted as V, is optimally represented as:

[0081]

[0082] v i ∈V∈R n

[0083] Wherein, V is the rotor rotation position query code representation vector, R represents a real number set, and v i represents the eigenvalue of the i-th position of the rotor rotation position query code representation vector, n represents the number of eigenvalues ​​of the rotor rotation position query code representation vector, α represents the first rotor rotation position query code and modulation value, β represents the second rotor rotation position query code and modulation value, ⊙ represents the position point multiplication, Indicates positional subtraction, (·) ⊙-1 represents the inverse of each eigenvalue of the calculated eigenvector, V1 represents the first rotor rotation position query encoding phase conversion vector, V2 represents the second rotor rotation position query encoding phase conversion vector, ω represents a weighted hyperparameter, and V' represents the optimized rotor rotation position query encoding representation vector.

[0084] That is, considering that the first electric signal differential feature vector, the second electric signal differential feature vector, and the third electric signal feature vector are differential features between the central electric signal feature vector and each of the first electric signal feature vector, the second electric signal feature vector, and the third electric signal feature vector, thus, when the central electric signal feature vector is used as the query feature vector and the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector, and the third electric signal differential feature vector are used as the feature vector. When the sequence of feature vectors is used as the sequence of analog key feature vectors, and the query feature vector and the sequence of analog key feature vectors are input into the attention gated query network based on 1D convolution local neighborhood implicit optimization, considering that the first electric signal difference feature vector, the second electric signal difference feature vector, and the third electric signal difference feature vector share a common component (i.e., the central electric signal feature vector), this will cause some feature redundancy within the sequence of analog key feature vectors. At the same time, during the gated query of the query feature vector and the analog key feature vector, some features in the query feature vector and the sequence of analog key feature vectors have overlapping responses. Based on the feature redundancy and overlapping response differences, the rotor rotation position query encoding representation vector will also have cross-semantic space aggregation differences. Therefore, it is expected to improve the detail semantic aggregation expression effect of the rotor rotation position query encoding representation vector based on cross-semantic space aggregation.

[0085] Based on this, in the preferred embodiment, the difference of the characteristic value of the rotor rotation position query code representation vector relative to the feature set of the whole vector of the rotor rotation position query code representation vector and the difference of the modulation representation are used as the semantic change intensity information, and the phase-like conversion corresponding to the position-based intensity modulation is performed through the different and modulation representation forms, so that the aggregation enhancement of the semantic change phase perception can improve the axial aggregation receptive field along the feature aggregation direction by performing the space translation operation based on the alternating stacking under the vector set scale balance of the rotor rotation position query code representation vector, so as to improve the perception effect of the aggregate semantics of the rotor rotation position query code representation vector for the detail semantic change, so as to improve the expression effect of the rotor rotation position query code representation vector, and improve the accuracy of the decoded value of the rotor rotation position obtained by the input decoder. In this way, the rotor rotation position can be determined more accurately, and the influence of noise interference and changes in external conditions is reduced, thereby improving the flexibility and generalization ability in a more intelligent DC brushless motor drive mode.

[0086] In summary, according to the DC brushless motor driving method of the embodiment of the present application, in the process of judging the rotation position of the rotor, the signal processing and analysis algorithm based on deep learning is used to analyze the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal of the motor stator winding, so as to capture the implicit features in each electrical signal, and after differential calculation of the central electrical signal feature and the other electrical signal features, the query coding representation of the rotor rotation position is obtained by feature query matching, thereby determining the rotor rotation position. In this way, accurate judgment of the rotor rotation position can be achieved through artificial intelligence and deep learning algorithms, rather than just through hardware components, reducing the influence of noise interference and changes in external conditions, thereby improving flexibility and generalization capabilities with a more intelligent DC brushless motor driving method.

[0087] Figure 5 FIG. 1 is a schematic block diagram of a brushless DC motor drive system according to an embodiment of the present application. Figure 5 As shown, the brushless DC motor drive system 100 includes: a center electrical signal acquisition module 110, which is used to acquire the center electrical signal of the center point of the three-phase coil in the motor stator winding; a three-phase coil electrical signal acquisition module 120, which is used to acquire the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the motor stator winding; a rotor rotation position determination module 130, which is used to determine the rotor rotation position based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the motor stator winding and the center electrical signal of the center point of the three-phase coil; a motor operation drive module 140, which is used to input the rotor rotation position and the voltage stabilization signal into a waveform synthesizer for processing, so as to output a level signal to an output module through the waveform synthesizer to drive the motor to operate.

[0088] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned brushless DC motor driving method have been described in detail above. Figures 1 to 4 The description of the DC brushless motor drive system has been described in detail, and therefore, its repeated description will be omitted.

[0089] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above embodiments are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

[0090] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0091] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0092] It is obvious to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present application. Any reference to a figure in a claim should not be regarded as limiting the claim to which it relates.

[0093] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the device claim can also be implemented by one unit through software or hardware.

[0094] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present application.

Claims

1. A method for driving a brushless DC motor, characterized in that: include: Obtaining the central electrical signal of the center point of the three-phase coil in the motor stator winding; Acquire a first electrical signal, a second electrical signal and a third electrical signal of a three-phase coil in a stator winding of the motor; Determine the rotation position of the rotor based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the stator winding of the motor and the central electrical signal of the central point of the three-phase coil; The rotor rotation position and the voltage stabilization signal are input into a waveform synthesizer for processing, so as to output a level signal to an output module through the waveform synthesizer to drive the motor to operate; Wherein, based on the first electrical signal, the second electrical signal and the third electrical signal of the three-phase coil in the stator winding of the motor and the central electrical signal of the center point of the three-phase coil, the rotor rotation position is judged, including: after feature extraction of the central electrical signal, the first electrical signal, the second electrical signal and the third electrical signal, feature differential comparison of the central electrical signal feature with the first electrical signal feature, the second electrical signal feature and the third electrical signal feature respectively to obtain the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature; using the feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature as a sequence of simulated key features, using the central electrical signal feature as a query feature, calculating the gated query matching feature between the query feature and the sequence of simulated key features to obtain a rotor rotation position query coding representation; determining the rotor rotation position based on the rotor rotation position query coding representation.

2. The brushless DC motor driving method according to claim 1, characterized in that: After extracting features of the central electrical signal, the first electrical signal, the second electrical signal, and the third electrical signal, the central electrical signal feature is respectively compared with the first electrical signal feature, the second electrical signal feature, and the third electrical signal feature to obtain a first electrical signal differential feature, a second electrical signal differential feature, and a third electrical signal differential feature, including: Inputting the central electric signal, the first electric signal, the second electric signal and the third electric signal into an electric signal feature extractor based on a dilated convolutional layer to obtain a central electric signal feature vector as the central electric signal feature, a first electric signal feature vector as the first electric signal differential feature, a second electric signal feature vector as the second electric signal feature and a third electric signal feature vector as the third electric signal feature; Calculate the differential features between the central electric signal feature vector and each of the first electric signal feature vector, the second electric signal feature vector and the third electric signal feature vector to obtain the first electric signal differential feature vector as the first electric signal differential feature, the second electric signal differential feature vector as the second electric signal differential feature, and the third electric signal differential feature vector as the third electric signal differential feature.

3. The brushless DC motor driving method according to claim 2, characterized in that: The method comprises: using a feature sequence composed of the first electrical signal feature, the second electrical signal feature, the third electrical signal feature, the first electrical signal differential feature, the second electrical signal differential feature and the third electrical signal differential feature as a sequence of simulated key features, using the central electrical signal feature as a query feature, and calculating a gated query matching feature between the query feature and the sequence of simulated key features to obtain a query encoding representation of the rotor rotation position, including: Arranging the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector and the third electric signal differential feature vector into a sequence of simulated key feature vectors as the sequence of simulated key features; Performing one-dimensional convolution coding on the sequence of the simulated key feature vectors to obtain a sequence of simulated key local neighborhood feature vectors, wherein the scale of the one-dimensional convolution coding is equal to an integer multiple of the length of the simulated key feature vector; Processing the central electrical signal feature vector using a value embedding coding matrix to obtain a central electrical signal value feature vector; The sequence of the simulated key feature vectors and the sequence of the simulated key local neighborhood feature vectors are respectively used as the sequence of key vectors, and the central electrical signal value feature vector is used as the value vector. Local neighborhood implicit optimization and attention-based gating processing are performed on the query vector, the value vector and the sequence of the key vector to obtain the rotor rotation position query encoding representation.

4. The brushless DC motor driving method according to claim 3, characterized in that: The central electric signal feature vector is processed using a value embedding coding matrix to obtain the central electric signal value feature vector, including: after matrix multiplication of the central electric signal feature vector and the value embedding coding matrix, the obtained central electric signal feature value time series embedding vector and the value bias vector are added by position to obtain the central electric signal value feature vector.

5. The brushless DC motor driving method according to claim 4, characterized in that: The sequence of the simulated key feature vectors and the sequence of the simulated key local neighborhood feature vectors are respectively used as the sequence of key vectors, and the central electrical signal value feature vector is used as the value vector, and the query vector, the value vector and the sequence of the key vector are subjected to local neighborhood implicit optimization and attention-based gating processing to obtain the rotor rotation position query encoding representation, including: Using the central electric signal feature vector as a query vector, the central electric signal value feature vector as a value vector, and using the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector, and the third electric signal differential feature vector as a sequence of key vectors, inputting the sequence of the query vector, the value vector, and the key vector into a first multi-head attention query module based on a converter structure to obtain a sequence of electric signal key granularity query matching implicit vectors; Taking the central electric signal feature vector as a query vector, the central electric signal value feature vector as a value vector and the sequence of the key-like local neighborhood feature vectors as a sequence of key vectors, inputting the query vector, the value vector and the sequence of the key vectors into a second multi-head attention query module based on a converter structure to obtain a sequence of electric signal key neighborhood granularity query matching implicit vectors; Calculating the positional mean vector of the sequence of the electric signal key granularity query matching implicit vectors and the sequence of the electric signal key neighborhood granularity query matching implicit vectors to obtain the electric signal key granularity query matching feature vector and the electric signal key neighborhood granularity query matching feature vector; The electrical signal key granularity query matching feature vector and the electrical signal key neighborhood granularity query matching feature vector are fused to obtain a rotor rotation position query encoding representation vector as the rotor rotation position query encoding representation.

6. The brushless DC motor driving method according to claim 5, characterized in that: Taking the central electric signal feature vector as a query vector, the central electric signal value feature vector as a value vector, and taking the first electric signal feature vector, the second electric signal feature vector, the third electric signal feature vector, the first electric signal differential feature vector, the second electric signal differential feature vector, and the third electric signal differential feature vector as a sequence of key vectors, inputting the sequence of the query vector, the value vector, and the key vector into a first multi-head attention query module based on a converter structure to obtain a sequence of electric signal key granularity query matching implicit vectors, including: After calculating the product between the central electric signal feature vector and the transposed vector of the key vector to obtain a central electric signal-rest electric signal feature association matrix, the central electric signal-rest electric signal feature association matrix is ​​divided by the square root of the length of the key vector by position to obtain a central electric signal-rest electric signal feature association modulation matrix; Input the central electrical signal-other electrical signal feature correlation modulation matrix into a softmax function to obtain a central electrical signal-other electrical signal feature correlation activation matrix; The central electric signal-rest electric signal feature association activation matrix is ​​multiplied by the central electric signal value feature vector to obtain the electric signal key granularity query matching implicit vector.

7. The brushless DC motor driving method according to claim 6, characterized in that: The electrical signal key granularity query matching feature vector and the electrical signal key neighborhood granularity query matching feature vector are merged to obtain a rotor rotation position query code representation vector as the rotor rotation position query code representation, including: Respectively calculating the positional subtraction, positional dot product and positional addition between the electric signal key granularity query matching feature vector and the electric signal key neighborhood granularity query matching feature vector to obtain an electric signal key granularity-key neighborhood granularity difference vector, an electric signal key granularity-key neighborhood granularity dot product vector and an electric signal key granularity-key neighborhood granularity sum vector; Cascading the electrical signal key granularity-key neighborhood granularity difference vector, the electrical signal key granularity-key neighborhood granularity dot product vector and the electrical signal key granularity-key neighborhood granularity sum vector to obtain an electrical signal key granularity-key neighborhood granularity cascade vector; After performing one-dimensional convolution coding on the electrical signal key granularity-key neighborhood granularity cascade vector, the obtained electrical signal key granularity-key neighborhood granularity convolution coding vector is subjected to local window-based maximum pooling processing to obtain the rotor rotation position query coding representation vector.

8. The brushless DC motor driving method according to claim 7, characterized in that: Determining the rotor rotational position based on the rotor rotational position query coded representation includes: inputting the rotor rotational position query coded representation vector into a decoder to obtain a decoded value of the rotor rotational position.

Images