EKF-based online parameter identification method for igbt thermal network of photovoltaic inverter

Abstract

The application discloses a kind of EKF-based photovoltaic inverter IGBT thermal network online parameter identification method, comprising the following steps: S1, using third-order Cauer network, the heat transfer path of IGBT module establishes IGBT thermal network discrete state space model;S2, based on IGBT thermal network discrete state space model, design EKF algorithm for thermal parameter joint estimation for the synchronous estimation of state and parameter;S3, using current sensor and voltage sensor real-time acquisition IGBT current waveform, voltage waveform, and preprocessing;S4, in the embedded microcontroller of photovoltaic inverter, with fixed sampling period, EKF algorithm is operated, time update and measurement update are sequentially executed in each cycle, and the optimal estimation value of current time updated thermal parameter and junction temperature estimation value are output.The application can track the actual thermal characteristics of IGBT in real time, solve the parameter mismatch problem, and realize dynamic online identification of thermal parameters.

Description

Technical Field

[0001] This invention relates to the field of power electronic equipment condition monitoring and parameter identification technology, and in particular to an online parameter identification method for the IGBT thermal network of a photovoltaic inverter based on EKF. Background Technology

[0002] As a key power conversion device connecting photovoltaic arrays and the power grid, the long-term reliability of photovoltaic inverters directly affects the efficiency and stability of the entire photovoltaic power generation system. Insulated-gate bipolar transistors (IGBTs), as the core power switching devices in inverters, are continuously subjected to electrothermal stress during operation. Junction temperature fluctuations and thermal fatigue are the main causes of device performance degradation and even failure. Therefore, accurately understanding the thermal behavior of IGBTs and establishing accurate thermal characteristic models are prerequisites for achieving real-time status monitoring, lifetime prediction, and proactive thermal management. Thermal network models such as the Cauer or Foster models describe the dynamic relationship between junction temperature and power loss through parameters such as thermal resistance and thermal capacity. They are effective tools for analyzing device thermal characteristics, and the accuracy of their parameters directly determines the reliability of junction temperature estimation and thermal design.

[0003] Currently, obtaining IGBT thermal network parameters in engineering mainly relies on data from manufacturer-provided datasheets or offline identification methods based on standard tests (such as transient thermal testing). These methods are typically conducted under specific, stable laboratory conditions, yielding a set of fixed parameters. However, the actual operating environment of photovoltaic inverters is complex and variable, with continuous fluctuations in factors such as solar irradiance, ambient temperature, and load power, leading to constant changes in IGBT losses and heat dissipation conditions. Furthermore, during long-term operation, the thermal characteristic parameters of the devices slowly drift due to factors such as material aging, solder fatigue, and degradation of heat dissipation interface performance. Traditional offline fixed-parameter models cannot reflect these dynamic changes and slow drifts, resulting in significant "model mismatch" problems in actual operation. Junction temperature estimation based on mismatched models will produce large errors, significantly reducing the effectiveness of functions such as junction temperature-based lifetime prediction, health status assessment, and over-temperature protection, potentially leading to unnecessary maintenance or failure to provide timely fault warnings.

[0004] To address the problem of time-varying parameters, online parameter identification technology has emerged as a promising solution. The Extended Kalman Filter (EKF) algorithm, a classic joint state and parameter estimation algorithm for nonlinear systems, possesses good real-time recursion and a certain degree of noise suppression capability, and has been successfully applied in fields such as battery state estimation and motor parameter identification. Its core advantage lies in its ability to utilize real-time input and output measurement data of the system to update the internal state and parameters of the model online, enabling the model to continuously track the real dynamics of the system. However, applying EKF to the online identification of IGBT thermal network parameters in photovoltaic inverters still faces specific challenges: it requires establishing a discrete state-space model of the thermal network suitable for the EKF framework; it requires reasonably handling the influence of uncertainties such as loss calculation errors and temperature measurement noise; and it requires implementing a real-time algorithm with low computational complexity in embedded systems to adapt to the operating environment of the inverter controller. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide an online parameter identification method for the IGBT thermal network of a photovoltaic inverter based on EKF, so as to realize dynamic online identification of thermal parameters, improve the accuracy of junction temperature estimation and system reliability.

[0006] Technical Solution: An online parameter identification method for IGBT thermal network of photovoltaic inverters based on EKF, wherein the identification system includes a power conversion unit, a sensing and measurement unit, and a data processing and identification unit; the power conversion unit includes a photovoltaic inverter and its internal IGBT modules; the sensing and measurement unit includes current sensors, voltage sensors, and temperature sensors; the data processing and identification unit includes signal conditioning circuits, analog-to-digital converters, and embedded microprocessors; the system collects the operating electrical parameters, case temperature, and ambient temperature of the IGBT modules in real time through sensors; the instantaneous power loss of the IGBT modules is calculated based on the electrical parameters, and the instantaneous power loss and temperature measurement values ​​are input together to the embedded processor with a built-in EKF algorithm; the EKF algorithm estimates the thermal network parameters online and recursively based on the established IGBT heat dissipation network model, and updates the junction temperature estimate in real time; the steps include the following:

[0007] S1. A third-order Cauer network is used to establish a discrete state-space model of the IGBT thermal network for the heat transfer path of the IGBT module. S2. Based on the discrete state-space model of IGBT thermal network, an EKF algorithm for joint estimation of thermal parameters is designed for the synchronous estimation of state and parameters. S3 uses current and voltage sensors to acquire the current and voltage waveforms of the IGBT in real time and performs preprocessing. S4 runs the EKF algorithm in the embedded microcontroller of the photovoltaic inverter with a fixed sampling period. In each period, it sequentially performs time update and measurement update, and finally outputs the optimal estimate of thermal parameters and junction temperature at the current time.

[0008] Furthermore, in step S1, the third-order Cauer network includes three series-connected RC units, which respectively characterize the heat transfer processes from the junction to the shell, from the shell to the heat sink, and from the heat sink to the environment; the continuous-time thermal network model is discretized, and thermal resistance and thermal capacity parameters are selected as state variables to be identified, which together with the node temperature form an extended state vector to establish a discrete state space model of the IGBT thermal network with power loss as input and measurable temperature as output.

[0009] Furthermore, the implementation of the EKF algorithm includes: deriving the state equation and measurement equation of the photovoltaic inverter IGBT thermal network, calculating the Jacobian matrix required for model linearization, reasonably initializing the state vector and its error covariance matrix, and setting the covariance of process noise and measurement noise according to sensor accuracy and model uncertainty. At each recursive time The EKF algorithm is executed according to the following steps: Step A1, Time Update; Calculate state prediction: , in, for Prediction of prior states at time t. for Posterior state estimation at time 10:00. It is a nonlinear state transition function. for Time-based control input; Prediction error covariance matrix: , in, for The prior error covariance matrix at time t. Let Jacobian matrix be the state transition function. for The posterior error covariance matrix at time t. The process noise covariance matrix; Step A2, measurement update; Calculate the Kalman gain: , Calculate the measurement residuals: , Updated state estimate: , Update error covariance: , in, Here is the Kalman gain matrix. for The prior error covariance matrix at time t. For the observation matrix, To observe the noise covariance matrix, To observe the residuals, These are actual observed values. To predict the observed values, This is the actual observed value of the outer casing temperature. For predicted observations of the outer shell temperature, For posterior state estimation, For prior state prediction, For the posterior error covariance, For the prior error covariance, It is an identity matrix.

[0010] Furthermore, in step S3, the instantaneous power loss sequence is calculated based on the current waveform and voltage waveform of the IGBT module. At the same time, the housing temperature is measured by an NTC thermistor mounted on the IGBT module substrate, and the temperature digital sequence is obtained after signal conditioning and analog-to-digital conversion. The instantaneous power loss sequence and the temperature digital sequence are strictly synchronized in time, and the original data is filtered and denoised.

[0011] Furthermore, the time update uses the thermal parameter estimate from the previous moment and the current power loss to predict the current thermal state; the measurement update uses the measured temperature value at the current moment to correct the predicted value.

[0012] Compared with the prior art, the significant advantages of this invention are as follows: 1. This invention uses the EKF framework to perform online joint estimation of thermal resistance and thermal capacity as state variables, enabling the thermal model to track the actual thermal characteristics of IGBTs in real time, fundamentally solving the parameter mismatch problem and realizing dynamic online identification of thermal parameters; the junction temperature estimate obtained thereby has high dynamic accuracy, providing a reliable basis for key functions such as over-temperature protection and life prediction. 2. This invention employs the optimal recursive estimation mechanism of the extended Kalman filter algorithm, which effectively integrates real-time measurement information and suppresses sensor noise and model uncertainty. It exhibits strong robustness and anti-interference capabilities, and can adapt to complex operating environments. The identification system designed in this invention maintains stable convergence even under actual operating conditions such as sudden power changes in photovoltaic inverters and fluctuations in ambient temperature, ensuring the continuity and reliability of parameter estimation results and meeting the needs of long-term online monitoring in industrial settings.

[0013] 3. The method of this invention can be directly embedded into existing inverter control systems without the need for expensive additional hardware, resulting in low cost. The real-time and accurate thermal status information provided can directly support precise lifespan prediction models, guide predictive maintenance, and optimize maintenance cycles and spare parts management. Simultaneously, it can provide decision input for proactive thermal management strategies, thereby improving system reliability, extending equipment lifespan, and reducing total lifespan maintenance costs, demonstrating significant economic and engineering application value. Attached Figure Description

[0014] Figure 1 The overall flowchart of the online parameter identification method for IGBT thermal network of photovoltaic inverter based on EKF provided by the present invention; Figure 2 This is a schematic diagram of the heat transfer model of the IGBT module in an example of the present invention; Figure 3 This is the equivalent third-order Cauer thermal network model of the IGBT module in the example of this invention; Figure 4 This is a flowchart of the extended Kalman filter algorithm provided by the present invention; Figure 5 This is a schematic diagram of the photovoltaic inverter system structure in an example of the present invention; Figure 6 This is a schematic diagram of the temperature acquisition principle of the built-in NTC resistor in the IGBT module in this invention example; Figure 7 This is a schematic diagram of the simulation platform for the IGBT thermal network online parameter identification system in an example of the present invention; Figure 8 This is a schematic diagram of simulated data for a typical day of a photovoltaic inverter in an example of the present invention, wherein (a) is a solar irradiance variation curve, (b) is an ambient temperature variation curve, and (c) is a schematic diagram of the daily power loss of the photovoltaic inverter. Figure 9 This is a schematic diagram illustrating the output results of the actual and estimated thermal resistance values ​​based on EKF online parameter identification and fixed parameter values ​​in an example of the present invention. In this diagram, (a) represents the thermal resistance. Actual value vs. estimated value, (b) is thermal resistance Actual value and estimated value, (c) is thermal resistance The true value and the estimated value, (d) are the true value and the estimated value of total thermal resistance, and (e) are the comparison results of aging test errors based on EKF online parameter identification and fixed parameter values; Figure 10 This is a schematic diagram illustrating the output results of the actual and estimated temperature values ​​based on EKF online parameter identification and fixed parameter values ​​in an example of the present invention, wherein (a) is the junction temperature. Actual value and estimated value, (b) is the shell temperature Actual value and estimated value, (c) is the radiator temperature Actual value and estimated value, (d) is the junction temperature The estimation error, (e) is the shell temperature. The estimation error, (f) is the radiator temperature. The estimation error; In the picture, For IGBT chip power loss, For chip junction temperature, For the outer casing temperature, For radiator temperature, For ambient temperature, The thermal resistance between the chip and the casing. The thermal resistance between the casing and the heat sink, Thermal resistance between the radiator and the environment; For the heat capacity between the chip and the casing, For the heat capacity between the casing and the heat sink, This refers to the heat capacity between the radiator and the environment. This refers to the output current on the inverter side. This is the capacitor voltage. For grid-side output current, For reference active power, For reference reactive power, To control the output frequency for drooping. This is for droop control of the output voltage. Detailed Implementation

[0015] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0016] An online parameter identification method for IGBT thermal networks in photovoltaic inverters based on EKF is disclosed. The identification system includes a power conversion unit, a sensing and measurement unit, and a data processing and identification unit. The power conversion unit includes the photovoltaic inverter and its internal IGBT modules; the sensing and measurement unit includes current sensors, voltage sensors, and temperature sensors; and the data processing and identification unit includes signal conditioning circuitry, an analog-to-digital converter, and an embedded microprocessor.

[0017] The working principle of an online parameter identification method for IGBT thermal networks in photovoltaic inverters based on EKF is as follows: The operating electrical parameters of the IGBT (such as collector current and collector-emitter voltage), as well as the case temperature and ambient temperature, are collected in real time by sensors; the instantaneous power loss of the IGBT is calculated based on the electrical parameters, and the instantaneous power loss and temperature measurement values ​​are input together to an embedded processor with a built-in extended Kalman filter algorithm; the EKF algorithm estimates the thermal network parameters (thermal resistance and thermal capacity) online and recursively based on the established IGBT heat dissipation network model, and updates the junction temperature estimate in real time; the updated thermal parameters can be used for condition monitoring, lifetime prediction, and thermal management strategy optimization.

[0018] like Figure 1 As shown, an online parameter identification method for the IGBT thermal network of a photovoltaic inverter based on EKF includes the following steps: Step 1: Establish a discrete state-space model of the IGBT thermal network; A third-order Cauer network is used to physically model the heat transfer path of the IGBT module. This network consists of three series-connected RC units, representing the heat transfer processes from the junction to the shell, from the shell to the heat sink, and from the heat sink to the environment, respectively. The continuous-time thermal network model is discretized, and thermal resistance and thermal capacity parameters are selected as state variables to be identified. Together with the node temperature, they form an extended state vector, thereby establishing a discrete state-space model of the IGBT thermal network with power loss as input and measurable temperature as output.

[0019] A schematic diagram of the heat transfer model of an IGBT module is shown below. Figure 2 As shown, a third-order Cauer network is used to physically model the heat transfer path of the IGBT module. The third-order Cauer network is as follows: Figure 3 As shown, the network consists of three series-connected resistor-capacitor (RC) units, where: First-stage resistor-capacitor unit ( This corresponds to the heat transfer from the chip junction to the casing; Second-stage resistor-capacitor unit ( This corresponds to the heat transfer from the casing to the radiator; Third-stage resistor-capacitor unit ( This corresponds to the heat transfer from the radiator to the environment.

[0020] Based on thermal network theory, a continuous-time state equation is established: (1) (2) (3) in, For the heat capacity between the chip and the casing, For the heat capacity between the casing and the heat sink, This refers to the heat capacity between the radiator and the environment. For IGBT chip power loss, For chip junction temperature, For the outer casing temperature, For radiator temperature, For ambient temperature, The thermal resistance between the chip and the casing. The thermal resistance between the casing and the heat sink, This refers to the thermal resistance between the heat sink and the environment.

[0021] Use the forward Euler method to fix the sampling period. Discretization yields a discrete state-space model: (4) (5) (6) in, for Constant chip junction temperature, for Constant chip junction temperature, for Constantly monitor the outer casing temperature. for Constantly monitor the outer casing temperature. for Constantly monitor radiator temperature. for Constantly monitor radiator temperature. for ambient temperature at all times for The thermal capacity between the chip and the casing at any given time. for The heat capacity between the casing and the heat sink is constantly monitored. for The heat capacity between the heat sink and the environment is constantly monitored. for Power loss of IGBT chip at all times for The thermal resistance between the chip and the casing at any given moment. for The thermal resistance between the casing and the heatsink is constantly monitored. for The thermal resistance between the heat sink and the environment at all times.

[0022] Thermal resistance and heat capacity parameters are selected as the state variables to be identified, which, together with the nodal temperature, form a nine-dimensional extended state vector. The expression is as follows: (7) Where T represents the matrix transpose.

[0023] Thus, a system based on power loss is established. For input, with measurable casing temperature The output discrete state-space model of the IGBT thermal network has the following general form: (8) (9) in, It is a nonlinear vector function defined by a discrete state-space model, assuming that the thermal parameters remain constant between adjacent sampling points, i.e. This serves as its state equation. Output matrix , used to select from the state vector As an observable quantity. Process noise. and measurement noise Modeled as zero-mean Gaussian white noise, with covariance matrices as follows: and .

[0024] Step 2: Design an extended Kalman filter for joint estimation of thermal parameters; Based on the discrete state-space model of the IGBT thermal network established in step one, the EKF algorithm (Extended Kalman Filter) is designed for the synchronous estimation of state and parameters. This includes deriving the state equation and measurement equation of the photovoltaic inverter IGBT thermal network, calculating the Jacobian matrix required for model linearization, reasonably initializing the state vector and its error covariance matrix, and setting the covariance of process noise and measurement noise according to the sensor accuracy and model uncertainty.

[0025] The flowchart of the Extended Kalman Filter (EKF) algorithm is as follows: Figure 4 As shown, the algorithm is first initialized: the initial estimate of the extended state vector is set. The initial temperature setting is set to the ambient temperature at startup. .in, This refers to the chip junction temperature at startup. The casing temperature at startup. This refers to the radiator temperature at startup. The initial thermal parameters are given at the ambient temperature during startup, referencing typical values ​​from the IGBT datasheet. Initial estimation error covariance matrix. Let this be a diagonal matrix, where the diagonal elements reflect the uncertainty of the initial estimates for each state. A smaller variance can be assigned to the temperature state, and a larger variance to the thermal parameters. Process noise covariance matrix. Measurement noise covariance matrix The process noise covariance matrix needs to be set appropriately based on the system characteristics. This characterizes the uncertainties caused by model simplification, unmodeled dynamics, and slow time-varying parameters. Since the thermal parameters are assumed to be slowly varying, the corresponding process noise variance should be set very small, while the process noise variance corresponding to the temperature state can be set based on model error estimation. Measurement noise covariance matrix. It is a scalar quantity that directly reflects the measurement noise variance of the temperature sensor, according to the accuracy specifications in the NTC (Negative Temperature Coefficient) sensor datasheet. Calculation, assuming the noise follows a Gaussian distribution, then .

[0026] At each recursive time The EKF algorithm is executed according to the following steps: Step A1, Time Update; Calculate state prediction: (10) in, for Prediction of prior states at time t. for Posterior state estimation at time 10:00. It is a nonlinear state transition function. for Time-based control input.

[0027] Calculate the state transition Jacobian matrix : (11) The Jacobian matrix is ​​a A matrix whose elements are nonlinear functions. The partial derivatives with respect to each state variable are in The value at that location. Regarding and The partial derivative is: (12) (13) Other elements can be derived similarly.

[0028] Prediction error covariance matrix: (14) in, for The prior error covariance matrix at time t. Let Jacobian matrix be the state transition function. for The posterior error covariance matrix at time t. Let be the process noise covariance matrix.

[0029] Step A2, measurement update; Since the measurement equation is linear, the measurement Jacobian matrix... It is a constant.

[0030] Calculate the Kalman gain: (15) Calculate the measurement residuals: (16) Updated state estimate: (17) Update error covariance: (18) in, Here is the Kalman gain matrix. for The prior error covariance matrix at time t. For the observation matrix, To observe the noise covariance matrix, To observe the residuals, These are actual observed values. To predict the observed values, This is the actual observed value of the outer casing temperature. For predicted observations of the outer shell temperature, For posterior state estimation, For prior state prediction, For the posterior error covariance, For the prior error covariance, It is an identity matrix.

[0031] Step 3: Collect and preprocess operational data in real time. Current and voltage waveforms of the IGBT module are acquired in real time using current and voltage sensors to calculate the instantaneous power loss sequence. Simultaneously, the casing temperature is measured using an NTC thermistor mounted on the IGBT module substrate, and the temperature digital sequence is obtained after signal conditioning and analog-to-digital conversion. Strict time synchronization between the instantaneous power loss sequence and the temperature digital sequence is ensured, and the raw data undergoes preprocessing such as filtering and noise reduction to eliminate high-frequency interference and measurement noise.

[0032] C1) Power loss measurement and calculation; The current and voltage waveforms of the IGBT module are acquired in real time using current and voltage sensors. The voltage and current are expressed as follows: (19) (20) (twenty one) in, This is the effective value of the actual voltage. This is the effective value of the actual current. The phase angle is the independent variable. The phase angle between the actual voltage and the actual current. In order to adjust the system, This represents the duty cycle.

[0033] The power loss of an IGBT module is calculated using voltage and current. The power loss of an IGBT module includes conduction loss and switching loss.

[0034] As the temperature changes, the on-state voltage drop becomes: (twenty two) in, This represents the actual on-state voltage drop of the IGBT module. For IGBT modules in Rated on-state voltage drop at that time This refers to the junction temperature of the IGBT module. For IGBT modules in The rated resistance at that time, This represents the temperature coefficient of resistance and on-state voltage drop of the IGBT module in relation to temperature.

[0035] Therefore, conduction loss for: (twenty three) Switching losses for: (twenty four) in, These are the cutoff loss and conduction loss of the IGBT module under rated conditions, respectively. For carrier frequency, These are the reference voltage and the reference current, respectively. For bridge arm voltage, These are the voltage coefficients representing the effects of bridge arm voltage and current amplitude on IGBT losses, respectively. This represents the temperature coefficient of the effect of temperature on IGBT losses.

[0036] C2) Temperature measurement; The housing temperature is measured using an NTC thermistor mounted on the IGBT module substrate. Figure 6This is a schematic diagram for temperature acquisition. A 12V power supply is used, which is divided by a 100-ohm resistor. The voltage division value is then acquired by an ADC (Analog-to-Digital Converter) to calculate the NTC resistance. The calculation formula is as follows: (25) in, The voltage divider value of the NTC resistor acquired by the ADC. This is the power supply voltage. It is a 100-ohm resistor. This refers to the resistance value of the NTC resistor.

[0037] Then, based on the relationship between NTC resistance and temperature, we can conclude that: (26) in, , , , , , .

[0038] Finally, after signal conditioning and analog-to-digital conversion, the temperature digital sequence is obtained.

[0039] C3) Data synchronization and preprocessing; Ensure strict time synchronization between power and temperature data by using the same timer to simultaneously initiate voltage, current, and temperature sampling. Noise in the raw data can affect the performance of the EKF; therefore, preprocessing the raw data, such as filtering and denoising, is performed to eliminate high-frequency interference and measurement noise.

[0040] Step 4: Execute the EKF algorithm online recursively in the embedded system; The EKF algorithm designed in step two is implemented in the embedded microcontroller of the photovoltaic inverter. The EKF algorithm runs with a fixed sampling period, performing time update and measurement update sequentially in each period. The time update uses the thermal parameter estimate from the previous moment and the current power loss to predict the current thermal state; the measurement update uses the measured temperature value at the current moment to correct the predicted value, and finally outputs the optimal estimate of the thermal parameters and the junction temperature estimate at the current moment. This process runs continuously online to achieve dynamic tracking of thermal parameters.

[0041] A computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the circuit-level simulation method of the present invention. The processor contains a kernel that retrieves corresponding program units from memory.

[0042] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed, it implements the steps of the online parameter identification method of the present invention.

[0043] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0044] A computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of the online parameter identification method of the present invention.

[0045] The modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular hardware and software combination.

[0046] Embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may be implemented as a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] The structure diagram of the photovoltaic inverter system is as follows: Figure 5 As shown, a complete simulation platform for online parameter identification of IGBT thermal networks was built using the co-simulation capabilities of PLECS Blockset and MATLAB / Simulink. The specific implementation process is as follows: First, a simulation model of the photovoltaic inverter's main circuit is built in PLECS Blockset. The IGBT module uses a refined power device model, which not only provides electrical characteristic simulation but also outputs internally calculated instantaneous power loss data as input for subsequent thermal network and EKF algorithms. The photovoltaic inverter control part is then built in MATLAB / Simulink. The complete EKF algorithm designed in step two is implemented in code and embedded into the simulation using the MATLAB Function module. The complete IGBT thermal network online parameter identification system simulation platform is shown below. Figure 7 As shown.

[0048] Then, parameter configuration and initialization were performed, and the simulation time was set to 12 hours to simulate the power curve of a typical day for a photovoltaic inverter. Considering the complex and variable actual working environment of photovoltaic inverters, variations in solar irradiance and cloud shading effects were simulated, along with changes in ambient temperature and load, to generate daily power loss data that closely approximates the actual operating conditions of the photovoltaic inverter. Figure 8 As shown. The inverter control frequency is 15kHz. Due to the slow rate of temperature change, the EKF algorithm uses a fixed step size. The simulation is run with parameter identification triggered every 15 control cycles, ensuring both accuracy and computational efficiency. Considering the slow aging effect and the fact that thermal network parameters will not change significantly in a short period, to verify the effectiveness of the EKF algorithm, the initial aging rate of the IGBT modules was set to 5%, and a faster aging rate was implemented in the latter half of the simulation by adjusting the thermal resistance.

[0049] Finally, the simulation is run and online recursive identification is performed. Within each cycle, time updates and measurement updates are executed sequentially. The time update uses the thermal parameter estimates from the previous moment and the current power loss to predict the current thermal state; the measurement update uses the measured temperature at the current moment to correct the predicted values, ultimately outputting the optimal estimates of the thermal parameters and the junction temperature at the current moment. This process runs continuously online, achieving dynamic tracking of the thermal parameters.

[0050] Step 5: Verification, comparative analysis, and application of the identification results; The time-varying thermal parameters obtained through online identification are substituted into the thermal network model to calculate the dynamic junction temperature estimate. The junction temperature estimate based on the dynamically identified parameters is then compared with the calculation results of a traditional thermal model using fixed parameters. Through comparative analysis, the superiority and estimation accuracy of this method in tracking device aging and changes in operating conditions are verified. Ultimately, the real-time and accurate thermal parameters and junction temperature information output can directly serve the inverter's precise thermal management, reliability assessment, and predictive maintenance system.

[0051] To verify the effectiveness of the method of this invention, the junction temperature estimation results obtained from online parameter identification are compared and analyzed with the calculation results of a traditional thermal model using fixed parameters. The simulation results are as follows: Figure 9 and Figure 10 As shown. Figure 9The online estimation results of thermal resistance parameters are presented. It can be seen that the method proposed in this invention can quickly and accurately track the changes in thermal resistance caused by IGBT aging, and its estimated value almost perfectly matches the actual value. In contrast, the traditional model using fixed parameters maintains a constant thermal resistance value throughout the simulation process, failing to reflect the time-varying characteristics of thermal resistance. In the early stages of aging, the aging state detection error of the traditional model due to parameter mismatch is 4.76%; as the aging process intensifies, this error further expands to 12.43%. This comparative result clearly demonstrates the significant limitations of the traditional fixed-parameter model in device aging scenarios, thus highlighting the necessity of this method.

[0052] Figure 10 The estimated junction temperature results are presented. Analysis shows that the temperature estimation curve of the method proposed in this invention is highly consistent with the true value, with an estimation error close to zero. In contrast, the traditional model using fixed parameters shows a significant deviation between the estimated result and the true value, and this deviation increases continuously with the aging of the device. Through the above comparative analysis, it is strongly demonstrated that the online parameter identification method proposed in this invention can significantly improve the accuracy of IGBT junction temperature estimation in photovoltaic inverters, providing an effective solution to the thermal management problems caused by parameter mismatch and device aging.

Claims

1. A method for online parameter identification of IGBT thermal network in photovoltaic inverters based on EKF, wherein, The identification system includes a power conversion unit, a sensing and measurement unit, and a data processing and identification unit. The power conversion unit includes a photovoltaic inverter and its internal IGBT modules. The sensing and measurement unit includes a current sensor, a voltage sensor, and a temperature sensor. The data processing and identification unit includes a signal conditioning circuit, an analog-to-digital converter, and an embedded microprocessor. The system collects the operating electrical parameters, case temperature, and ambient temperature of the IGBT modules in real time through the sensors. Based on the electrical parameters, it calculates the instantaneous power loss of the IGBT modules and inputs the instantaneous power loss and temperature measurement values ​​together into the embedded processor with a built-in EKF algorithm. The EKF algorithm estimates the thermal network parameters online and recursively based on the established IGBT heat dissipation network model and updates the junction temperature estimate in real time. The system is characterized by the following steps: S1. A third-order Cauer network is used to establish a discrete state-space model of the IGBT thermal network for the heat transfer path of the IGBT module. S2. Based on the discrete state-space model of IGBT thermal network, an EKF algorithm for joint estimation of thermal parameters is designed for the synchronous estimation of state and parameters. S3 uses current and voltage sensors to acquire the current and voltage waveforms of the IGBT in real time and performs preprocessing. S4 runs the EKF algorithm in the embedded microcontroller of the photovoltaic inverter with a fixed sampling period. In each period, it sequentially performs time update and measurement update, and finally outputs the optimal estimate of thermal parameters and junction temperature at the current time.

2. The online parameter identification method for IGBT thermal network of photovoltaic inverter based on EKF according to claim 1, characterized in that, In step S1, the third-order Cauer network includes three series-connected RC units, which respectively represent the heat transfer processes from the junction to the shell, from the shell to the heat sink, and from the heat sink to the environment. The continuous-time thermal network model is discretized, and thermal resistance and thermal capacity parameters are selected as state variables to be identified. Together with the node temperature, they form an extended state vector, and an IGBT thermal network discrete state space model is established with power loss as input and measurable temperature as output.

3. The online parameter identification method for IGBT thermal network of photovoltaic inverter based on EKF according to claim 1, characterized in that, The implementation of the EKF algorithm includes: deriving the state equation and measurement equation of the photovoltaic inverter IGBT thermal network, calculating the Jacobian matrix required for model linearization, reasonably initializing the state vector and its error covariance matrix, and setting the covariance of process noise and measurement noise according to sensor accuracy and model uncertainty. At each recursive time The EKF algorithm is executed according to the following steps: Step A1, Time Update; Calculate state prediction: in, for Prediction of prior states at time t. for Posterior state estimation at time 10:

00. It is a nonlinear state transition function. for Time-based control input; Prediction error covariance matrix: in, for The prior error covariance matrix at time t. Let Jacobian matrix be the state transition function. for The posterior error covariance matrix at time t. The process noise covariance matrix; Step A2, measurement update; Calculate the Kalman gain: Calculate the measurement residuals: Updated state estimate: Update error covariance: in, Here is the Kalman gain matrix. for The prior error covariance matrix at time t. For the observation matrix, To observe the noise covariance matrix, To observe the residuals, These are actual observed values. To predict the observed values, This is the actual observed value of the outer casing temperature. For predicted observations of the outer shell temperature, For posterior state estimation, For prior state prediction, For the posterior error covariance, For the prior error covariance, It is an identity matrix.

4. The online parameter identification method for IGBT thermal network of photovoltaic inverter based on EKF according to claim 1, characterized in that, In step S3, the instantaneous power loss sequence is calculated based on the current waveform and voltage waveform of the IGBT module. At the same time, the housing temperature is measured by an NTC thermistor mounted on the IGBT module substrate, and the temperature digital sequence is obtained after signal conditioning and analog-to-digital conversion. The instantaneous power loss sequence and the temperature digital sequence are strictly synchronized in time, and the original data is filtered and denoised.

5. The online parameter identification method for IGBT thermal network of photovoltaic inverter based on EKF according to claim 1, characterized in that, The time update uses the thermal parameter estimate from the previous moment and the current power loss to predict the current thermal state; the measurement update uses the measured temperature value at the current moment to correct the predicted value.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 5.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed, it implements the steps of the method as described in any one of claims 1 to 5.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

Images