WO2023221740A1 - 永磁同步电机效率优化控制方法及系统 - Google Patents
永磁同步电机效率优化控制方法及系统 Download PDFInfo
- Publication number
- WO2023221740A1 WO2023221740A1 PCT/CN2023/090074 CN2023090074W WO2023221740A1 WO 2023221740 A1 WO2023221740 A1 WO 2023221740A1 CN 2023090074 W CN2023090074 W CN 2023090074W WO 2023221740 A1 WO2023221740 A1 WO 2023221740A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- permanent magnet
- magnet synchronous
- synchronous motor
- optimal
- value
- Prior art date
Links
- 230000001360 synchronised effect Effects 0.000 title claims abstract description 118
- 238000000034 method Methods 0.000 title claims abstract description 87
- 238000005457 optimization Methods 0.000 title claims abstract description 41
- 230000002787 reinforcement Effects 0.000 claims abstract description 56
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 24
- 230000009471 action Effects 0.000 claims description 52
- 230000001133 acceleration Effects 0.000 claims description 26
- 230000002159 abnormal effect Effects 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 21
- 230000008859 change Effects 0.000 claims description 8
- 230000004907 flux Effects 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 6
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000004088 simulation Methods 0.000 claims description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 32
- 229910052742 iron Inorganic materials 0.000 description 15
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 8
- 229910052802 copper Inorganic materials 0.000 description 8
- 239000010949 copper Substances 0.000 description 8
- 230000010349 pulsation Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 3
- 229910052799 carbon Inorganic materials 0.000 description 3
- 239000003795 chemical substances by application Substances 0.000 description 3
- 239000002131 composite material Substances 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- WCUXLLCKKVVCTQ-UHFFFAOYSA-M Potassium chloride Chemical compound [Cl-].[K+] WCUXLLCKKVVCTQ-UHFFFAOYSA-M 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000011162 core material Substances 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 230000001052 transient effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
- H02P21/0003—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control
- H02P21/0014—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control using neural networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
- H02P21/0003—Control strategies in general, e.g. linear type, e.g. P, PI, PID, using robust control
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
- H02P21/50—Vector control arrangements or methods not otherwise provided for in H02P21/00- H02P21/36
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
- H02P21/14—Estimation or adaptation of machine parameters, e.g. flux, current or voltage
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P25/00—Arrangements or methods for the control of AC motors characterised by the kind of AC motor or by structural details
- H02P25/02—Arrangements or methods for the control of AC motors characterised by the kind of AC motor or by structural details characterised by the kind of motor
- H02P25/022—Synchronous motors
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P6/00—Arrangements for controlling synchronous motors or other dynamo-electric motors using electronic commutation dependent on the rotor position; Electronic commutators therefor
- H02P6/34—Modelling or simulation for control purposes
Definitions
- the invention belongs to the field of permanent magnet synchronous motor control, and relates to a composite permanent magnet synchronous motor efficiency optimization control method and system based on a combination of loss model method and deep reinforcement learning.
- Permanent magnet synchronous motors have the advantages of small size and high power density, and are widely used in many industries. How to further improve their operating efficiency is a persistent goal of scientific researchers. Permanent magnet synchronous motor efficiency optimization control technology can be summarized into two categories: efficiency optimization control method based on loss model and efficiency optimization control method based on search technology.
- the loss model method is currently the mainstream method. It establishes the maximum efficiency function of the motor or the minimum loss function of the motor, derives the excitation current and excitation flux linkage at optimal efficiency, and controls them to achieve efficiency optimization. It can achieve control targets quickly and achieve real-time control, but it ignores the impact of nonlinear factors such as saturation effects on parameter changes, and often does not consider the impact of the driver's switching action on motor efficiency.
- the salient pole permanent magnet synchronous motor's orthogonal and direct axis parameters are not equal. Obtaining the optimal direct axis current initial value is complex and highly dependent on the motor parameters. These parameters change with the operating conditions of the motor, requiring online identification of the motor parameters. Therefore, it is difficult to implement. In practice, the model often needs to be simplified, and the approximate optimal solution obtained still deviates from the actual optimal value. In the end, only suboptimal control of efficiency can be achieved.
- the search method can achieve the global optimum of system efficiency and is not affected by changes in motor parameters.
- the algorithm takes a long time to converge and the convergence speed is slow, which easily causes pulsation of magnetic flux and torque.
- convergence speed and torque pulsation There is a contradiction between convergence speed and torque pulsation.
- improper selection of the step size of the control algorithm will cause the algorithm to converge too slowly or even not converge, resulting in a decrease in overall efficiency and seriously affecting the optimal efficiency control effect.
- the present invention provides a permanent magnet synchronous motor efficiency optimization control method and system, which can further improve the operating efficiency of the permanent magnet synchronous motor.
- a permanent magnet synchronous motor efficiency optimization control method including the following steps:
- Step 1 Use the loss model method to obtain the approximately optimal d-axis current of the permanent magnet synchronous motor
- Step 2 Taking the approximately optimal d-axis current as the initial value, use the deep reinforcement learning (deep Q network, DQN) algorithm to optimize the d-axis current and build the optimal deep reinforcement learning model;
- DQN deep Q network
- Step 3 Send the currently collected permanent magnet synchronous motor status data to the optimal deep reinforcement learning model. Based on this model, predict the control parameter value (optimal d-axis current) that enables the motor to operate at optimal efficiency; according to the control parameters The value controls the permanent magnet synchronous motor.
- step 1 includes the following steps:
- the losses of permanent magnet synchronous motor include mechanical loss, copper loss and iron loss. Among them, mechanical losses continue to change with different rotation speeds and working conditions, and are uncontrollable. Therefore, this invention only considers the controllable part of electrical losses, including iron losses and copper losses. Combined with the d, q-axis equivalent circuit of the permanent magnet synchronous motor considering copper loss and iron loss, the dynamic voltage equation of the permanent magnet synchronous motor can be written as:
- u d and u q are the d and q-axis armature voltages respectively; i d and i q are the d and q-axis currents (armature current) respectively; i wd and i wq are the active components of the d and q-axis stator currents; R s is the armature resistance; L d and L q are the armature inductances of the d and q axes respectively; ⁇ m is the flux linkage of the rotor permanent magnet; ⁇ r is the mechanical angular speed of the rotor; n p is the number of pole pairs of the motor;
- i Feq is the d, q-axis iron loss current component, R Fe is the equivalent iron loss resistance;
- the total loss of the permanent magnet synchronous motor can be obtained as:
- Te is the electromagnetic torque of the permanent magnet synchronous motor.
- the parameters will continue to change.
- the obtained d-axis current can only be approximately the optimal d-axis current due to the influence of the parameters.
- this paper The invention continues to optimize parameters through deep reinforcement learning to obtain optimal parameters for motor control.
- step 2 includes the following steps:
- Step 2.1 Use the deep reinforcement learning algorithm as the intelligent agent of the permanent magnet synchronous motor to construct a deep reinforcement learning model, including the value network Q v and the target network Q T with the same structure, the state space S, the action space A and the reward function Reward.
- Each state in the state space S is represented as a four-dimensional vector (T, v, acc, T1), where T, v, acc, and T1 respectively represent the electromagnetic torque, rotation speed, acceleration and load torque of the permanent magnet synchronous motor.
- each action in the action space A is a value of the d-axis current i d ;
- Step 2.2 Use the current deep reinforcement learning model to optimize the suboptimal data, that is, the approximately optimal d-axis current, to obtain the global optimal data pool;
- Step 2.3 Detect and eliminate abnormal data in the global optimal data pool
- Step 2.4 Use the sample data in the global optimal data pool after eliminating abnormal data to train the deep reinforcement learning model, and determine whether the termination condition is met. If so, terminate and obtain the optimal deep reinforcement learning model; otherwise, return Step 2.2.
- step 2.2 specifically includes the following steps:
- Step 2.2.1 Construct data pool D with capacity C;
- Step 2.2.2 Measure the current state s of the motor, including the electromagnetic torque, speed, acceleration and load torque of the permanent magnet synchronous motor; record the approximately optimal d-axis current as action a, and execute action a, that is, according to the approximate optimal
- the optimal d-axis current controls the permanent magnet synchronous motor.
- the permanent magnet synchronous motor automatically updates (transfers) to the next state s' (this state is the sub-optimal state of the permanent magnet synchronous motor).
- the reward function is used to calculate the actions performed in the current state s.
- the reward value r of a store the current sample data (s, a, r, s') in the data pool D;
- Step 2.2.3 Update the current state s to s', input the current state s into the value network Q v , and use the parameters ⁇ of the value network Q v to obtain the predicted Q value of each action in the state space under the current state s, using The ⁇ -greedy strategy selects the most suitable action, denoted as a;
- Step 2.2.4 Execute action a, that is, control the permanent magnet synchronous motor according to the approximately optimal d-axis current, and the permanent magnet synchronous motor automatically updates to the next state s'; calculate the cost of executing action a in the current state s through the reward function Reward value r; store the current sample data (s, a, r, s') into the data pool D;
- Step 2.2.5 Return to 2.2.3 to loop until data pool D is full (the number of sample data in data pool D reaches its capacity C), and data pool D at this time is regarded as the global optimal data pool.
- the state of the permanent magnet synchronous motor (including electromagnetic torque, rotation speed, acceleration and load torque) obtained by the approximately optimal d-axis current obtained by the loss model method is a fixed value and is a sub-optimal state of the permanent magnet synchronous motor; this
- the state space S (state value range) is designed, in which each state is a value of electromagnetic torque, rotation speed, acceleration and load torque expressed as a permanent magnet synchronous motor, and is judged through state information and rewards.
- the deep reinforcement learning model is used to optimize the data, and the motor control can be adjusted in a timely manner to obtain the global optimal data pool.
- T(k), v(k), acc(k), and T1(k) respectively represent the electromagnetic torque, speed, acceleration and load torque of the permanent magnet synchronous motor before executing action a;
- T(k+1) , v(k+1), acc(k+1), T1(k+1) respectively represent the electromagnetic torque, speed, acceleration and load torque of the permanent magnet synchronous motor after executing action a;
- Reward T(k+1 ), Reward v(k+1), Reward acc(k+1), Reward T1(k+1) respectively represent the electromagnetic torque, rotation speed, acceleration and load torque reward of action a; ⁇ 1 , ⁇ 2 , ⁇ 3.
- step 2.3 the density-based local abnormal factor method is used to detect and eliminate abnormal data in the global optimal data pool
- abnormal points In actual engineering, the motor system sometimes produces data that is obviously inconsistent with common sense, called abnormal points.
- search methods based on deep reinforcement learning whether anomalies are transient or permanent, they will have a catastrophic impact on the results of data-driven methods. This is because abnormal data will bring "confusion" to the data-driven method, making it impossible to mine the real information inside the data;
- the local anomaly factor method uses the density-based local anomaly factor method to detect and eliminate outliers.
- the local anomaly factor method based on density determines whether the point is an anomaly point by comparing the density of each point with its neighbor points. The lower the density of point p, the more likely it is to be identified as an anomaly point. As for density, it is calculated by the distance between points. The farther the distance between points, the lower the density. The closer the distance, the higher the density. Because this method calculates the density through the k-th neighborhood of the point instead of a global calculation, it is called the "local" anomaly factor, and will not mistakenly determine normal points as abnormal points due to different data density dispersions. .
- the specific steps are:
- d k (o) is the k-th distance from neighborhood point o
- d (o, p) is the distance from neighborhood point o to point p;
- Step 2.3.2 Calculate the kth local reach density p reach (p) of each data point p:
- N k (p) is the k-th distance neighborhood of point p;
- Step 2.3.3 Calculate the k-th local outlier factor for each data point:
- Step 2.3.4 Analyze the local outlier factor. If the local outlier factor is less than 1, it is not an outlier. If the local outlier factor is greater than 1, it is an outlier;
- Step 2.3.5 Eliminate abnormal points.
- the step 2.4 specifically includes the following steps:
- Step 2.4.2 Use randomly selected n pieces of sample data as inputs to the value network Q v and the target network Q T ;
- Step 2.4.3 Calculate the loss function L through Q predict(i) and Q Target(i) :
- Step 2.4.4 Minimize the loss function L by the error inversion method, where the parameters of the value network Q v are used using the gradient descent method ⁇ is updated;
- ⁇ is the learning law and is the empirical parameter
- Step 2.4.6 Return to step 2.2.3.
- ⁇ is set to 0.01, so that the error value has a small impact on the new Q value, while having a certain impact.
- the target network parameter update period K is set to 3 simulation cycles.
- the target network parameters are updated every 3 cycles to obtain a new target network in time, while avoiding the problem caused by too long time.
- the target network is never updated, causing the Q value to fall into a local optimum.
- the control parameter values include the optimal d-axis current and the optimal q-axis current; first, based on the optimal deep reinforcement learning model, the optimal d-axis current that enables the motor to operate at optimal efficiency is predicted; according to the optimal The optimal d-axis current controls the permanent magnet synchronous motor, and at the same time, the q-axis current of the permanent magnet synchronous motor is compensated according to the optimal d-axis current to achieve optimal control of the efficiency of the permanent magnet synchronous motor under steady state.
- this solution uses adjusting the d-axis current while compensating the q-axis current. Thereby maintaining the constant torque;
- the electromagnetic torque before d-axis current adjustment is:
- the q-axis current is adjusted accordingly to i q * , and the adjusted electromagnetic torque can be obtained as:
- a permanent magnet synchronous motor efficiency optimization control system which includes a memory and a processor.
- a computer program is stored in the memory. The feature is that when the computer program is executed by the processor, the computer program causes the The processor implements the above-mentioned permanent magnet synchronous motor efficiency optimization control method.
- the system also includes a data acquisition module for collecting the current status data of the permanent magnet synchronous motor, including the current electromagnetic torque, rotation speed, acceleration and load torque of the permanent magnet synchronous motor, and sending it to the processor. middle.
- This invention combines the deep reinforcement learning algorithm and the loss model method for the first time, and proposes a new composite method to solve the problem of motor efficiency optimization.
- This method can inherit the advantages of both and overcome the shortcomings of both.
- the driver and permanent magnet synchronous motor are regarded as a whole and a unified efficiency optimization model is established.
- the efficiency will not reach the global optimum; the deep reinforcement learning method is to further optimize based on the model method.
- the algorithm takes a long time to converge and the convergence speed is slow, which easily causes pulsation of magnetic flux and torque. There is a contradiction between convergence speed and torque pulsation.
- This Improper selection of the step size of the time control algorithm will cause the algorithm to converge too slowly or even not converge, resulting in a decrease in overall efficiency and seriously affecting the efficiency optimal control effect.
- the present invention obtains the suboptimal efficiency d-axis current through the loss model method, and uses this current as the initial value of the deep reinforcement learning search, which can reduce the scope of data search, shorten the algorithm convergence time, and at the same time avoid the impact of changes in motor parameters on the results.
- the density-based local anomaly factor method is adopted to avoid the "confusion" caused by abnormal data to the deep reinforcement learning method, making it impossible to mine the real information inside the data, thereby reducing the algorithm search accuracy.
- the use of q-axis current compensation can reduce the torque ripple and optimize the q-axis current in time.
- the electromagnetic torque, rotation speed, acceleration and load torque obtained through the loss model method are fixed values
- the present invention combines deep reinforcement learning to set the state space as a four-dimensional vector B (T, v, acc, T1), It includes the current electromagnetic torque, speed, acceleration and load torque of the motor, and calculates the reward value based on the current status of the motor, so as to select the most suitable action and achieve optimal control of the motor.
- the present invention proposes a permanent magnet synchronous motor efficiency optimization method, which combines the strengths of the current efficiency optimization control method based on a loss model and the efficiency optimization control method based on search technology, and overcomes the shortcomings of the two.
- the motor loss model Through the motor loss model, the total loss of the motor can be intuitively obtained, speeding up the motor response speed, and being applicable to various occasions.
- the introduction of deep reinforcement learning can enable the motor to obtain the global optimal solution when the parameters change, and realize real-time control of the motor operation.
- the motor loss model method can be used to establish a sub-optimal data pool, and then search for global optimization based on the sub-optimal method, which greatly reduces the scope of data search, shortens the algorithm convergence time, and avoids the need for motor optimization at the same time.
- the impact of parameter changes on the results can Real-time optimal control of the motor.
- Figure 1 is a vector control block diagram for efficiency optimization of permanent magnet synchronous motor.
- Figure 2 shows the d-axis equivalent circuit considering iron loss.
- Figure 3 shows the q-axis equivalent circuit considering iron loss.
- Figure 4 shows the deep reinforcement learning framework for motor efficiency optimization.
- the motor model automatically updates to the next state based on the action and current state.
- the state is passed to the reward function determined based on the target model to obtain the corresponding reward value.
- the state vector and reward value are placed into the data pool
- Figure 5 shows the DQN algorithm framework.
- the present invention proposes a permanent magnet synchronous motor efficiency optimization control method, which includes the following steps:
- This embodiment provides a permanent magnet synchronous motor efficiency optimization control method, which includes the following steps:
- Step 1 Use the loss model method to obtain the approximately optimal d-axis current of the permanent magnet synchronous motor
- Step 2 Taking the approximately optimal d-axis current as the initial value, use the deep reinforcement learning (deep Q network, DQN) algorithm to optimize the d-axis current and build the optimal deep reinforcement learning model;
- DQN deep Q network
- Step 3 Send the currently collected permanent magnet synchronous motor status data to the optimal deep reinforcement learning model. Based on this model, predict the control parameter value (optimal d-axis current) that enables the motor to operate at optimal efficiency; according to the control parameters The value controls the permanent magnet synchronous motor.
- the step 1 includes the following steps:
- the losses of permanent magnet synchronous motor include copper loss, iron loss, stray loss and mechanical loss.
- Copper loss refers to the heat loss generated when current passes through the stator winding. Since the motor stator is equipped with a large number of coils, and the current passing through the coils is large during operation, copper loss is one of the main losses of the motor; iron loss refers to the magnetic flux in the The hysteresis and eddy current losses caused by the alternating flow in the motor core depend on the core material itself and are related to the current frequency and speed. Mechanical losses and stray losses account for a relatively small proportion of the total loss of the motor and are difficult to control. detection. Therefore, this solution only considers iron loss and copper loss. Combining the d and q-axis equivalent circuits of the permanent magnet synchronous motor in Figures 2 and 3, considering copper loss and iron loss, the dynamic voltage equation of the permanent magnet synchronous motor can be written as:
- u d and u q are the d and q-axis armature voltages respectively; i d and i q are the d and q-axis currents (armature current) respectively; i wd and i wq are the active components of the d and q-axis stator currents; R s is the armature resistance; L d and L q are the armature inductances of the d and q axes respectively; ⁇ m is the flux linkage of the rotor permanent magnet; ⁇ r is the mechanical angular speed of the rotor; n p is the number of pole pairs of the motor;
- i Feq is the d, q-axis iron loss current component, R Fe is the equivalent iron loss resistance;
- the total loss of the permanent magnet synchronous motor can be obtained as:
- Te is the electromagnetic torque of the permanent magnet synchronous motor.
- the parameters will continue to change.
- the obtained d-axis current can only be approximately the optimal d-axis current due to the influence of the parameters.
- this paper The invention continues to optimize parameters through deep reinforcement learning to obtain optimal parameters for motor control.
- step 2 includes the following steps:
- Step 2.1 Use the deep reinforcement learning algorithm as the intelligent agent of the permanent magnet synchronous motor to construct a deep reinforcement learning model, including the value network Q v and the target network Q T with the same structure, the state space S, the action space A and the reward function Reward.
- Each state in the state space S is represented as a four-dimensional vector (T, v, acc, T1), where T, v, acc, and T1 respectively represent the electromagnetic torque, rotation speed, acceleration and load torque of the permanent magnet synchronous motor.
- each action in the action space A is a value of the d-axis current i d ;
- Step 2.2 Use the current deep reinforcement learning model to optimize the suboptimal data, that is, the approximately optimal d-axis current, to obtain the global optimal data pool;
- Step 2.3 Detect and eliminate abnormal data in the global optimal data pool
- Step 2.4 Use the sample data in the global optimal data pool after eliminating abnormal data to train the deep reinforcement learning model, and determine whether the termination condition is met. If so, terminate and obtain the optimal deep reinforcement learning model; otherwise, return Step 2.2.
- step 2.2 specifically includes the following steps:
- Step 2.2.1 Construct data pool D with capacity C;
- Step 2.2.2 Measure the current state s of the motor, including the electromagnetic torque, speed, acceleration and load torque of the permanent magnet synchronous motor; record the approximately optimal d-axis current as action a, and execute action a, that is, according to the approximate optimal
- the optimal d-axis current controls the permanent magnet synchronous motor.
- the permanent magnet synchronous motor automatically updates (transfers) to the next state s' (this state is the sub-optimal state of the permanent magnet synchronous motor).
- the reward function is used to calculate the actions performed in the current state s.
- the reward value r of a store the current sample data (s, a, r, s') in the data pool D;
- Step 2.2.3 Update the current state s to s', input the current state s into the value network Q v , and use the parameters ⁇ of the value network Q v to obtain the predicted Q value of each action in the state space under the current state s, using The ⁇ -greedy strategy selects the most suitable action, denoted as a;
- Step 2.2.4 Execute action a, that is, control the permanent magnet synchronous motor according to the approximately optimal d-axis current, and the permanent magnet synchronous motor automatically updates to the next state s'; calculate the cost of executing action a in the current state s through the reward function Reward value r; store the current sample data (s, a, r, s') into the data pool D;
- Step 2.2.5 Return to 2.2.3 to loop until data pool D is full (the number of sample data in data pool D reaches its capacity C), and data pool D at this time is regarded as the global optimal data pool.
- the state of the permanent magnet synchronous motor (including electromagnetic torque, rotation speed, acceleration and load torque) obtained by the approximately optimal d-axis current obtained by the loss model method is a fixed value and is a sub-optimal state of the permanent magnet synchronous motor; this
- the state space S state value range
- each state is a value of electromagnetic torque, rotation speed, acceleration and load torque represented as a permanent magnet synchronous motor
- the deep reinforcement learning model is used to By optimizing the data, the global optimal data pool can be obtained.
- Using the motor state obtained by the loss model method as the initial value of the deep reinforcement learning algorithm can reduce the number of trials and errors during the motor search process, search for global optimization on the basis of suboptimality, and greatly reduce training and learning time.
- the value of C in step 2.2.1 is set to 200, that is, 200 sample data can be stored in data pool D.
- T(k), v(k), acc(k), and T1(k) respectively represent the electromagnetic torque of the permanent magnet synchronous motor before executing action a, Speed, acceleration and load torque
- T(k+1), v(k+1), acc(k+1), T1(k+1) respectively represent the electromagnetic torque of the permanent magnet synchronous motor after executing action a, Rotation speed, acceleration and load torque
- Reward T(k+1), Reward v(k+1), Reward acc(k+1), Reward T1(k+1) respectively represent the electromagnetic torque, rotation speed, Acceleration and load torque rewards
- ⁇ 1 , ⁇ 2 , ⁇ 3 , ⁇ 4 are the weight parameters of the four reward factors, which can be determined based on experience
- step 2.3 the density-based local abnormal factor method is used to detect and eliminate abnormal data in the global optimal data pool;
- abnormal points In actual engineering, the motor system sometimes produces data that is obviously inconsistent with common sense, called abnormal points.
- anomalies For search methods based on deep reinforcement learning, whether anomalies are transient or permanent, they will have a catastrophic impact on the results of data-driven methods. This is because abnormal data will bring "confusion" to the data-driven method, making it impossible to mine the real information inside the data. By detecting and eliminating abnormal points, the real information inside the data can be mined to make the motor run more smoothly.
- the local anomaly factor method uses the density-based local anomaly factor method to detect and eliminate outliers.
- the local anomaly factor method based on density determines whether the point is an anomaly point by comparing the density of each point with its neighbor points. The lower the density of point p, the more likely it is to be identified as an anomaly point. As for density, it is calculated by the distance between points. The farther the distance between points, the lower the density. The closer the distance, the higher the density. Because this method calculates the density through the k-th neighborhood of the point instead of a global calculation, it is called the "local" anomaly factor, and will not mistakenly determine normal points as abnormal points due to different data density dispersions. .
- the specific steps are:
- d k (o) is the k-th distance from neighborhood point o
- d (o, p) is the distance from neighborhood point o to point p;
- Step 2.3.2 Calculate the kth local reach density p reach (p) of each data point p:
- N k (p) is the k-th distance neighborhood of point p;
- Step 2.3.3 Calculate the k-th local outlier factor for each data point:
- Step 2.3.4 Analyze the local outlier factor. If the local outlier factor is less than 1, it is not an outlier. If the local outlier factor is greater than 1, it is an outlier;
- Step 2.3.5 Eliminate abnormal points.
- step 2.4 specifically includes the following steps:
- Step 2.4.2 Use randomly selected n pieces of sample data as inputs to the value network Q v and the target network Q T ;
- Step 2.4.3 Calculate the loss function L through Q predict(i) and Q Target(i) :
- Step 2.4.4 Minimize the loss function L through the error inverse method, in which the gradient descent method is used to update the parameters ⁇ of the value network Q v ;
- ⁇ is the learning law and is the empirical parameter
- Step 2.4.6 Return to step 2.2.3.
- step 2.4.4 ⁇ is set to 0.01, so that the error value has a small impact on the new Q value and has a certain impact.
- step 2.4.5 the target network parameter update period K is set to 3 simulation cycles.
- the target network parameters are updated every 3 cycles to obtain new targets in time. network, while avoiding the target network not being updated for too long, causing the Q value to fall into a local optimum.
- the control parameter values include the optimal d-axis current and the optimal q-axis current; first, the currently collected real-time data is sent to the optimal deep reinforcement learning model, based on The optimal deep reinforcement learning model predicts the optimal d-axis current that enables the motor to operate at optimal efficiency; the permanent magnet synchronous motor is controlled based on the optimal d-axis current, and the q-axis of the permanent magnet synchronous motor is controlled based on the optimal d-axis current. The current is compensated to achieve optimal control of the efficiency of the permanent magnet synchronous motor under steady state.
- This solution is considered to be implemented in an actual motor, so it is necessary to ensure the steady-state operation of the motor to achieve optimal efficiency.
- changes in the d-axis current will cause fluctuations in rotation speed and electromagnetic torque.
- this solution uses adjusting the d-axis current while compensating the q-axis current. Thereby maintaining the constant torque;
- the electromagnetic torque before d-axis current adjustment is:
- the q-axis current is adjusted accordingly to i q * , and the adjusted electromagnetic torque can be obtained as:
- This embodiment provides a permanent magnet synchronous motor efficiency optimization control system, which includes a memory and a processor.
- a computer program is stored in the memory. The feature is that when the computer program is executed by the processor, the processing The device implements the efficiency optimization control method of the permanent magnet synchronous motor described in the above embodiment.
- the system further includes a data collection module for collecting the current status data of the permanent magnet synchronous motor, including the current electromagnetic torque, rotation speed, acceleration and load torque of the permanent magnet synchronous motor. , and sent to the processor.
- This invention combines the deep reinforcement learning algorithm and the loss model method for the first time, and proposes a new composite method to solve the problem of motor efficiency optimization.
- This method can inherit the advantages of both and overcome the shortcomings of both.
- the driver and permanent magnet synchronous motor are regarded as a whole and a unified efficiency optimization model is established.
- the efficiency will not reach the global optimum; the deep reinforcement learning method is to further optimize based on the model method.
- the algorithm takes a long time to converge and the convergence speed is slow, which easily causes pulsation of magnetic flux and torque. There is a contradiction between convergence speed and torque pulsation.
- This Improper selection of the step size of the time control algorithm will cause the algorithm to converge too slowly or even not converge, resulting in a decrease in overall efficiency and seriously affecting the efficiency optimal control effect.
- the present invention obtains the suboptimal efficiency d-axis current through the loss model method, and uses this current as the initial value of the deep reinforcement learning search, which can reduce the scope of data search, shorten the algorithm convergence time, and at the same time avoid the impact of changes in motor parameters on the results.
- the density-based local anomaly factor method is adopted to avoid the "confusion" caused by abnormal data to the deep reinforcement learning method, making it impossible to mine the real information inside the data, thereby reducing the algorithm search accuracy.
- the use of q-axis current compensation can reduce the torque ripple and optimize the q-axis current in time.
- the electromagnetic torque, rotation speed, acceleration and load torque obtained through the loss model method are fixed values
- the present invention combines deep reinforcement learning to set the state space as a four-dimensional vector B (T, v, acc, T1), It includes the current electromagnetic torque, speed, acceleration and load torque of the motor, and calculates the reward value based on the current status of the motor, so as to select the most suitable action and achieve optimal control of the motor.
- the present invention provides a permanent magnet synchronous motor efficiency optimization control method and system, which can enable the motor to On the basis of achieving global optimal efficiency, it has the advantages of short search time, fast convergence speed, real-time control of motor operation and is suitable for various occasions; it can achieve energy saving and emission reduction, reduce losses and improve efficiency.
- the method and system proposed by the present invention are simple, easy to understand, easy to implement, and highly accurate, and are a feasible solution that can effectively improve the efficiency of the permanent magnet synchronous motor.
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Control Of Ac Motors In General (AREA)
Abstract
本发明提供了一种永磁同步电机效率优化控制方法及系统,所述方法包括以下步骤:步骤1、利用损耗模型法求得永磁同步电机的近似最优d轴电流;步骤2、以近似最优d轴电流为初始值,利用深度强化学习算法对d轴电流进行寻优,构建最优深度强化学习模型;步骤3、将当前采集的永磁同步电机状态数据送入最优深度强化学习模型,基于该模型预测出使电机运行在最优效率的控制参数值;根据该控制参数值对永磁同步电机进行控制。本发明能够对电机进行实时最优控制。
Description
本发明属于永磁同步电机控制领域,涉及一种基于损耗模型法和深度强化学习相结合的复合式永磁同步电机效率优化控制方法及系统。
随着社会经济的快速发展和电力需求的不断增加,能源短缺和环境污染问题变得越来越严重。2030年碳达峰,2060年碳中和,是我国制定的宏伟目标,也是对国际社会的庄重承诺。2020年全球电机系统用电量占比超过50%,我国电动机系统用电量占比达到了64%,提高电机的工作效率对缓解全球能源短缺和降低碳排放量有着至关重要的作用。
永磁同步电机具有体积小,功率密度高等优点,在多个行业广泛应用,如何进一步提高其运行效率,是科研工作者持之以恒的追求目标。永磁同步电机效率优化控制技术归纳为两类:基于损耗模型的效率优化控制方法和基于搜索技术的效率优化控制方法。
损耗模型法是目前的主流方法,其建立电机的效率最大函数或电机的损耗最小函数,推导出最优效率时的励磁电流和励磁磁链,并加以控制实现效率优化。它能够较快达到控制目标,实现实时控制,但它忽略饱和效应等非线性因素对参数变化的影响,往往也没有考虑驱动器的开关动作对于电机效率的影响。特别是凸极永磁同步电机交直轴参数不相等,得到最优直轴电流初始值比较复杂,并且对电机参数依赖大,而这些参数随电机运行状况而改变,需要对电机参数进行在线辨识。因此,实现起来难度大,实践中经常要对模型简化,得到的近似最优解与实际最优值仍存在偏差,最终只能实现效率次优控制。
搜索法能够实现系统效率的全局最优,且不受电机参数变化的影响,但算法收敛时间长,收敛速度慢,容易引起磁通和转矩的脉动,存在收敛速度和转矩脉动的矛盾,此时控制算法步长选择不当,则会使算法收敛过慢甚至不收敛,从而导致总体效率下降,严重影响效率最优控制效果。
针对现有技术存在的不足,有必要设计一种新的永磁同步电机效率优化控制方法,进一步提高永磁同步电机的运行效率。
发明内容
针对现有技术中存在的问题,本发明提供了一种永磁同步电机效率优化控制方法及系统,能进一步提高永磁同步电机的运行效率。
本发明提供的技术方案如下:
一方面,提供一种永磁同步电机效率优化控制方法,包括以下步骤:
步骤1、利用损耗模型法求得永磁同步电机的近似最优d轴电流;
步骤2、以近似最优d轴电流为初始值,利用深度强化学习(深度Q网络,DQN)算法对d轴电流进行寻优,构建最优深度强化学习模型;
步骤3、将当前采集的永磁同步电机状态数据送入最优深度强化学习模型,基于该模型预测出使电机运行在最优效率的控制参数值(最优d轴电流);根据该控制参数值对永磁同步电机进行控制。
进一步地,所述步骤1包括以下步骤:
首先,对电机损耗进行分析评估并构建永磁同步电机的损耗模型;
永磁同步电机的损耗包括机械损耗、铜损和铁损等。其中机械损耗随转速和工况的不同不断改变,是不可控的。因此本发明只考虑可控部分的电气损耗,包括铁损和铜损。结合永磁同步电机考虑铜损和铁损的d,q轴等效电路,可以写出永磁同步电机的动态电压方程为:
其中,ud与uq分别为d、q轴电枢电压;id与iq分别为d、q轴电流(电枢电流);iwd与iwq为d、q轴定子电流有功分量;Rs为电枢电阻;Ld与Lq分别为d、q轴的电枢电感;ψm为转子永磁体的磁链;ωr为转子的机械角速度;np为电机极对数;
由永磁同步电机等效电路回路电压方程(KVL),可得:
其中,iFeq为d,q轴铁损电流分量,RFe为等效铁损电阻;
结合d,q轴铁损电流分量表达式,可得永磁同步电机总损耗为:
Te为永磁同步电机电磁转矩。
然后,基于永磁同步电机总损耗模型求得近似最优d轴电流;
稳态运行时,电机转速,转矩保持不变,从总损耗关系式可知,电机损耗只与iwd有关,要想总损耗最小,只需满足:
可得iwd为:
由永磁同步电机等效电路节点电流方程(KCL),可得:
id=iwd+iFed
id=iwd+iFed
由此得到近似最优d轴电流为:
电机在运行过程中,参数会不断发生变化,当采用损耗模型法时,所得d轴电流因为参数的影响,只能为近似最优d轴电流,为实现永磁同步电机效率最优控制,本发明继续通过深度强化学习对参数进行寻优,以得到电机控制的最优参数。
进一步地,所述步骤2包括以下步骤:
步骤2.1:以深度强化学习算法为永磁同步电机的智能体Agent,构建深度强化学习模型,包括结构相同的值网络Qv和目标网络QT,状态空间S、动作空间A和奖励函数Reward,状态空间S中的每个状态为表示为一个四维向量(T,v,acc,T1),其中T、v、acc、T1分别表示永磁同步电机的电磁转矩、转速、加速度和负载转矩,动作空间A中的每个动作为d轴电流id的一种取值;
步骤2.2、利用当前的深度强化学习模型对次优数据即近似最优d轴电流进行寻优,得到全局最优数据池;
步骤2.3、对全局最优数据池中的异常数据进行检测和剔除;
步骤2.4、将剔除异常数据后的全局最优数据池中的样本数据,对深度强化学习模型进行训练,并判断是否满足终止条件,若满足,则终止,得到最优深度强化学习模型;否则返回
步骤2.2。
进一步地,所述步骤2.2具体包括以下步骤:
步骤2.2.1:构建数据池D,其容量为C;
步骤2.2.2:测得电机当前状态s,包括永磁同步电机的电磁转矩、转速、加速度和负载转矩;将近似最优d轴电流记为动作a,执行动作a,即根据近似最优d轴电流对永磁同步电机进行控制,永磁同步电机自动更新(转移)至下一状态s'(该状态为永磁同步电机次优状态),通过奖励函数计算当前状态s下执行动作a的奖励值r;将当前样本数据(s,a,r,s')存入数据池D中;
步骤2.2.3:将当前状态s更新为s',将当前状态s输入至值网络Qv中,利用值网络Qv的参数θ得到当前状态s下状态空间中各个动作的预测Q值,采用ε-greedy策略选取最适合的动作,记为a;
步骤2.2.4:执行动作a,即根据近似最优d轴电流对永磁同步电机进行控制,永磁同步电机自动更新至下一状态s';通过奖励函数计算当前状态s下执行动作a的奖励值r;将当前样本数据(s,a,r,s')存入数据池D中;
步骤2.2.5:返回2.2.3进行循环,直至数据池D满(数据池D中的样本数据条数达到其容量C),将此时的数据池D作为全局最优数据池。
通过损耗模型法求得的近似最优d轴电流进而获得的永磁同步电机状态(包括电磁转矩,转速,加速度和负载转矩)为定值,且为永磁同步电机次优状态;本方案中通过设计状态空间S(状态取值范围),其中的每个状态为表示为永磁同步电机的一种电磁转矩、转速、加速度和负载转矩的取值,通过状态信息和奖励判断电机运行状态的好坏,通过深度强化学习模型对数据进行寻优,能够及时的对电机控制进行调整,得到全局最优数据池。
进一步地,所述奖励函数设计如下:
r(s,a)=α1·Reward T(k+1)+α2·Reward v(k+1)+α3·Reward acc(k+1)+α4·Reward T1(k+1)
r(s,a)=α1·Reward T(k+1)+α2·Reward v(k+1)+α3·Reward acc(k+1)+α4·Reward T1(k+1)
其中,T(k)、v(k)、acc(k)、T1(k)分别表示执行动作a之前永磁同步电机的电磁转矩、转速、加速度和负载转矩;T(k+1)、v(k+1)、acc(k+1)、T1(k+1)分别表示执行动作a之后永磁同步电机的电磁转矩、转速、加速度和负载转矩;Reward T(k+1)、Reward v(k+1)、Reward acc(k+1)、Reward T1(k+1)分别表示动作a的电磁转矩、转速、加速度和负载转矩奖励;α1、α2、α3、α4为四种奖励因子的权重参数,其可以根据经验取值;r(s,a)表示当前状态s={T(k),v(k),acc(k),T1(k)}下执行动作a的奖励值。
进一步地,所述步骤2.3中,采用基于密度的局部异常因子法对全局最优数据池中的异常数据进行检测和剔除;
在实际工程中,电机系统有时会产生明显不符合常理的数据,称为异常点。对于基于深度强化学习的搜索方法而言,无论异常点是一过性的还是永久性的,都会对数据驱动方法的结果带来灾难性的影响。这是因为异常数据会给数据驱动方法带来“困惑”,使得无法挖掘数据内部的真实信息;
本方案采用基于密度的局部异常因子法对异常点进行检测和剔除。基于密度的局部异常因子法是通过比较每个点和其邻域点的密度来判断该点是否为异常点,如果点p的密度越低,越可能被认定是异常点。至于密度,则通过点与点之间的距离来计算,间距离越远,密度越低,距离越近,密度越高。因为该方法对密度是通过点的第k邻域来计算,而不是全局计算,因此得名为“局部”异常因子,而不会因为数据密度分散情况不同而错误的将正常点判定为异常点。具体步骤为:
步骤2.3.1:对于全局最优数据池中的每个数据点p,计算其第k距离邻域内各点的第k可达距离:
kreach(o,p)=max(dk(o),d(o,p))
kreach(o,p)=max(dk(o),d(o,p))
其中,dk(o)为邻域点o的第k距离,d(o,p)为邻域点o到点p的距离;
步骤2.3.2:计算每个数据点p的第k局部可达密度preach(p):
Nk(p)为点p的第k距离邻域;|Nk(p)|为Nk(p)的模;
步骤2.3.3:计算每个数据点的第k局部离群因子:
步骤2.3.4:分析局部离群因子,局部离群因子小于1,则不为异常点,若局部离群因子大于1,则为异常点;
步骤2.3.5:对异常点进行剔除。
所述步骤2.4具体包括以下步骤:
步骤2.4.1:初始化迭代次数t=1;从剔除异常数据后的全局最优数据池中随机选择n条样本数据,若n条样本数据中的奖励值r均大于所设定阈值,则结束,此时得到的值网络Qv即为最优深度强化学习模型;否则进入步骤2.4.2;
步骤2.4.2:将随机选择的n条样本数据作为值网络Qv和目标网络QT的输入;
针对每一条样本数据(s,a,r,s'),基于其中的状态s、动作a以及值网络Qv的参数θ,得到对应的预测Q值,其中第i条样本数据对应的预测Q值记为Qpredict(i),其中i=1,2,···,n;
针对每一条样本数据(s,a,r,s'),基于其中的下一状态s'及目标网络QT的参数θT,获得状态s'下执行状态空间中不同动作a'对应的Q值Q(s',a',θT),并取其中最大值并计算第i条样本数据对应的目标Q值QTarget(i):
其中,γ为奖励衰减因子,取[0,1]之间的数;ri为第i条样本数据中的奖励值;
步骤2.4.3:通过Qpredict(i)和QTarget(i)计算损失函数L:
步骤2.4.4:通过误差反向法最小化损失函数L,其中使用梯度下降法对值网络Qv的参数
θ进行更新;
其中,α为学习律,为经验参数;
步骤2.4.5:判断是否满足t=K,若满足,则更新目标网络QT的参数θT,令θT=θ,并令t=1;否则令t=t+1;其中K为目标网络参数更新周期;
步骤2.4.6:返回步骤2.2.3。
进一步地,所述步骤2.4.4中,α取0.01,以使误差值对新的Q值影响较小,同时具有有一定影响。
进一步地,所述步骤2.4.5中,设置目标网络参数更新周期K为3个仿真周期,每隔3个周期对目标网络参数进行更新能够及时求得新的目标网络,同时避免时间过长导致一直不更新目标网络,导致Q值陷入局部最优。
进一步,所述步骤3中,控制参数值包括最优d轴电流和最优q轴电流;首先基于最优深度强化学习模型预测出使电机运行在最优效率的最优d轴电流;根据最优d轴电流对永磁同步电机进行控制,同时根据最优d轴电流对永磁同步电机q轴电流进行补偿,实现永磁同步电机稳态下的效率最优控制。
电机在寻优过程中,d轴电流的变化会导致转速,电磁转矩发生波动,为了保持永磁同步电机控制系统的稳态,本方案采用调整d轴电流的同时,对q轴电流进行补偿从而维持转矩的恒定;
d轴电流调整之前电磁转矩为:
d轴电流变化Δid后,q轴电流相应调整为iq
*,可得调整后的电磁转矩为:
令调整前后的电磁转矩相等,可得最优q轴电流为:
q轴电流变化量Δiq为:
另一方面,提供一种永磁同步电机效率优化控制系统,包括存储器及处理器,所述存储器中存储有计算机程序,其特征在于,所述计算机程序被所述处理器执行时,使得所述处理器实现上述的永磁同步电机效率优化控制方法。
进一步地,所述系统还包括数据采集模块,用于采集永磁同步电机当前的状态数据,包括永磁同步电机当前的电磁转矩、转速、加速度和负载转矩,并送至所述处理器中。
本发明首次将深度强化学习算法与损耗模型法相结合,提出一种新的复合式方法用以解决电机效率优化问题,该方法能够继承两者的优点,克服双方的缺陷。电机损耗模型中将驱动器和永磁同步电机视为整体,建立统一效率优化模型,但是受到参数的影响,会导致效率不能达到全局最优;深度强化学习法是在模型法的基础上进一步寻优,能够实现系统效率的全局最优,且不受电机参数变化的影响,但算法收敛时间长,收敛速度慢,容易引起磁通和转矩的脉动,存在收敛速度和转矩脉动的矛盾,此时控制算法步长选择不当,则会使算法收敛过慢甚至不收敛,从而导致总体效率下降,严重影响效率最优控制效果。
本发明通过损耗模型法得到次优效率d轴电流,并将该电流作为深度强化学习搜索初始值,能够减少数据搜索的范围,缩短算法收敛时间,同时又避免了电机参数变化对结果的影响。并且采用了基于密度的局部异常因子法,能够避免异常数据会给深度强化学习方法带来“困惑”,使得无法挖掘数据内部的真实信息,从而减少算法搜索精度。同时,因为深度强化学习会带来转矩脉动,采用对q轴电流补偿能够减小转矩脉动并及时使q轴电流达到最优。此外,通过损耗模型法所得到的电磁转矩,转速,加速度和负载转矩为定值,而本发明结合深度强化学习,设定状态空间为四维向量B(T,v,acc,T1),其包含了电机当前电磁转矩、转速、加速度和负载转矩,结合电机当前状态计算奖励值,从而选取最适合的动作,实现电机的最优控制。
本发明提出永磁同步电机效率优化方法,兼具目前基于损耗模型的效率优化控制方法和基于搜索技术的效率优化控制方法之长,克服了两者其短。通过电机损耗模型,可以直观得到电机总损耗,加快电机响应速度,并能适用于各种场合。引入深度强化学习,能够使电机在参数变化的情况下,获得全局最优解,实现电机运行实时控制。将损耗模型法结合深度强化学习,能够利用电机损耗模型法建立次优数据池,然后在次优的基础上全局寻优,大大减少了数据搜索的范围,缩短算法收敛时间,同时又避免了电机参数变化对结果的影响,能够
对电机进行实时最优控制。
图1为永磁同步电机效率优化的矢量控制框图。
图2为考虑铁损的d轴等效电路。
图3为考虑铁损的q轴等效电路。
图4为电机效率优化深度强化学习框架。电机接收来自智能体的动作时,电机模型根据动作和当前状态,自动更新至下一步状态,同时将状态传递到根据目标模型确定的奖励函数,得到对应的奖励值,将状态向量和奖励值放入到数据池中
图5为DQN算法框架。
参照图1,本发明提出永磁同步电机效率优化控制方法,包括以下步骤:
以下结合附图和具体实施方式对本发明进行进一步具体说明。
实施例1:
本实施例提供一种永磁同步电机效率优化控制方法,包括以下步骤:
步骤1、利用损耗模型法求得永磁同步电机的近似最优d轴电流;
步骤2、以近似最优d轴电流为初始值,利用深度强化学习(深度Q网络,DQN)算法对d轴电流进行寻优,构建最优深度强化学习模型;
步骤3、将当前采集的永磁同步电机状态数据送入最优深度强化学习模型,基于该模型预测出使电机运行在最优效率的控制参数值(最优d轴电流);根据该控制参数值对永磁同步电机进行控制。
所述步骤1包括以下步骤:
首先,对电机损耗进行分析评估并构建永磁同步电机的损耗模型;
永磁同步电机的损耗包括铜损、铁损、杂散损耗和机械损耗。铜损是指电流通过定子绕组时产生的热损耗,由于电机定子中装有大量线圈,且工作时通过线圈的电流较大,使得铜耗是电机主要损耗之一;铁损是指磁通在电机铁芯中交变流动产生的磁滞与涡流损耗,它的大小取决于铁芯材料本身,与电流频率和转速有关;机械损耗和杂散损耗在电机总损耗中占比较小且不易控制与检测。因此本方案只考虑铁损和铜损。结合结合图2和图3的永磁同步电机考虑铜损和铁损的d,q轴等效电路,可以写出永磁同步电机的动态电压方程为:
其中,ud与uq分别为d、q轴电枢电压;id与iq分别为d、q轴电流(电枢电流);iwd与iwq为d、q轴定子电流有功分量;Rs为电枢电阻;Ld与Lq分别为d、q轴的电枢电感;ψm为转子永磁体的磁链;ωr为转子的机械角速度;np为电机极对数;
由永磁同步电机等效电路回路电压方程(KVL),可得:
其中,iFeq为d,q轴铁损电流分量,RFe为等效铁损电阻;
结合d,q轴铁损电流分量表达式,可得永磁同步电机总损耗为:
Te为永磁同步电机电磁转矩。
然后,基于永磁同步电机总损耗模型求得近似最优d轴电流;
稳态运行时,电机转速,转矩保持不变,从总损耗关系式可知,电机损耗只与iwd有关,要想总损耗最小,只需满足:
可得iwd为:
由永磁同步电机等效电路节点电流方程(KCL),可得:
id=iwd+iFed
由此得到近似最优d轴电流为:
电机在运行过程中,参数会不断发生变化,当采用损耗模型法时,所得d轴电流因为参数的影响,只能为近似最优d轴电流,为实现永磁同步电机效率最优控制,本发明继续通过深度强化学习对参数进行寻优,以得到电机控制的最优参数。
实施例2:
本实施例在实施例1的基础上,所述步骤2包括以下步骤:
步骤2.1:以深度强化学习算法为永磁同步电机的智能体Agent,构建深度强化学习模型,包括结构相同的值网络Qv和目标网络QT,状态空间S、动作空间A和奖励函数Reward,状态空间S中的每个状态为表示为一个四维向量(T,v,acc,T1),其中T、v、acc、T1分别表示永磁同步电机的电磁转矩、转速、加速度和负载转矩,动作空间A中的每个动作为d轴电流id的一种取值;
步骤2.2、利用当前的深度强化学习模型对次优数据即近似最优d轴电流进行寻优,得到全局最优数据池;
步骤2.3、对全局最优数据池中的异常数据进行检测和剔除;
步骤2.4、将剔除异常数据后的全局最优数据池中的样本数据,对深度强化学习模型进行训练,并判断是否满足终止条件,若满足,则终止,得到最优深度强化学习模型;否则返回步骤2.2。
实施例3:
本实施例在实施例2的基础上,,所述步骤2.2具体包括以下步骤:
步骤2.2.1:构建数据池D,其容量为C;
步骤2.2.2:测得电机当前状态s,包括永磁同步电机的电磁转矩、转速、加速度和负载转矩;将近似最优d轴电流记为动作a,执行动作a,即根据近似最优d轴电流对永磁同步电机进行控制,永磁同步电机自动更新(转移)至下一状态s'(该状态为永磁同步电机次优状态),通过奖励函数计算当前状态s下执行动作a的奖励值r;将当前样本数据(s,a,r,s')存入数据池D中;
步骤2.2.3:将当前状态s更新为s',将当前状态s输入至值网络Qv中,利用值网络Qv的参数θ得到当前状态s下状态空间中各个动作的预测Q值,采用ε-greedy策略选取最适合的动作,记为a;
步骤2.2.4:执行动作a,即根据近似最优d轴电流对永磁同步电机进行控制,永磁同步电机自动更新至下一状态s';通过奖励函数计算当前状态s下执行动作a的奖励值r;将当前样本数据(s,a,r,s')存入数据池D中;
步骤2.2.5:返回2.2.3进行循环,直至数据池D满(数据池D中的样本数据条数达到其容量C),将此时的数据池D作为全局最优数据池。
通过损耗模型法求得的近似最优d轴电流进而获得的永磁同步电机状态(包括电磁转矩,转速,加速度和负载转矩)为定值,且为永磁同步电机次优状态;本方案中通过设计状态空间S(状态取值范围),其中的每个状态为表示为永磁同步电机的一种电磁转矩、转速、加速度和负载转矩的取值,通过深度强化学习模型对数据进行寻优,可以得到全局最优数据池。以损耗模型法所得电机状态为深度强化学习算法的初始值,能够在电机搜索过程中减少试错次数,在次优的基础上全局寻优,大大减少训练学习时间。
实施例4:
本实施例在实施例3的基础上,将步骤2.2.1中的C取值为200,即数据池D中可以存放200个样本数据。
实施例5:
本实施例在实施例3的基础上,所述奖励函数设计如下:
r(s,a)=α1·Reward T(k+1)+α2·Reward v(k+1)+α3·Reward acc(k+1)+α4·Reward T1(k+1)
r(s,a)=α1·Reward T(k+1)+α2·Reward v(k+1)+α3·Reward acc(k+1)+α4·Reward T1(k+1)
其中,T(k)、v(k)、acc(k)、T1(k)分别表示执行动作a之前永磁同步电机的电磁转矩、
转速、加速度和负载转矩;T(k+1)、v(k+1)、acc(k+1)、T1(k+1)分别表示执行动作a之后永磁同步电机的电磁转矩、转速、加速度和负载转矩;Reward T(k+1)、Reward v(k+1)、Reward acc(k+1)、Reward T1(k+1)分别表示动作a的电磁转矩、转速、加速度和负载转矩奖励;α1、α2、α3、α4为四种奖励因子的权重参数,其可以根据经验取值;r(s,a)表示当前状态s={T(k),v(k),acc(k),T1(k)}下执行动作a的奖励值。
实施例6:
本实施例在实施例2的基础上,所述步骤2.3中,采用基于密度的局部异常因子法对全局最优数据池中的异常数据进行检测和剔除;
在实际工程中,电机系统有时会产生明显不符合常理的数据,称为异常点。对于基于深度强化学习的搜索方法而言,无论异常点是一过性的还是永久性的,都会对数据驱动方法的结果带来灾难性的影响。这是因为异常数据会给数据驱动方法带来“困惑”,使得无法挖掘数据内部的真实信息。通过对异常点进行检测和剔除,能够挖掘数据内部的真实信息,使电机运行状态更加平稳。
本方案采用基于密度的局部异常因子法对异常点进行检测和剔除。基于密度的局部异常因子法是通过比较每个点和其邻域点的密度来判断该点是否为异常点,如果点p的密度越低,越可能被认定是异常点。至于密度,则通过点与点之间的距离来计算,间距离越远,密度越低,距离越近,密度越高。因为该方法对密度是通过点的第k邻域来计算,而不是全局计算,因此得名为“局部”异常因子,而不会因为数据密度分散情况不同而错误的将正常点判定为异常点。具体步骤为:
步骤2.3.1:对于全局最优数据池中的每个数据点p,计算其第k距离邻域内各点的第k可达距离:
kreach(o,p)=max(dk(o),d(o,p))
kreach(o,p)=max(dk(o),d(o,p))
其中,dk(o)为邻域点o的第k距离,d(o,p)为邻域点o到点p的距离;
步骤2.3.2:计算每个数据点p的第k局部可达密度preach(p):
Nk(p)为点p的第k距离邻域;|Nk(p)|为Nk(p)的模;
步骤2.3.3:计算每个数据点的第k局部离群因子:
步骤2.3.4:分析局部离群因子,局部离群因子小于1,则不为异常点,若局部离群因子大于1,则为异常点;
步骤2.3.5:对异常点进行剔除。
实施例6:
本实施例在实施例3的基础上,所述步骤2.4具体包括以下步骤:
步骤2.4.1:初始化迭代次数t=1;从剔除异常数据后的全局最优数据池中随机选择n条样本数据,若n条样本数据中的奖励值r均大于所设定阈值,则结束,此时得到的值网络Qv即为最优深度强化学习模型;否则进入步骤2.4.2;
步骤2.4.2:将随机选择的n条样本数据作为值网络Qv和目标网络QT的输入;
针对每一条样本数据(s,a,r,s'),基于其中的状态s、动作a以及值网络Qv的参数θ,得到对应的预测Q值,其中第i条样本数据对应的预测Q值记为Qpredict(i),其中i=1,2,···,n;
针对每一条样本数据(s,a,r,s'),基于其中的下一状态s'及目标网络QT的参数θT,获得状态s'下执行状态空间中不同动作a'对应的Q值Q(s',a',θT),并取其中最大值并计算第i条样本数据对应的目标Q值QTarget(i):
其中,γ为奖励衰减因子,取[0,1]之间的数;ri为第i条样本数据中的奖励值;
步骤2.4.3:通过Qpredict(i)和QTarget(i)计算损失函数L:
步骤2.4.4:通过误差反向法最小化损失函数L,其中使用梯度下降法对值网络Qv的参数θ进行更新;
其中,α为学习律,为经验参数;
步骤2.4.5:判断是否满足t=K,若满足,则更新目标网络QT的参数θT,令θT=θ,并令t=1;否则令t=t+1;其中K为目标网络参数更新周期;
步骤2.4.6:返回步骤2.2.3。
实施例7:
本实施例在实施例6的基础上,所述步骤2.4.4中,α取0.01,以使误差值对新的Q值影响较小,同时具有有一定影响。
实施例8:
本实施例在实施例6的基础上,所述步骤2.4.5中,设置目标网络参数更新周期K为3个仿真周期,每隔3个周期对目标网络参数进行更新能够及时求得新的目标网络,同时避免时间过长导致一直不更新目标网络,导致Q值陷入局部最优。
实施例9:
本实施例在实施例2的基础上,所述步骤3中,控制参数值包括最优d轴电流和最优q轴电流;首先将当前采集的实时数据送入最优深度强化学习模型,基于最优深度强化学习模型预测出使电机运行在最优效率的最优d轴电流;根据最优d轴电流对永磁同步电机进行控制,同时根据最优d轴电流对永磁同步电机q轴电流进行补偿,实现永磁同步电机稳态下的效率最优控制。
本方案考虑是在实际电机中实现,所以需要保证电机稳态运行才能实现效率最优。电机在寻优过程中,d轴电流的变化会导致转速,电磁转矩发生波动,为了保持永磁同步电机控制系统的稳态,本方案采用调整d轴电流的同时,对q轴电流进行补偿从而维持转矩的恒定;
d轴电流调整之前电磁转矩为:
d轴电流变化Δid后,q轴电流相应调整为iq
*,可得调整后的电磁转矩为:
令调整前后的电磁转矩相等,可得最优q轴电流为:
q轴电流变化量Δiq为:
实施例10:
本实施例提供一种永磁同步电机效率优化控制系统,包括存储器及处理器,所述存储器中存储有计算机程序,其特征在于,所述计算机程序被所述处理器执行时,使得所述处理器实现上述实施例所述的永磁同步电机效率优化控制方法。
实施例11:
本实施例在实施例10的基础上,所述系统还包括数据采集模块,用于采集永磁同步电机当前的状态数据,包括永磁同步电机当前的电磁转矩、转速、加速度和负载转矩,并送至所述处理器中。
本发明首次将深度强化学习算法与损耗模型法相结合,提出一种新的复合式方法用以解决电机效率优化问题,该方法能够继承两者的优点,克服双方的缺陷。电机损耗模型中将驱动器和永磁同步电机视为整体,建立统一效率优化模型,但是受到参数的影响,会导致效率不能达到全局最优;深度强化学习法是在模型法的基础上进一步寻优,能够实现系统效率的全局最优,且不受电机参数变化的影响,但算法收敛时间长,收敛速度慢,容易引起磁通和转矩的脉动,存在收敛速度和转矩脉动的矛盾,此时控制算法步长选择不当,则会使算法收敛过慢甚至不收敛,从而导致总体效率下降,严重影响效率最优控制效果。
本发明通过损耗模型法得到次优效率d轴电流,并将该电流作为深度强化学习搜索初始值,能够减少数据搜索的范围,缩短算法收敛时间,同时又避免了电机参数变化对结果的影响。并且采用了基于密度的局部异常因子法,能够避免异常数据会给深度强化学习方法带来“困惑”,使得无法挖掘数据内部的真实信息,从而减少算法搜索精度。同时,因为深度强化学习会带来转矩脉动,采用对q轴电流补偿能够减小转矩脉动并及时使q轴电流达到最优。此外,通过损耗模型法所得到的电磁转矩,转速,加速度和负载转矩为定值,而本发明结合深度强化学习,设定状态空间为四维向量B(T,v,acc,T1),其包含了电机当前电磁转矩、转速、加速度和负载转矩,结合电机当前状态计算奖励值,从而选取最适合的动作,实现电机的最优控制。
综上所述,本发明提供一种永磁同步电机效率优化控制方法和系统,能使电机在运行时,
在效率达到全局最优的基础上,具有搜索时间短、收敛速度快、能够实现电机运行实时控制并适用于各种场合等优点;达到节能减排、降低损耗并且提高效率。本发明提出的方法和系统简单、容易理解、易于实现、准确度高、是一种能够有效提高永磁同步电机效率的可行性方案。
以上所述,仅为本发明较佳的具体实施方式,但本发明的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本发明揭露的技术范围内,根据本发明的技术方案及其发明构思加以等同替换或改变,都应涵盖在本发明的保护范围之内。
Claims (10)
- 一种永磁同步电机效率优化控制方法,其特征在于,包括以下步骤:步骤1、利用损耗模型法求得永磁同步电机的近似最优d轴电流;步骤2、以近似最优d轴电流为初始值,利用深度强化学习算法对d轴电流进行寻优,构建最优深度强化学习模型;步骤3、将当前采集的永磁同步电机状态数据送入最优深度强化学习模型,基于该模型预测出使电机运行在最优效率的控制参数值;根据该控制参数值对永磁同步电机进行控制。
- 根据权利要求1所述的永磁同步电机效率优化控制方法,其特征在于,所述步骤2包括以下步骤:步骤2.1:构建深度强化学习模型,包括结构相同的值网络Qv和目标网络QT,状态空间S、动作空间A和奖励函数Reward,状态空间S中的每个状态为表示为一个四维向量(T,v,acc,T1),其中T、v、acc、T1分别表示永磁同步电机的电磁转矩、转速、加速度和负载转矩,动作空间A中的每个动作为d轴电流id的一种取值;步骤2.2、利用当前的深度强化学习模型对次优数据即近似最优d轴电流进行寻优,得到全局最优数据池;步骤2.3、对全局最优数据池中的异常数据进行检测和剔除;步骤2.4、将剔除异常数据后的全局最优数据池中的样本数据,对深度强化学习模型进行训练,并判断是否满足终止条件,若满足,则终止,得到最优深度强化学习模型;否则返回步骤2.2。
- 根据权利要求2所述的永磁同步电机效率优化控制方法,其特征在于,所述步骤2.2具体包括以下步骤:步骤2.2.1:构建数据池D,其容量为C;步骤2.2.2:测得电机当前状态s,包括永磁同步电机的电磁转矩、转速、加速度和负载转矩;将近似最优d轴电流记为动作a,执行动作a,即根据近似最优d轴电流对永磁同步电机进行控制,永磁同步电机自动更新至下一状态s',通过奖励函数计算当前状态s下执行动作a的奖励值r;将当前样本数据(s,a,r,s')存入数据池D中;步骤2.2.3:将当前状态s更新为s',将当前状态s输入至值网络Qv中,利用值网络Qv的参数θ得到当前状态s下状态空间中各个动作的预测Q值,采用ε-greedy策略选取最适合的动作,记为a;步骤2.2.4:执行动作a,即根据近似最优d轴电流对永磁同步电机进行控制,永磁同步电机自动更新至下一状态s';通过奖励函数计算当前状态s下执行动作a的奖励值r;将当前样本数据(s,a,r,s')存入数据池D中;步骤2.2.5:返回2.2.3进行循环,直至数据池D满,将此时的数据池D作为全局最优数据池。
- 根据权利要求3所述的永磁同步电机效率优化控制方法,其特征在于,所述奖励函数设计如下:
其中,T(k)、v(k)、acc(k)、T1(k)分别表示执行动作a之前永磁同步电机的电磁转矩、转速、加速度和负载转矩;T(k+1)、v(k+1)、acc(k+1)、T1(k+1)分别表示执行动作a之后永磁同步电机的电磁转矩、转速、加速度和负载转矩;Reward T(k+1)、Reward v(k+1)、Reward acc(k+1)、Reward T1(k+1)分别表示动作a的电磁转矩、转速、加速度和负载转矩奖励;α1、α2、α3、α4为四种奖励因子的权重参数;r(s,a)表示当前状态s={T(k),v(k),acc(k),T1(k)}下执行动作a的奖励值。 - 根据权利要求3所述的永磁同步电机效率优化控制方法,其特征在于,所述步骤2.4具体包括以下步骤:步骤2.4.1:初始化迭代次数t=1;从剔除异常数据后的全局最优数据池中随机选择n条样本数据,若n条样本数据中的奖励值r均大于所设定阈值,则结束,此时得到的值网络Qv即为最优深度强化学习模型;否则进入步骤2.4.2;步骤2.4.2:将随机选择的n条样本数据作为值网络Qv和目标网络QT的输入;针对每一条样本数据(s,a,r,s'),基于其中的状态s、动作a以及值网络Qv的参数θ,得到对应的预测Q值,其中第i条样本数据对应的预测Q值记为Qpredict(i),其中i=1,2,···,n;针对每一条样本数据(s,a,r,s'),基于其中的下一状态s'及目标网络QT的参数θT,获得状态s'下执行状态空间中不同动作a'对应的Q值Q(s',a',θT),并取其中最大值并计算第i条样本数据对应的目标Q值QTarget(i):
其中,γ为奖励衰减因子,取[0,1]之间的数;ri为第i条样本数据中的奖励值;步骤2.4.3:通过Qpredict(i)和QTarget(i)计算损失函数L:
步骤2.4.4:通过误差反向法最小化损失函数L,其中使用梯度下降法对值网络Qv的参数θ进行更新;
其中,α为学习律;步骤2.4.5:判断是否满足t=K,若满足,则更新目标网络QT的参数θT,令θT=θ,并令t=1;否则令t=t+1;其中K为目标网络参数更新周期;步骤2.4.6:返回步骤2.2.3。 - 根据权利要求5所述的永磁同步电机效率优化控制方法,其特征在于,所述步骤2.4.4中,α取0.01。
- 根据权利要求5所述的永磁同步电机效率优化控制方法,其特征在于,所述步骤2.4.5中,设置目标网络参数更新周期K为3个仿真周期。
- 根据权利要求5所述的永磁同步电机效率优化控制方法,其特征在于,所述步骤3中,控制参数值包括最优d轴电流和最优q轴电流;首先基于最优深度强化学习模型预测出使电机运行在最优效率的最优d轴电流;根据最优d轴电流对永磁同步电机进行控制,同时根据 最优d轴电流对永磁同步电机q轴电流进行补偿,实现永磁同步电机稳态下的效率最优控制,q轴电流补偿量Δiq为:
其中,Ld与Lq分别为d、q轴的电枢电感;ψm为转子永磁体的磁链;id为调整之前的d轴电流,Δid为调整之后d轴电流变化量。 - 一种永磁同步电机效率优化控制方法,包括存储器及处理器,所述存储器中存储有计算机程序,其特征在于,所述计算机程序被所述处理器执行时,使得所述处理器实现如权利要求1~8中任一项所述的永磁同步电机效率优化控制方法。
- 根据权利要求9所述的永磁同步电机效率优化控制方法,其特征在于,还包括数据采集模块,用于采集永磁同步电机当前的状态数据,包括永磁同步电机当前的电磁转矩、转速、加速度和负载转矩,并送至所述处理器中。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US18/239,994 US20230412098A1 (en) | 2022-05-18 | 2023-08-30 | Efficiency optimization control method and system for permanent magnet synchronous motor |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210542923.4 | 2022-05-18 | ||
CN202210542923.4A CN115149859A (zh) | 2022-05-18 | 2022-05-18 | 永磁同步电机效率优化控制方法及系统 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US18/239,994 Continuation US20230412098A1 (en) | 2022-05-18 | 2023-08-30 | Efficiency optimization control method and system for permanent magnet synchronous motor |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023221740A1 true WO2023221740A1 (zh) | 2023-11-23 |
Family
ID=83405903
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2023/090074 WO2023221740A1 (zh) | 2022-05-18 | 2023-04-23 | 永磁同步电机效率优化控制方法及系统 |
Country Status (3)
Country | Link |
---|---|
US (1) | US20230412098A1 (zh) |
CN (1) | CN115149859A (zh) |
WO (1) | WO2023221740A1 (zh) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115149859A (zh) * | 2022-05-18 | 2022-10-04 | 湘潭大学 | 永磁同步电机效率优化控制方法及系统 |
CN116526922A (zh) * | 2023-07-04 | 2023-08-01 | 浙江大学 | 一种基于改进拟牛顿法的混合型最小损耗控制方法及系统 |
CN116931431B (zh) * | 2023-07-26 | 2024-06-07 | 江南大学 | 一种永磁同步电机模型预测控制权重系数设计方法和系统 |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090237014A1 (en) * | 2007-05-24 | 2009-09-24 | Aisin Seiki Kabushiki Kaisha | Synchronous motor control device and method for optimizing synchronous motor control |
CN110995076A (zh) * | 2019-12-11 | 2020-04-10 | 湖南大学 | 一种永磁同步电机模型预测电流控制方法 |
CN113014170A (zh) * | 2021-04-21 | 2021-06-22 | 昆明理工大学 | 一种基于鲸鱼优化算法的永磁同步电机最小损耗控制方法 |
CN113726253A (zh) * | 2021-09-03 | 2021-11-30 | 安徽大学 | 一种提高电动汽车用永磁同步电机效率的方法 |
CN114123896A (zh) * | 2021-11-30 | 2022-03-01 | 江南大学 | 基于增量式宽度学习系统的永磁同步电机控制方法及系统 |
CN115149859A (zh) * | 2022-05-18 | 2022-10-04 | 湘潭大学 | 永磁同步电机效率优化控制方法及系统 |
-
2022
- 2022-05-18 CN CN202210542923.4A patent/CN115149859A/zh active Pending
-
2023
- 2023-04-23 WO PCT/CN2023/090074 patent/WO2023221740A1/zh unknown
- 2023-08-30 US US18/239,994 patent/US20230412098A1/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090237014A1 (en) * | 2007-05-24 | 2009-09-24 | Aisin Seiki Kabushiki Kaisha | Synchronous motor control device and method for optimizing synchronous motor control |
CN110995076A (zh) * | 2019-12-11 | 2020-04-10 | 湖南大学 | 一种永磁同步电机模型预测电流控制方法 |
CN113014170A (zh) * | 2021-04-21 | 2021-06-22 | 昆明理工大学 | 一种基于鲸鱼优化算法的永磁同步电机最小损耗控制方法 |
CN113726253A (zh) * | 2021-09-03 | 2021-11-30 | 安徽大学 | 一种提高电动汽车用永磁同步电机效率的方法 |
CN114123896A (zh) * | 2021-11-30 | 2022-03-01 | 江南大学 | 基于增量式宽度学习系统的永磁同步电机控制方法及系统 |
CN115149859A (zh) * | 2022-05-18 | 2022-10-04 | 湘潭大学 | 永磁同步电机效率优化控制方法及系统 |
Also Published As
Publication number | Publication date |
---|---|
US20230412098A1 (en) | 2023-12-21 |
CN115149859A (zh) | 2022-10-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2023221740A1 (zh) | 永磁同步电机效率优化控制方法及系统 | |
Chen et al. | Intelligent maximum power factor searching control using recurrent Chebyshev fuzzy neural network current angle controller for SynRM drive system | |
Kumar et al. | Efficiency optimization of induction motor drive using soft computing techniques | |
Liu et al. | Second-order ESO-based current sensor fault-tolerant strategy for sensorless control of PMSM with B-phase current | |
CN114938173A (zh) | 永磁同步电机效率优化控制方法及装置 | |
Senjyu et al. | High efficiency control of synchronous reluctance motors using extended Kalman filter | |
Sun et al. | Efficient feedback linearization control for an IPMSM of EVs based on improved firefly algorithm | |
Cao et al. | Optimal control strategy of state feedback control for surface-mounted PMSM drives based on auto-tuning of seeker optimization algorithm | |
Zhou et al. | Parameter identification of permanent magnet synchronous motor based on modified-fuzzy particle swarm optimization | |
Chaouch et al. | Optimized torque control via backstepping using genetic algorithm of induction motor | |
Guo et al. | Research on a new adaptive integral sliding mode controller based on a small BLDC | |
Kim et al. | Vector control for loss minimization of induction motor using GA–PSO | |
CN111162714A (zh) | 考虑参数失配的多相储能永磁电机鲁棒预测转矩控制方法、系统及介质 | |
Ben et al. | Torque ripple reduction for switched reluctance motors using global optimization algorithm | |
CN113343592B (zh) | 一种新能源飞机的永磁同步电机dqn智能控制方法 | |
Kortas et al. | Analytical solution of optimized energy consumption of Double Star Induction Motor operating in transient regime using a Hamilton–Jacobi–Bellman equation | |
CN113691179A (zh) | 一种基于固定时间的变幂次指数趋近律的永磁同步电机滑模控制方法 | |
Peng | Predictive Current Control of Permanent Magnet Synchronous Motor Based on Parameter Identification | |
Zhang et al. | Vector control of permanent magnet synchronous motor drive system based on new sliding mode control | |
Xiao et al. | Loss minimization control for interior permanent magnet synchronous motor in a wide speed range | |
Mahmoudi et al. | Neuro-Genetic sensorless sliding mode control of a permanent magnet synchronous motor using Luenberger observer | |
Song et al. | Design of PMSM Dual‐Loop Control Systems Integrating LADRC and PI Controllers via an Improved PSO Algorithm | |
Zhu et al. | Traction system research of High-Speed Maglev Train Based on ADRC | |
Zhang et al. | Parameter Identification of PMSWG Based on ASMDRPSO. | |
Cao et al. | Digital implementation of DTC based on PSO for induction Motors |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23806693 Country of ref document: EP Kind code of ref document: A1 |