A motor thermal state estimation method and system based on dual-target parameter identification and Kalman filtering, and related equipment
Patent Information
- Application Number
- CN202611517102.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-29
- Publication Date
- 2026-10-30
AI Technical Summary
[0024]本发明实施例提供的上述技术方案的有益效果至少包括：
Smart Images
Figure CN123019572A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of motor thermal management technology, and more specifically to a method, system and related equipment for estimating the thermal state of a motor based on dual-objective parameter identification and Kalman filtering. Background Technology
[0002] Permanent magnet synchronous motors (PMSMs) are widely used in high-performance applications such as new energy vehicles due to their advantages of high power density, high efficiency, and wide speed range. Under high speed and heavy load conditions, continuous accumulation of losses leads to a rapid increase in the internal temperature of the motor. Achieving high-precision online estimation of the critical thermal node temperature of PMSMs is of significant engineering importance for motor thermal protection and safe operation.
[0003] Lumped Parameter Thermal Network (LPTN) models are widely used for PMSM temperature prediction and thermal state estimation due to their high computational efficiency and clear physical meaning. LPTN abstracts the complex heat transfer process inside the motor into an equivalent network composed of thermal resistance and thermal capacity, and obtains the temperature of each node by solving the thermal network equations.
[0004] However, existing LPTN-based thermal state estimation techniques suffer from the following drawbacks: First, most existing LPTN model parameter identification methods employ a single-objective optimization framework, resulting in strong coupling between thermal resistance and thermal capacity parameters. The identified models often struggle to simultaneously achieve agile transient response and accurate steady-state temperature prediction, leading to a significant decrease in estimation accuracy under conditions of drastic load changes and large temperature fluctuations. Second, treating thermal network parameter identification and Kalman Filter (KF) state estimation as two independent stages results in estimation performance highly dependent on the quality of the identified parameters. When changes in operating conditions cause parameter drift, parameter mismatch will further propagate and amplify through the KF prediction-correction process, degrading temperature estimation accuracy. Third, although some methods have attempted to introduce multi-objective optimization to construct a Pareto front to balance different performance indicators, in high-dimensional, strongly coupled, and nonlinear thermal network parameter identification problems, traditional Particle Swarm Optimization (PSO) algorithms are prone to premature convergence, insufficient local optimization capabilities, and uneven solution set distribution. The Pareto optimal solution set obtained in this way still has shortcomings in terms of convergence, diversity and local accuracy, which affects the selection of parameters for subsequent thermal network models and the performance of online thermal state estimation. Summary of the Invention
[0005] This invention proposes a method, system, and related equipment for estimating the thermal state of a motor based on dual-objective parameter identification and Kalman filtering, used for motor temperature estimation. This method decouples the inherent contradiction between transient response performance and steady-state temperature accuracy during the thermal network parameter identification process, organically integrating parameter identification with Kalman filtering state estimation. This avoids the propagation and amplification of parameter mismatch during the estimation process, improves the quality of the Pareto optimal solution set obtained through multi-objective optimization, and yields an optimal thermal network model that balances transient response and steady-state accuracy.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] In a first aspect, embodiments of the present invention provide a method for estimating the thermal state of a motor based on dual-target parameter identification and Kalman filtering, comprising the following steps: S1. Obtain motor operation data, construct a five-node low-order thermal equivalent network model and the corresponding continuous-time state-space model, and discretize the continuous-time state-space model to obtain a discrete-time state-space model. S2. Based on the preset temperature change rate threshold, the motor temperature rise process is divided into transient stage and steady-state stage; based on the discrete-time state-space model, the response deviation of the transient stage and the thermal balance deviation of the steady-state stage are used as dual-objective optimization criteria to construct a dual-objective Pareto parameter identification model. S3. The particle swarm-whale collaborative optimization search strategy is used to solve the bi-objective Pareto parameter identification model to obtain the Pareto optimal solution set, and the optimal thermal network parameters that take into account both transient response performance and steady-state balance performance are determined from the Pareto optimal solution set according to the knee point criterion. S4. Substitute the optimal thermal network parameters into the discrete-time state-space model, and predict and correct the state variables at the current moment based on the Kalman filter algorithm to recursively estimate the temperature state of the key nodes of the motor online.
[0008] Furthermore, in step S1, constructing a five-node low-order thermal equivalent network model and its corresponding continuous-time state-space model specifically includes: For a five-node low-order thermal network model, a continuous-time state-space model is established based on the thermal network topology:
[0009] in, This indicates the rate of temperature change per unit time at each thermal node; Represents the system state vector. Indicates the temperature of the stator teeth. Indicates the stator winding temperature. The stator yoke temperature, Indicates the temperature of the permanent magnet. Indicates the coolant temperature. Represents the continuous-time state transition matrix. Represents the hot input matrix; This represents the input vector of node heat loss.
[0010] Furthermore, step S2 specifically includes: The temperature change rate of each thermal node is calculated based on historical temperature data, and the motor temperature rise process is divided into transient and steady-state stages according to a preset temperature change rate threshold. When the temperature change rate of each thermal node is lower than the set threshold, the temperature rise process continues. If the motor thermal process is in a steady state, it is determined to be in a transient state; otherwise, it is determined to be in a transient state. Temperature error objective functions are constructed for both the transient and steady-state phases. Using the dynamic response deviation in the transient phase and the thermal balance deviation in the steady-state phase as dual-objective optimization criteria, a dual-objective Pareto optimization model for thermal network parameter identification is established, with the following expression:
[0011] in, This represents the vector of thermal network parameters to be identified, including thermal resistance and heat capacity parameters; The sum of squared errors between the model output temperature and the measured temperature during the transient phase is represented by the first optimization objective. The sum of squared errors between the model output temperature and the measured temperature during the steady-state phase represents the second optimization objective.
[0012] Furthermore, the sum of squared errors between the transient phase model output temperature and the measured temperature, and the sum of squared errors between the steady-state phase model output temperature and the measured temperature are specifically defined as follows:
[0013]
[0014] in, This represents the total number of sampling points during the transient phase. This indicates that under given thermal network parameters In the case of the LPTN model, the first... The hot node is at the _ ... Predicted temperature at each moment; Indicates the first The node at the th The temperature actually collected by the sensor at a given moment; This represents the total number of sampling points during the steady-state phase; Indicates the total length of the data.
[0015] Furthermore, in step S3, a particle swarm-whale collaborative optimization search strategy is used to solve the bi-objective Pareto parameter identification model to obtain the Pareto optimal solution set, specifically including: The particle swarm optimization algorithm is used to globally explore the parameter space of the thermal network, construct the Pareto front skeleton, and use an external archive to save all non-dominated solutions generated during the iteration process. Based on the non-dominated solutions in the external archive, the whale optimization algorithm is used to perform a local fine search of the neighborhood of the non-dominated solutions, update the candidate parameter positions, and update the external archive according to the non-dominated relationship.
[0016] Furthermore, in step S3, the optimal thermal network parameters that balance transient response performance and steady-state equilibrium performance are determined from the Pareto optimal solution set according to the knee point criterion, specifically including: The ideal solution is taken as the point where both transient and steady-state errors are zero. Calculate the Euclidean distance from each solution in the Pareto optimal solution set to the ideal solution; Calculate the Euclidean distance between each Pareto optimal solution and the defined ideal point, and take the Pareto optimal solution with the smallest Euclidean distance as the optimal heat network parameter, expressed as:
[0017] in, Indicates the optimal thermal network parameters. This represents the Euclidean distance from the Pareto solution to the ideal point.
[0018] Furthermore, step S4 specifically includes: Substitute the optimal thermal network parameters into the discrete state-space model to establish the discrete system state equations and observation equations. Perform prior state prediction and prior error covariance prediction; Calculate the Kalman gain and use the temperature measurement values to perform posterior correction on the prior state prediction values; The stator winding temperature is extracted from the updated state vector as the temperature estimate of the key thermodynamic node.
[0019] Furthermore, the state equations and observation equations of the discrete system are specifically expressed as follows:
[0020]
[0021] in, This represents the state vector indicating the temperature of each key component of the motor at the next moment. This represents the state vector indicating the temperature of each key component of the motor at the current moment. The key components of the motor include: stator teeth, stator windings, stator yoke, permanent magnets, and coolant. Represents the discrete state transition matrix. , Represents the identity matrix. Represents the heat loss input vector. Indicates process noise. Indicates measurement noise. This represents the temperature measurement output vector. This represents the observation matrix.
[0022] In a second aspect, embodiments of the present invention provide a motor thermal state estimation system based on dual-target parameter identification and Kalman filtering, applying the method described in the first aspect, the system comprising: The model building module is used to acquire motor operation data, construct a five-node low-order thermal equivalent network model and the corresponding continuous-time state-space model, and discretize the model to obtain a discrete-time state-space model. The phase division and optimization modeling module is used to divide the motor temperature rise process into transient and steady-state stages based on a preset temperature change rate threshold. Based on the discrete-time state-space model, the response deviation of the transient stage and the thermal balance deviation of the steady-state stage are used as dual-objective optimization criteria to construct a dual-objective Pareto parameter identification model. The collaborative optimization solution module is used to solve the bi-objective Pareto parameter identification model using a particle swarm-whale collaborative optimization search strategy, obtain the Pareto optimal solution set, and determine the optimal thermal network parameters that take into account both transient response performance and steady-state balance performance from the Pareto optimal solution set according to the knee point criterion. The state estimation module is used to substitute the optimal thermal network parameters into the discrete-time state-space model, and predict and correct the state variables at the current moment based on the Kalman filter algorithm, so as to recursively estimate the temperature state of the key nodes of the motor online.
[0023] Thirdly, embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the motor thermal state estimation method based on dual-target parameter identification and Kalman filtering as described in the first aspect.
[0024] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following: This invention constructs a dual-objective Pareto optimization model targeting both transient and steady-state errors. It employs a particle swarm optimization-whale joint optimization approach to perform global skeleton search and local fine-tuning of the Pareto front. Finally, the knee criterion is used to select optimal thermal network parameters from the Pareto non-dominated solution set and substitute them into the Kalman filter state-space model, achieving high-precision online estimation of the temperature of key motor components.
[0025] This invention effectively overcomes the inherent contradiction between transient and steady-state constraints in single-objective identification methods by decoupling and optimizing the transient response performance dominated by heat capacity parameters and the steady-state temperature accuracy dominated by thermal resistance parameters during temperature rise. It also effectively overcomes the premature convergence defect of traditional particle swarm optimization algorithms in high-dimensional nonlinear parameter identification problems. The optimal thermal network parameters obtained based on the knee criterion are embedded into the Kalman state-space model, enabling parameter identification and thermal state estimation to work collaboratively within the same framework. This avoids the problem of parameter mismatch propagating and amplifying through the Kalman prediction-correction process, which is common in traditional separate strategies. Attached Figure Description
[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0027] Figure 1 This is a flowchart of a method for estimating the thermal state of a motor based on dual-target parameter identification and Kalman filtering, provided in an embodiment of the present invention.
[0028] Figure 2 The diagram shows the Pareto front distribution results obtained from the bi-objective Pareto optimization model provided in this embodiment of the invention.
[0029] Figure 3 This is a diagram showing the dynamic tracking results of different thermal state estimation methods provided in the embodiments of the present invention at the stator winding temperature nodes.
[0030] Figure 4 This is a structural diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0032] Example 1: like Figure 1 As shown, this embodiment of the invention discloses a method for estimating the thermal state of a motor based on dual-target parameter identification and Kalman filtering, including the following steps: S1. Obtain motor operating data, construct a five-node low-order thermal equivalent network model and the corresponding continuous-time state-space model, and discretize the continuous-time state-space model to obtain a discrete-time state-space model; specifically: identify key thermal dynamic nodes based on the internal heat transfer mechanism of the permanent magnet synchronous motor, and determine the thermal resistance path and thermal capacity parameters that characterize the heat transfer capability between each node; establish the thermal balance equation of the five-node low-order thermal network model based on the law of conservation of energy and discretize the model, and construct the thermal resistance and thermal capacity as the set of parameters to be identified in the thermal dynamic system.
[0033] S2. Based on the preset temperature change rate threshold, the motor temperature rise process is divided into transient and steady-state stages. Based on the discrete-time state-space model, the response deviation in the transient stage and the thermal balance deviation in the steady-state stage are used as dual-objective optimization criteria to construct a dual-objective Pareto parameter identification model. Specifically, the transient and steady-state stages are divided according to the temperature change rate threshold. A dual-objective Pareto optimization model is established with minimizing the transient temperature error as the first optimization objective and minimizing the steady-state temperature error as the second optimization objective, thus establishing an evaluation system for subsequent thermal network parameter identification.
[0034] S3. A particle swarm-whale collaborative optimization search strategy is adopted to solve the bi-objective Pareto parameter identification model to obtain the Pareto optimal solution set. Based on the knee criterion, the optimal thermal network parameters that take into account both transient response performance and steady-state balance performance are determined from the Pareto optimal solution set. Specifically, a particle swarm-whale collaborative search strategy is adopted to perform multi-objective optimization on the set of parameters to be identified, to obtain the Pareto optimal solution set that represents the trade-off between the first and second optimization objectives. Based on the knee criterion, the optimal thermal network parameters that take into account both transient response and steady-state balance are determined from the Pareto optimal solution set.
[0035] S4. Substitute the optimal thermal network parameters into the discrete-time state-space model, and use the Kalman filter algorithm to predict and correct the state variables at the current moment to recursively estimate the temperature state of key nodes in the motor online. Specifically, substitute the optimized thermal network parameters into the discrete-time thermal state prediction model, and use the Kalman filter algorithm to fuse the model prediction values and temperature measurement values. Through prior state prediction, Kalman gain calculation, and posterior state correction, the temperature state of key thermodynamic nodes in the permanent magnet synchronous motor is recursively estimated online.
[0036] This invention decouples the inherent contradiction between transient response performance and steady-state temperature accuracy during the thermal network parameter identification process when estimating motor temperature. It effectively solves the problem of decreased identification accuracy caused by the coupling of thermal resistance and thermal capacity parameters in traditional single-objective parameter identification. A particle swarm optimization-whale collaborative search strategy is employed to solve the dual-objective optimization model, obtaining a Pareto optimal solution set. Based on the knee criterion, the optimal thermal network parameters that balance transient response and steady-state equilibrium are determined from this set and embedded into a discrete-time thermal state-space model. Combined with Kalman filtering, this enables online recursive estimation of the temperatures of key thermal nodes.
[0037] The following example, using the stator windings of a motor, provides a detailed explanation of the motor thermal state estimation method based on dual-objective parameter identification and Kalman filtering of the present invention: Step S1: Obtain motor operating data, including motor speed, d-axis current, q-axis current, and coolant temperature; construct a five-node low-order thermal equivalent network model and its corresponding continuous-time state-space model; discretize the model to obtain a discrete-time state-space model, specifically including: Step S11: Based on the thermal network topology, using the principle of energy conservation, the rate of temperature change for any thermal node i is described by the heat balance equation, expressed as:
[0038] in, This represents the equivalent heat capacity (J / K) of node i. This represents the temperature (K) of node j. Indicates time (s). The temperature (K) of node i is represented. This represents the heat generation power (W) at node i. This represents the equivalent thermal resistance (K / W) between node i and node j. This represents the set of neighboring nodes that are thermally connected to node i.
[0039] Step S12: Assemble the temperature of each node into a state vector. Based on the topological relationships of the thermal network, a continuous-time state-space model can be further established:
[0040] in, This indicates the rate of temperature change per unit time at each thermal node; Represents the system state vector. Indicates the temperature of the stator teeth. Indicates the stator winding temperature. Indicates the stator yoke temperature. Indicates the temperature of the permanent magnet. Indicates the coolant temperature; Represents the continuous-time state transition matrix; Represents the hot input matrix, Represents the heat source allocation matrix; This represents the input vector of node heat loss.
[0041] The heat capacity matrix is represented as:
[0042] in, This represents the equivalent heat capacity of the stator winding; This represents the equivalent heat capacity of the stator core; This represents the equivalent heat capacity of the stator yoke; This represents the equivalent heat capacity of a permanent magnet; This indicates the equivalent heat capacity of the coolant.
[0043] Thermal conductivity matrix Based on the network topology, its elements are determined by the thermal conductivity between nodes, and are expressed as:
[0044] in, Indicates the thermal conductivity between the stator core and the stator windings; Indicates the thermal conductivity between the stator core and the stator yoke; This indicates the thermal conductivity between the stator core and the permanent magnet; This indicates the thermal conductivity between the stator yoke and the coolant. This indicates the thermal conductivity between the permanent magnet and the coolant; This indicates the thermal conductivity between the stator yoke and the environment. This indicates the thermal conductivity between the coolant and the environment.
[0045] To achieve online recursive calculations in digital control systems, the matrix exponentiation method is used to discretize the continuous model:
[0046] in, Represents the discrete state transition matrix. , Represents the identity matrix.
[0047] Step S13: In the low-order thermal network model, the heat source term is physically related to the motor operating variables.
[0048] The heat loss input mainly consists of copper loss, iron loss, and permanent magnet loss, and their expressions are as follows:
[0049]
[0050]
[0051] in, and They represent PMSM respectively shaft and shaft current, This represents the stator winding resistance. This indicates the mechanical speed of the motor (rad / s). and These represent the empirical coefficients for iron loss, This represents the loss coefficient of the permanent magnet.
[0052] Step S2: Based on a preset temperature change rate threshold, the motor temperature rise process is divided into a transient stage and a steady-state stage; based on the discrete-time state-space model, using the response deviation of the transient stage and the thermal balance deviation of the steady-state stage as dual-objective optimization criteria, a dual-objective Pareto parameter identification model is constructed, specifically including: Step S21: Perform transient-steady-state decoupling on the temperature rise process. This occurs when the temperature change rate at all nodes simultaneously falls below a set threshold. When the system is considered to have entered a thermal steady state, the criterion is defined as follows:
[0053] in, / step is a set threshold, when the condition is met The conditions are maintained continuously The system is considered to have entered a steady state after one sampling period. Indicates the first Each node at time... The rate of temperature change divides the temperature rise process into transient intervals. and steady-state range From the perspective of thermal mechanism, the time decoupling of the heat capacity-dominated stage and the thermal resistance-dominated stage is realized, which establishes a physical basis for bi-objective Pareto optimization.
[0054] Step S22: Employing a bi-objective Pareto optimization approach, the dynamic response performance and steady-state thermal equilibrium performance are modeled separately. Let the vector of parameters to be identified be:
[0055] in, This represents the thermal resistance parameter. This represents the heat capacity parameter.
[0056] Step S23: For the transient phase, define a dynamic response objective function to constrain the heat capacity parameter, so that the model can accurately reflect the dynamic inertia and rate of change during the temperature rise process, expressed as:
[0057] in, This represents the total number of sampling points during the transient phase; This indicates that under given thermal network parameters In the case of the LPTN model, the first... The hot node is at the _ ... Predicted temperature at each moment; Indicates the first The node at the th The temperature actually collected by the sensor at a given moment.
[0058] For the steady-state phase, a thermal equilibrium objective function is defined to constrain the thermal resistance parameter, thereby improving the model's ability to fit the final thermal equilibrium temperature, expressed as:
[0059] in, This represents the total number of sampling points during the steady-state phase; Indicates the total length of the data.
[0060] The overall multi-objective optimization model can be expressed as:
[0061] in, This represents the vector of thermal network parameters to be identified, including thermal resistance and heat capacity parameters; The sum of squared errors between the model output temperature and the measured temperature during the transient phase is represented by the first optimization objective. The sum of squared errors between the model output temperature and the measured temperature during the steady-state phase represents the second optimization objective.
[0062] Step S3: Solve the bi-objective Pareto parameter identification model using a particle swarm optimization-whale cooperative search strategy to obtain the Pareto optimal solution set. Then, based on the knee point criterion, determine the optimal thermal network parameters from the Pareto optimal solution set that balance transient response performance and steady-state equilibrium performance. Specifically, this includes: Step S31: Use the particle swarm optimization algorithm to globally explore the parameter space of the thermal network, construct the Pareto front framework, and use an external archive to save all non-dominated solutions generated during the iteration process; represented as: Speed update formula:
[0063] Particle position update:
[0064] in, Inertial weights determine the historical motion trend of particles; and These represent individual learning factors and group learning factors, respectively. and Represents a random disturbance term; Indicates the best position in the particle's history; Indicates the globally optimal position; and They represent the first The particle in the first The position and velocity at the next iteration; in this embodiment, the maximum number of iterations of the particle swarm optimization algorithm is set to 50 generations. If the particle swarm optimization algorithm meets any of the following conditions within 8 consecutive generations: the number of non-dominated solutions does not increase, the Pareto front change rate is less than 0.5%, and the objective function improvement rate is less than 0.1%, then the particle swarm optimization algorithm will end early and enter the whale optimization algorithm.
[0065] Step S32: Use the whale optimization algorithm to perform a local fine-grained search on the non-dominated solutions in the external archive to improve the continuity of the Pareto front and the accuracy of the solutions, expressed as: When random number At this time, a shrinking encirclement mechanism is adopted:
[0066]
[0067] when At that time, a spiral update mechanism is adopted:
[0068]
[0069] in, This represents the solution selected from the non-dominated solution set; This indicates the deviation of the current solution from the Pareto skeleton point; Denotes the vector of random perturbation coefficients; This represents the contraction control vector. This represents the convergence factor that decreases linearly with iteration; This represents the true deviation of the current solution from the Pareto skeleton point; This indicates the spiral shape control parameter, which controls the spiral contraction speed; The variable represents a random spiral, which controls the position of the spiral trajectory. In this embodiment, the whale optimization algorithm terminates early if it has been executed for 30 generations or if it has met any of the following conditions for 5 consecutive generations: the rate of change of the frontier is less than 0.2%, the improvement rate of the knee point is less than 0.05%, and there are no new non-dominated solutions.
[0070] Step S33: Use the knee point criterion to screen the optimal thermal network parameters for the Pareto front.
[0071] Let the ideal solution be:
[0072] in, Indicates transient error: This represents the steady-state error.
[0073] Calculate the Euclidean distance between each Pareto optimal solution and the defined ideal point, and take the Pareto optimal solution with the smallest Euclidean distance as the optimal heat network parameter, expressed as:
[0074] in, Indicates the optimal thermal network parameters. This represents the Euclidean distance from the Pareto solution to the ideal point, achieving an optimal trade-off between transient response performance and steady-state temperature accuracy.
[0075] Step S4: Substitute the optimal thermal network parameters into the discrete-time state-space model, and predict and correct the state variables at the current moment based on the Kalman filter algorithm to recursively estimate the temperature state of key nodes of the motor online. Specifically: Step S41: Optimal heat network parameters Substituting into the state-space model of the thermal network, the continuous-time temperature state vector is... After discretization, it is denoted as a discrete-time state vector. Establish a discrete system:
[0076]
[0077] in, This represents the state vector indicating the temperature of each key component of the motor at the next moment. This represents the state vector indicating the temperature of each key component of the motor at the current moment. The key components of the motor include: stator teeth, stator windings, stator yoke, permanent magnets, and coolant. Represents the discrete state transition matrix. , Represents the identity matrix; Represents the heat loss input vector; Representing process noise, it is used to characterize the stochastic effects of factors such as unmodeled dynamics in the thermal network model, environmental disturbances, parameter uncertainties, and load fluctuations on the thermal state evolution process; Indicates measurement noise. This represents the temperature measurement output vector. This represents the observation matrix.
[0078] Based on the prediction results of the thermal network model on historical operating data, the prediction residual between the model-predicted temperature and the actual measured temperature is calculated:
[0079] Based on the noise covariance matrix of the statistical process of the predicted residual samples:
[0080] in, This represents the process noise covariance matrix.
[0081] Step S42: Based on the discrete state-space model, Kalman filtering is used to realize the recursive estimation of the system state. The prediction stage is represented as follows:
[0082]
[0083] in, Indicates that in the known Given all information at any given time, for Prior prediction of the thermal state at any given time; Represents the prior error covariance matrix; Let represent the posterior error covariance matrix.
[0084] The Kalman gain is calculated and expressed as:
[0085] in, This indicates the measurement standard deviation obtained from the static calibration experiment of the temperature sensor. The constructed measurement noise covariance matrix.
[0086] The final state update is calculated as follows:
[0087] This update process enables dynamic fusion between physical model predictions and measurement information.
[0088] Step S43: The final estimated critical hot node temperature is directly extracted from the state vector, and expressed as:
[0089] in, Indicates the stator winding temperature; This represents the temperature component of the corresponding hot node in the stator winding in the state vector.
[0090] To further verify the effectiveness of the present invention, Figure 2The Pareto front distribution obtained from solving the bi-objective Pareto optimization model is presented. As shown in the figure, the solution set of the front is uniformly distributed and exhibits good convergence, indicating that the particle swarm-whale cooperative search strategy can effectively explore the parameter space. Furthermore, the parameters corresponding to the knee point in the figure successfully achieve the optimal trade-off between transient response error and steady-state thermal equilibrium error, proving the applicability of the knee criterion in this multi-objective problem.
[0091] Figure 3 The dynamic tracking curves of the proposed method and different thermal state estimation methods at the stator winding temperature nodes were compared. The results show that the proposed method, by combining global search and local refinement, and establishing optimization objectives in both transient and steady-state stages, effectively alleviates the coupling conflict between heat capacity and thermal resistance parameters. Compared to the comparative methods, LPTN can reflect the basic thermal change law, but is significantly affected by model parameter errors; LPTN-KF improves estimation accuracy through measurement correction, but is still limited by model accuracy; LPTN-WOA-Pareto-KF improves the accuracy of the thermal network model by using dual-objective parameter identification, but still suffers from local optima due to the limited global search capability of a single optimization algorithm; the proposed method employs a PSO-WOA collaborative optimization strategy, which further enhances local optimization performance while ensuring global search capability, and combines Kalman filtering to correct model errors online, enabling the predicted curve to closely match the actual temperature change trend across all time scales. This results in not only faster transient response but also higher steady-state thermal balance accuracy, verifying the strong robustness and high reliability of the proposed method under complex operating conditions.
[0092] As can be seen from the above, the present invention can effectively alleviate the physical conflict between heat capacity and thermal resistance parameters, significantly suppress the interference of unmodeled dynamics and measurement noise, so that the motor thermal state estimation has both fast response capability, steady-state balance accuracy and online operation robustness, and realizes accurate recursive tracking of key node temperature.
[0093] Example 2: Based on the same inventive concept, embodiments of the present invention also provide a motor thermal state estimation system based on dual-target parameter identification and Kalman filtering, the system comprising: The model building module is used to acquire motor operation data, construct a five-node low-order thermal equivalent network model and the corresponding continuous-time state-space model, and discretize the model to obtain a discrete-time state-space model. The phase division and optimization modeling module is used to divide the motor temperature rise process into transient and steady-state stages based on a preset temperature change rate threshold. Based on the discrete-time state-space model, the response deviation of the transient stage and the thermal balance deviation of the steady-state stage are used as dual-objective optimization criteria to construct a dual-objective Pareto parameter identification model. The collaborative optimization solution module is used to solve the bi-objective Pareto parameter identification model using a particle swarm-whale collaborative optimization search strategy, obtain the Pareto optimal solution set, and determine the optimal thermal network parameters that take into account both transient response performance and steady-state balance performance from the Pareto optimal solution set according to the knee point criterion. State estimation module: It is used to substitute the optimal thermal network parameters into the discrete-time state-space model, and predict and correct the state variables at the current moment based on the Kalman filter algorithm, so as to recursively estimate the temperature state of the key nodes of the motor online.
[0094] Since these systems and the principles underlying the problems they solve are similar to the aforementioned method for estimating the thermal state of a motor based on dual-objective parameter identification and Kalman filtering, the implementation of this system can be found in the implementation of the aforementioned method, and the repetitions will not be repeated.
[0095] Example 3: Based on the same inventive concept, the present invention also provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the program stored in the memory, it is able to implement the motor thermal state estimation method based on dual-target parameter identification and Kalman filtering as described in any one of Embodiments 1.
[0096] like Figure 4 As shown, the electronic device may include: a processor 10, a communication interface 20, a memory 30, and a communication bus 40, wherein the processor 10, the communication interface 20, and the memory 30 communicate with each other via the communication bus 40. The processor 10 can call logic instructions in the memory 30 to execute a motor thermal state estimation method based on dual-target parameter identification and Kalman filtering. This method includes: S1. Obtain motor operation data, construct a five-node low-order thermal equivalent network model and the corresponding continuous-time state-space model, and discretize the continuous-time state-space model to obtain a discrete-time state-space model. S2. Based on the preset temperature change rate threshold, the motor temperature rise process is divided into transient stage and steady-state stage; based on the discrete-time state-space model, the response deviation of the transient stage and the thermal balance deviation of the steady-state stage are used as dual-objective optimization criteria to construct a dual-objective Pareto parameter identification model. S3. The particle swarm-whale collaborative optimization search strategy is used to solve the bi-objective Pareto parameter identification model to obtain the Pareto optimal solution set, and the optimal thermal network parameters that take into account both transient response performance and steady-state balance performance are determined from the Pareto optimal solution set according to the knee point criterion. S4. Substitute the optimal thermal network parameters into the discrete-time state-space model, and predict and correct the state variables at the current moment based on the Kalman filter algorithm to recursively estimate the temperature state of the key nodes of the motor online.
[0097] Example 4: This invention also provides a computer-readable storage medium, in which a program is stored for executing the motor thermal state estimation method based on dual-target parameter identification and Kalman filtering described in Embodiment 1 above. This program can be executed on a processor.
[0098] Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0099] The program stored on this medium is loaded into the processor's memory and executed to perform various functions. This storage medium, connected to hardware devices, enables the computer to perform the steps of Embodiment 1 described above.
[0100] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.
[0101] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for estimating the thermal state of a motor based on dual-objective parameter identification and Kalman filtering, characterized in that, Includes the following steps: S1. Obtain motor operation data, construct a five-node low-order thermal equivalent network model and the corresponding continuous-time state-space model, and discretize the continuous-time state-space model to obtain a discrete-time state-space model. S2. Based on the preset temperature change rate threshold, the motor temperature rise process is divided into transient stage and steady-state stage; based on the discrete-time state-space model, the response deviation of the transient stage and the thermal balance deviation of the steady-state stage are used as dual-objective optimization criteria to construct a dual-objective Pareto parameter identification model. S3. The particle swarm-whale collaborative optimization search strategy is used to solve the bi-objective Pareto parameter identification model to obtain the Pareto optimal solution set, and the optimal thermal network parameters that take into account both transient response performance and steady-state balance performance are determined from the Pareto optimal solution set according to the knee point criterion. S4. Substitute the optimal thermal network parameters into the discrete-time state-space model, and predict and correct the state variables at the current moment based on the Kalman filter algorithm to recursively estimate the temperature state of the key nodes of the motor online.
2. The method as described in claim 1, characterized in that, In step S1, constructing a five-node low-order thermal equivalent network model and its corresponding continuous-time state-space model specifically includes: For a five-node low-order thermal network model, a continuous-time state-space model is established based on the thermal network topology: in, This indicates the rate of temperature change per unit time at each thermal node; Represents the system state vector. Indicates the temperature of the stator teeth. Indicates the stator winding temperature. The stator yoke temperature, Indicates the temperature of the permanent magnet. Indicates the coolant temperature. Represents the continuous-time state transition matrix. Represents the hot input matrix; This represents the input vector of node heat loss.
3. The method as described in claim 1, characterized in that, Step S2 specifically includes: The temperature change rate of each thermal node is calculated based on historical temperature data, and the motor temperature rise process is divided into transient and steady-state stages according to a preset temperature change rate threshold. When the temperature change rate of each thermal node is lower than the set threshold, the temperature rise process continues. If the motor thermal process is in a steady state, it is determined to be in a transient state; otherwise, it is determined to be in a transient state. Temperature error objective functions are constructed for both the transient and steady-state phases. Using the dynamic response deviation in the transient phase and the thermal balance deviation in the steady-state phase as dual-objective optimization criteria, a dual-objective Pareto optimization model for thermal network parameter identification is established, with the following expression: in, This represents the vector of thermal network parameters to be identified, including thermal resistance and heat capacity parameters; The sum of squared errors between the model output temperature and the measured temperature during the transient phase is represented by the first optimization objective. The sum of squared errors between the model output temperature and the measured temperature during the steady-state phase represents the second optimization objective.
4. The method as described in claim 3, characterized in that, The sum of squared errors between the transient phase model output temperature and the measured temperature, and the sum of squared errors between the steady-state phase model output temperature and the measured temperature, are specifically defined as follows: in, This represents the total number of sampling points during the transient phase. This indicates that under given thermal network parameters In the case of the LPTN model, the first... The hot node is at the _ ... Predicted temperature at each moment; Indicates the first The node at the th The temperature actually collected by the sensor at a given moment; This represents the total number of sampling points during the steady-state phase; Indicates the total length of the data.
5. The method as described in claim 1, characterized in that, In step S3, a particle swarm-whale collaborative optimization search strategy is used to solve the bi-objective Pareto parameter identification model to obtain the Pareto optimal solution set, specifically including: The particle swarm optimization algorithm is used to globally explore the parameter space of the thermal network, construct the Pareto front skeleton, and use an external archive to save all non-dominated solutions generated during the iteration process. Based on the non-dominated solutions in the external archive, the whale optimization algorithm is used to perform a local fine search of the neighborhood of the non-dominated solutions, update the candidate parameter positions, and update the external archive according to the non-dominated relationship.
6. The method as described in claim 1, characterized in that, In step S3, the optimal thermal network parameters that balance transient response performance and steady-state equilibrium performance are determined from the Pareto optimal solution set according to the knee point criterion. Specifically, this includes: The ideal solution is taken as the point where both transient and steady-state errors are zero. Calculate the Euclidean distance from each solution in the Pareto optimal solution set to the ideal solution; Calculate the Euclidean distance between each Pareto optimal solution and the defined ideal point, and take the Pareto optimal solution with the smallest Euclidean distance as the optimal heat network parameter, expressed as: in, Indicates the optimal thermal network parameters. This represents the Euclidean distance from the Pareto solution to the ideal point.
7. The method as described in claim 1, characterized in that, Step S4 specifically includes: Substitute the optimal thermal network parameters into the discrete state-space model to establish the discrete system state equations and observation equations. Perform prior state prediction and prior error covariance prediction; Calculate the Kalman gain and use the temperature measurement values to perform posterior correction on the prior state prediction values; The stator winding temperature is extracted from the updated state vector as the temperature estimate of the key thermodynamic node.
8. The method as described in claim 7, characterized in that, The state equations and observation equations of the discrete system are specifically expressed as follows: in, This represents the state vector indicating the temperature of each key component of the motor at the next moment. This represents the state vector indicating the temperature of each key component of the motor at the current moment. The key components of the motor include: stator teeth, stator windings, stator yoke, permanent magnets, and coolant. Represents the discrete state transition matrix. , Represents the identity matrix. Represents the heat loss input vector. Indicates process noise. Indicates measurement noise. This represents the temperature measurement output vector. This represents the observation matrix.
9. A motor thermal state estimation system based on dual-target parameter identification and Kalman filtering, characterized in that, The system comprises: The model building module is used to acquire motor operation data, construct a five-node low-order thermal equivalent network model and the corresponding continuous-time state-space model, and discretize the model to obtain a discrete-time state-space model. The phase division and optimization modeling module is used to divide the motor temperature rise process into transient and steady-state stages based on a preset temperature change rate threshold. Based on the discrete-time state-space model, the response deviation of the transient stage and the thermal balance deviation of the steady-state stage are used as dual-objective optimization criteria to construct a dual-objective Pareto parameter identification model. The collaborative optimization solution module is used to solve the bi-objective Pareto parameter identification model using a particle swarm-whale collaborative optimization search strategy, obtain the Pareto optimal solution set, and determine the optimal thermal network parameters that take into account both transient response performance and steady-state balance performance from the Pareto optimal solution set according to the knee point criterion. The state estimation module is used to substitute the optimal thermal network parameters into the discrete-time state-space model, and predict and correct the state variables at the current moment based on the Kalman filter algorithm, so as to recursively estimate the temperature state of the key nodes of the motor online.
10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the motor thermal state estimation method based on dual-target parameter identification and Kalman filtering as described in any one of claims 1 to 8.