Wind power system permanent magnet synchronous motor turn-to-turn short circuit fault diagnosis method considering wind speed fluctuation

Abstract

The invention discloses a wind power system permanent magnet synchronous motor turn-to-turn short circuit fault diagnosis method considering wind speed fluctuation. The method comprises the following steps: firstly, selecting a current space vector effective value as a diagnosis variable, and constructing a permanent magnet synchronous motor digital twinborn model by establishing a non-electrical quantity expression of a stator current space vector; the model outputs a calculated value of a stator current space vector effective value; and then a self-adaptive threshold value based on the wind speed change rate is designed based on analysis of a residual error between a measured value of a current space vector effective value of the permanent magnet synchronous motor and a model calculation value. And finally, the permanent magnet synchronous motor is monitored in real time in combination with a digital twin model and an adaptive threshold value, a fault phase is determined through a phase current effective value, and the severity is estimated by comparing a current space vector measurement value with a calculation value.

Description

Technical Field

[0001] The present invention relates to the field of fault maintenance of wind power systems, and specifically to a method for diagnosing inter-turn short circuit faults of permanent magnet synchronous motors in wind power systems considering wind speed fluctuations. Background Technique

[0002] In recent years, due to the characteristics of large reserves, wide distribution and easy access of wind energy, the installed capacity of wind power has achieved rapid growth and played an important role in promoting the global energy transformation. Among them, permanent magnet direct drive wind turbine generators, with their unique technical advantages such as no need for excitation windings, direct drive, multi-pole low-speed operation and excellent low voltage ride-through ability, have an expanding market share year by year. Especially in the field of large-power offshore wind power generation, they have become an indispensable main model.

[0003] However, the offshore operating environment is harsh and the meteorological conditions are complex and changeable, which poses a severe challenge to the long-term stable operation of permanent magnet synchronous generators. During the long-term operation process, permanent magnet synchronous generators are prone to mechanical faults, electrical faults and permanent magnet faults. Among them, electrical faults are most commonly inter-turn short circuit faults. Inter-turn short circuit faults are highly destructive and will generate large eddy currents in the short circuit loop. If not detected and processed in time, they may further evolve into more serious faults such as phase-to-phase short circuit, end short circuit or even phase-to-ground short circuit, ultimately leading to the paralysis of the wind power system and even posing a potential threat to the safety of personnel's lives. At the same time, fault diagnosis, as a pre-requisite technology for fault-tolerant operation, can provide necessary information for fault-tolerant operation in the case where the wind power system needs fault-tolerant operation. Therefore, diagnosing inter-turn short circuit faults and arranging preventive maintenance accordingly have very significant significance for improving the reliability and service life of the wind power system. Summary of the Invention

[0004] Object of the Invention. The object of the present invention is to provide a method for diagnosing inter-turn short circuit faults of permanent magnet synchronous motors in wind power systems considering wind speed fluctuations, so as to improve the operation reliability of offshore wind farms. This technology aims to achieve precise monitoring of inter-turn short circuit faults of permanent magnet synchronous motors in wind turbine generators.

[0005] Technical Solution. In order to achieve the above object of the invention, the present invention proposes a method for diagnosing inter-turn short circuit faults of permanent magnet synchronous motors in wind power systems considering wind speed fluctuations. The method includes the following steps:

[0006] S1, Select the effective value of the stator current space vector as the diagnostic variable, and construct a digital twin model of the permanent magnet synchronous motor by establishing a non-electrical quantity expression of the stator current space vector. The model outputs the calculated value of the effective value of the stator current space vector.

[0007] S2. Establish the residual expression between the measured value and the calculated value of the effective value of the stator current space vector of the permanent magnet synchronous motor, and design an adaptive threshold based on the wind speed change rate;

[0008] S3. Combine the digital twin model and the adaptive threshold to monitor the permanent magnet synchronous motor in real time.

[0009] Further, the specific method of step S1 is as follows:

[0010] Step 1.1) Select the effective value of the stator current space vector as the diagnostic variable, establish the non-electrical quantity expression of the stator current space vector, and establish the connection between the mechanical system and the electrical system according to the law of conservation of momentum T w = T e + T loss where, T w represents the mechanical torque of the wind turbine:

[0011]

[0012] In the formula, ρ is the air density, R is the radius of the wind turbine blade, v w represents the wind speed, C p represents the wind energy utilization coefficient, ω is the rotational angular velocity of the wind turbine;

[0013] T e represents the electromagnetic torque, and the electromagnetic torque T p of the permanent magnet synchronous motor with the number of pole pairs n e is:

[0014]

[0015] In the formula, ψ f is the permanent magnet flux linkage; L d , L q represent the direct-axis and quadrature-axis inductances of the motor respectively; i d , i q represent the direct-axis and quadrature-axis currents of the motor respectively. According to the actual system, L d ≈ L q , i d = ki q and k is a coefficient greater than 0 and much less than 1. Simplify T e to:

[0016]

[0017] where, T loss represents the loss torque including iron loss, stray loss and mechanical loss:

[0018] T loss = C1 + C2ω (4)

[0019] Among them, C1 represents the sum of the constant parts of iron loss, mechanical loss, and stray loss, and C2 represents the coefficient of the variable part of iron loss;

[0020] Substitute Equations (1), (3), and (4) into T w = T e + T loss Combined with the magnitude of the current space vector, i.e., Equation (5), the non-electrical quantity expression of the magnitude of the current space vector I m is shown in Equation (6) as follows:

[0021]

[0022] According to Equation (6) and the effective value of the current space vector under the normal state of the motor, which is equal to I m , Select the wind speed v w and the rotational speed n r as the main input parameters of the digital twin model, and select the mechanical torque T w and the DC bus voltage U dc as the auxiliary input parameters;

[0023] Step 1.2) Design a digital twin model of a permanent magnet synchronous generator. Use a 1D-CNN with a depth of 1 to build the model. The model consists of a multi-channel one-dimensional convolutional layer and a fully connected layer. The input parameters X = {v w , n r , T w , U dc}, and the output is the calculated value of the effective value of the stator current space vector i.e., the output The input X first passes through a one-dimensional convolutional layer with 32 channels, and activation, batch normalization, and pooling operations are performed. Then, it passes through a one-dimensional convolutional layer with 64 channels for the same operations, and the output Y is obtained through 3 fully connected layers;

[0024] Step 1.3) Train and save the model. Obtain wind speed data through a wind speed measurement system, use MATLAB / Simulink to simulate and model, obtain the dataset {X, Y}, divide the training set and the validation set in an 8:2 ratio after cleaning the data, use the Adam optimizer, with an initial learning rate of 0.0001, a batch size of 64, and iterate 50 times. If the mean square error of the model meets the preset requirements, then save it. Otherwise, adjust the hyperparameters and retrain.

[0025] Furthermore, the specific method of Step S2 is as follows:

[0026] Step 2.1) Represent the current space vector in the form of the cube of the wind speed according to Equation (6):

[0027]

[0028] Among them, a1 and a2 are constants used to characterize the relationship between the amplitude of the current space vector and the cube of the wind speed. The specific expressions are shown in Equations (8) and (9) as follows:

[0029]

[0030] Step 2.2) Regarding the smoothing effect of the PI controller and the moment of inertia on I under wind speed fluctuations, Equation (7) is extended to: m The smoothing degree of I under wind speed fluctuations, denoted as a', ranges from 0.01a1 to 0.1a1;

[0031]

[0032] where a' represents the smoothing degree of the PI controller and the moment of inertia on I under wind speed fluctuations, and its value is between 0.01a1 and 0.1a1; m The smoothing degree of I under wind speed fluctuations, denoted as a', ranges from 0.01a1 to 0.1a1;

[0033] Step 2.3) Combining the normal state of the motor and Equation (10), the measured value of the effective value of the stator current space vector of the motor under the normal state is obtained and The residual expression between them is as follows:

[0034]

[0035] In the formula, ε represents the error between and under constant wind speed;

[0036] Step 2.4) Design an adaptive threshold based on the wind speed change rate. When the wind speed change rate is less than v th , the diagnostic variable threshold is selected as ε th ; when the wind speed change rate is greater than or equal to v th , the diagnostic variable threshold is selected as where ε th represents the current residual threshold under steady state, and v th represents the wind speed change rate threshold.

[0037] Furthermore, the specific method of Step S3 is as follows:

[0038] Step 3.1) Fault monitoring and judgment: Measure the effective value of the motor current space vector in real time, and use the digital twin model to calculate the effective value of the space vector under the same operating conditions to obtain the residual; Design a time window Δt. If the time span during which the residual exceeds the adaptive threshold within Δt is greater than Δt / 2, it is determined that there is an inter-turn short circuit fault; otherwise, the motor is operating normally; If the motor is normal, collect the operating data and update the model regularly; If a fault occurs, an alarm is issued and Steps 3.2 and 3.3 are entered;

[0039] Step 3.2) Fault phase location: Locate the fault phase by measuring the effective values of the three-phase currents. Use the maximum value function to determine the phase where the maximum value lies. Design a time window Δt, and output the phase where the maximum value of the current with the largest proportion within Δt is located, which is the fault phase.

[0040] Step 3.3) Estimate the severity through the effective value of the space current. The specific formula is as follows

[0041]

[0042] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0043] The present invention discloses a method for diagnosing inter-turn short circuit faults of permanent magnet synchronous motors in a wind power system considering wind speed fluctuations. A method for diagnosing inter-turn short circuit faults of permanent magnet synchronous generators based on stator current space vectors, digital twin models, and adaptive thresholds is proposed. Innovatively, through the combined analysis of the mechanical system and the electrical system, the non-electrical quantity expression of the current space vector is derived, and based on this, the main parameters are selected to establish a digital twin model of a permanent magnet synchronous motor considering wind speed fluctuations. By analyzing the reason why the residual of the current space vector shows a negative correlation with the wind speed change rate, an adaptive threshold based on the wind speed change rate is innovatively designed, effectively realizing the diagnosis of minor faults. In addition, the digital twin model of this method is simply constructed and can effectively realize the real-time diagnosis of inter-turn short circuit faults. Description of the Drawings

[0044] Figure 1 It is the structure diagram of the digital twin model of the permanent magnet synchronous generator;

[0045] Figure 2 It is the effect diagram of the digital twin model;

[0046] Figure 3 It is the block diagram of the diagnosis strategy for inter-turn short circuit faults of the permanent magnet synchronous generator;

[0047] Figure 4 It is the curve of the current space vector and the fault flag diagram after the inter-turn short circuit fault occurs in the motor;

[0048] Figure 5 It is the curve diagram of the fault phase location flag and severity estimation after the inter-turn short circuit fault occurs in the motor. Specific Embodiments

[0049] The present invention will be described in detail below in combination with specific embodiments and the drawings.

[0050] The method for diagnosing the inter-turn short circuit fault of the permanent magnet synchronous motor in the wind power system considering wind speed fluctuation in this embodiment is as follows: First, obtain the stator current space vector as the diagnostic variable. Then, establish a 1D-CNN digital twin model considering wind speed fluctuation, calculate the reference value of the current space vector, and perform fault diagnosis by combining the residual between the measured value and the reference value with an adaptive threshold. Finally, determine the faulty phase through the effective value of the phase current, and estimate the severity by comparing the measured value and the reference value of the current space vector.

[0051] The present invention proposes a method for diagnosing the inter-turn short circuit fault of the permanent magnet synchronous motor in the wind power system considering wind speed fluctuation, and the method includes the following steps:

[0052] S1. Select the effective value of the stator current space vector as the diagnostic variable, and construct a digital twin model of the permanent magnet synchronous motor by establishing a non-electrical quantity expression of the stator current space vector. The model outputs the calculated value of the effective value of the stator current space vector.

[0053] S2. Establish a residual expression between the measured value and the calculated value of the effective value of the stator current space vector of the permanent magnet synchronous motor, and design an adaptive threshold based on the wind speed change rate.

[0054] S3. Combine the digital twin model and the adaptive threshold to perform real-time monitoring on the permanent magnet synchronous motor.

[0055] Furthermore, the specific method of step S1 is as follows:

[0056] Step 1.1) Select the effective value of the stator current space vector as the diagnostic variable, and establish a non-electrical quantity expression of the stator current space vector. Establish the connection between the mechanical system and the electrical system according to the law of conservation of momentum T w =T e +T loss , where T w represents the mechanical torque of the wind turbine:

[0057]

[0058] In the formula, ρ is the air density, R is the radius of the wind turbine blade, v w represents the wind speed, C p represents the wind energy utilization coefficient, and ω is the rotational angular velocity of the wind turbine;

[0059] T e represents the electromagnetic torque. The electromagnetic torque T p of the permanent magnet synchronous motor with the number of pole pairs n e is:

[0060]

[0061] In the formula, ψ f is the permanent magnet flux linkage; Ld , L q respectively represent the direct-axis and quadrature-axis inductances of the motor; i d , i q respectively represent the direct-axis and quadrature-axis currents of the motor. According to the actual system, L d ≈L q , i d = ki q and k is a coefficient greater than 0 and much less than 1. Then T e is simplified to:

[0062]

[0063] where T loss represents the loss torque including iron loss, stray loss, and mechanical loss:

[0064] T loss = C1 + C2ω (4)

[0065] where C1 represents the sum of the constant parts of iron loss, mechanical loss, and stray loss, and C2 represents the coefficient of the variable part of iron loss;

[0066] Substitute equations (1), (3), and (4) into T w = T e + T loss and combine with the current space vector amplitude, i.e., equation (5), to obtain the non-electrical quantity expression of the current space vector amplitude I m as shown in equation (6):

[0067]

[0068] According to equation (6) and the effective value of the current space vector equal to I m , select the wind speed v w and the rotational speed n r as the main input parameters of the digital twin model, and select the mechanical torque T w and the DC bus voltage U dc as the auxiliary input parameters;

[0069] Step 1.2) Design the digital twin model of the permanent magnet synchronous generator. The structure of the twin model is as Figure 1 shown. Use a depth 1D-CNN to build the model. The model consists of a multi-channel one-dimensional convolutional layer and a fully connected layer. The input parameters X = {v w , n r , T w , U dc}, and the output is the calculated value of the effective value of the stator current space vector i.e., the output The input X first passes through a one-dimensional convolutional layer with 32 channels, followed by activation, batch normalization, and pooling operations. Then it passes through a one-dimensional convolutional layer with 64 channels for the same operations, and finally outputs Y through three fully connected layers;

[0070] Step 1.3) Train and save the model. Obtain wind speed data through the wind speed measurement system, use MATLAB / Simulink for simulation modeling to obtain the dataset {X, Y}. After cleaning the data, divide it into a training set and a validation set in an 8:2 ratio. Use the Adam optimizer with an initial learning rate of 0.0001, a batch size of 64, and iterate 50 times. If the mean squared error of the model meets the preset requirements, save it; otherwise, adjust the hyperparameters and retrain. The mean squared error loss curves of the saved model on the training set and the test set are as shown in Figure 2 (a), and the current space vector residual curve is as shown in Figure 2 (b). It can be seen that the model converges quickly and the residual fluctuation range is very small.

[0071] Furthermore, the specific method of step S2 is as follows:

[0072] Step 2.1) Represent the current space vector in the form of the cube of the wind speed according to formula (6):

[0073]

[0074] where a1 and a2 are constants used to characterize the relationship between the magnitude of the current space vector and the cube of the wind speed. The specific expressions are shown in formulas (8) and (9):

[0075]

[0076] Step 2.2) Regarding the smoothing effect of the PI controller and the moment of inertia on I m under wind speed fluctuations, expand formula (7) to:

[0077]

[0078] where a' represents the smoothing degree of the PI controller and the moment of inertia on I m under wind speed fluctuations, and its value is between 0.01a1 and 0.1a1;

[0079] Step 2.3) Combine the normal state of the motor and formula (10) to obtain the measurement value of the effective value of the stator current space vector of the motor in the normal state and The residual expression between them is as follows:

[0080]

[0081] In the formula, ε represents under constant wind speed The error with ;

[0082] Step 2.4) Design an adaptive threshold based on the wind speed change rate. When the wind speed change rate is less than v th , the diagnostic variable threshold is selected as ε th ; when the wind speed change rate is greater than or equal to v th , the diagnostic variable threshold is selected as where ε th represents the current residual threshold under steady-state conditions, and v th represents the wind speed change rate threshold.

[0083] Furthermore, the overall strategy of step S3 is as Figure 3 shown, and the specific method is as follows:

[0084] Step 3.1) Fault monitoring and judgment. Measure the effective value of the motor current space vector in real time, and use the digital twin model to calculate the effective value of the space vector under the same operating conditions to obtain the residual between the two. Design a time window Δt. If the time span during which the residual exceeds the adaptive threshold within Δt is greater than Δt / 2, it is determined that there is a turn-to-turn short circuit fault; otherwise, the motor operates normally. If the motor is normal, collect operation data and update the model regularly; if a fault occurs, issue an alarm and enter steps 3.2 and 3.3. At 50 seconds, a 20% turn-to-turn short circuit fault occurs in phase A. The comparison curve of the measured value and the calculated value of the effective value of the current space vector is as Figure 4 (a) shown, and the fault flag is as Figure 4 (b) shown. It can be seen that there is an obvious difference between the calculated value and the measured value after the fault occurs, and the diagnostic algorithm can quickly output the diagnostic result.

[0085] Step 3.2) Fault phase location. Locate by measuring the effective values of the three-phase currents, and use the maximum value function to determine the phase where the maximum value is located. Design a time window Δt. The phase where the maximum value of the current, which has the largest proportion within Δt, is output is the fault phase. The fault phase location flag is as Figure 5 (a) shown. After a 20% turn-to-turn short circuit fault occurs in phase A at 50 seconds, the diagnostic algorithm accurately locates the fault phase.

[0086] Step 3.3) Estimate the severity through the effective value of the space current. The fault severity estimation curve is as Figure 5 (b) shown. After a 20% turn-to-turn short circuit fault occurs in phase A at 50 seconds, the diagnostic algorithm can estimate the severity more accurately. The specific estimation formula is as follows:

[0087]

Claims

1. A method for diagnosing inter-turn short-circuit faults of permanent magnet synchronous motors in wind power systems considering wind speed fluctuations, characterized in that: The method comprises the following steps: S1, select the effective value of the stator current space vector as the diagnostic variable, build a digital twin model of the permanent magnet synchronous motor by establishing a non-electrical quantity expression of the stator current space vector, and the model outputs the calculated value of the effective value of the stator current space vector; S2, establish the residual expression between the measured value and the calculated value of the effective value of the permanent magnet synchronous motor stator current space vector, and design an adaptive threshold based on the wind speed change rate; S3, real-time monitoring of permanent magnet synchronous motors using a digital twin model and adaptive thresholds.

2. A method for diagnosing inter-turn short-circuit faults of permanent magnet synchronous motors in wind power systems considering wind speed fluctuations according to claim 1, characterized in that: The specific method of step S1 is as follows: Step 1.1) Select the effective value of the stator current space vector as the diagnostic variable, and establish the non-electrical quantity expression of the stator current space vector. According to the law of conservation of momentum, the relationship between the mechanical system and the electrical system is established. w =T e +T loss , where T w Represents the mechanical torque of the wind turbine: Where ρ is the air density, R is the fan blade radius, and v w Indicates wind speed, C p represents the wind energy utilization coefficient, ω is the wind turbine rotation angular velocity; T e Represents electromagnetic torque, with the number of pole pairs being n p The electromagnetic torque T of the permanent magnet synchronous motor e for: In the formula, ψ f is the permanent magnet flux; L d , L q Respectively represent the direct and quadrature axis inductance of the motor; i d 、i q Respectively represent the motor direct and quadrature axis currents. According to the actual system, L d ≈L q ,i d =ki q And k is a coefficient greater than 0 and much less than 1, T e Simplified to: Among them, T loss Expresses the loss torque including iron loss, stray loss and mechanical loss: T loss =C1+C2ω (4) Where C1 represents the sum of the constant parts of iron loss, mechanical loss and stray loss, and C2 represents the coefficient of the variable part of iron loss; Substituting equations (1), (3) and (4) into T w =T e +T loss Combined with the current space vector amplitude, that is, equation (5), the current space vector amplitude I m The non-electrical quantity expression of is shown in formula (6): According to formula (6) and the effective value of the current space vector in the normal state of the motor Equal to I m , Select wind speed v w and speed n r As the main input parameter of the digital twin model, the mechanical torque T is selected w and DC bus voltage U dc As auxiliary input parameters; Step 1.2) Design the digital twin model of the permanent magnet synchronous generator and use deep 1D-CNN to build the model. The model consists of a multi-channel one-dimensional convolutional layer and a fully connected layer. The input parameter X = {v w , n r , T w , U dc }, the output is the calculated value of the effective value of the stator current space vector Output The input X first passes through a 32-channel one-dimensional convolutional layer for activation, batch normalization, and pooling operations, then passes through a 64-channel one-dimensional convolutional layer for the same operations, and outputs Y through three fully connected layers; Step 1.3) Train and save the model, obtain wind speed data through the wind speed measurement system, use MATLAB / Simulink simulation modeling, obtain the data set {X, Y}, clean the data and divide it into training set and validation set according to 8:2, use Adam optimizer, initial learning rate 0.0001, batch size 64, iteration 50 times, if the model mean square error meets the preset requirements, save it, otherwise, adjust the hyperparameters and retrain.

3. A method for diagnosing inter-turn short-circuit faults of permanent magnet synchronous motors in wind power systems considering wind speed fluctuations according to claim 2, characterized in that: The specific method of step S2 is as follows: Step 2.1) Express the current space vector as the cubic form of wind speed according to formula (6): Where a1 and a2 are constants used to characterize the relationship between the current space vector amplitude and the cubic wind speed. The specific expressions are shown in equations (8) and (9): Step 2.2) PI controller and moment of inertia for wind speed fluctuations m The smoothing effect of , formula (7) is expanded to: Where a' represents the effect of PI controller and moment of inertia on I under wind speed fluctuation. m The degree of smoothness is between 0.01a1 and 0.1a1; Step 2.3) Combined with the normal state of the motor And formula (10), the measured value of the effective value of the motor stator current space vector in the normal state is obtained and The residual expression between is as follows: In the formula, ε represents the wind speed under constant and The error of Step 2.4) Design an adaptive threshold based on the wind speed change rate. When the wind speed change rate is less than v th When the diagnostic variable threshold is ε th ; When the wind speed change rate is greater than or equal to v th When the diagnostic variable threshold is selected as Among them, ε th Indicates the current residual threshold in steady state, v th Indicates the wind speed change rate threshold.

4. A method for diagnosing turn-to-turn short-circuit faults of permanent magnet synchronous motors in wind power systems considering wind speed fluctuations according to claim 3, characterized in that: The specific method of step S3 is as follows: Step 3.1) Fault monitoring and judgment, real-time measurement of the effective value of the motor current space vector, and use the digital twin model to calculate the effective value of the space vector under the same operating conditions to obtain the residual; design a time window Δt, if the residual exceeds the adaptive threshold within Δt for a time span greater than Δt / 2, it is determined to be a turn-to-turn short circuit fault; otherwise, the motor runs normally; if the motor is normal, collect operating data and update the model regularly; if a fault occurs, issue an alarm and enter steps 3.2 and 3.3; Step 3.2) Fault phase location: locate the fault phase by measuring the effective value of the three-phase current and using the maximum value function to determine the phase where the maximum value is located; design a time window Δt, and output the phase where the current maximum value with the largest proportion within Δt is located, which is the fault phase; Step 3.3) Estimate the severity by using the effective value of the space current. The calculation formula is as follows:

Images