A method for monitoring the state of a resonant capacitor of a medium voltage direct current LLC resonant converter

Abstract

The method for monitoring the resonant capacitor state of the medium-voltage direct-current LLC resonant converter comprises the following steps: closing the switch tube S 11 ~S 41 of the full-bridge at the initial moment when the LLC resonant converter is stopped, then turning on the switch tube S 31 and S 41 after the closing process lasts for 500 ns, forming a discharge loop of a resonant cavity independent of the latter stage, then collecting the time sequence data of the resonant current in the resonant cavity, and estimating the resonant capacitor by using a nonlinear parameter identification algorithm, wherein the estimation is sequentially performed by encoding, initializing the population, calculating the individual fitness, selecting the cross variation evolution, and iterative optimization, finally searching for the optimal resonant current estimation value, and calculating the resonant capacitor estimation value based on the corresponding resonant frequency and resonant inductance. The present application estimates the resonant capacitor value based on the discharge curve of the limited sampling data, has low cost and high reliability, and the present application estimates the parameters by using the nonlinear parameter identification algorithm, and has high estimation accuracy.

Description

Technical Field

[0001] This invention relates to the field of DC-DC converter condition monitoring technology, and in particular to a method for monitoring the condition of the resonant capacitor in a medium-voltage DC LLC resonant converter. Background Technology

[0002] The LLC resonant converter is a reliable and efficient medium-voltage DC-DC converter structure that can be applied to medium-voltage DC applications. Compared with conventional resonators, the advantages of the LLC resonant converter can be summarized as follows: First, it has the ability to regulate a wide input voltage and a wide output load; second, it can achieve soft-switching technology, i.e., zero-voltage switching, within its operating range, thereby improving power efficiency; third, parasitic components in the circuit, including the junction capacitance of semiconductor devices, the leakage inductance and magnetizing inductance of the transformer, can be used to achieve zero-voltage switching.

[0003] Generally, a typical LLC resonant topology consists of four parts: a switching network, a resonant network, a transformer, and a rectifier-filter network. The resonant cavity of an LLC resonant converter is composed of a resonant capacitor and a resonant inductor. Energy transfer in the converter is primarily achieved through the resonance process of these two capacitors and inductors. Capacitors are among the components with the highest failure rates in modern power electronic systems, facing harsh operating environments such as high temperature and high humidity. The limited heat dissipation volume in high-power-density power electronic systems further exacerbates the reliability challenges. During resonance, the resonant capacitor inevitably ages due to frequent charging and discharging. In particular, when LLC resonant converters based on silicon carbide devices reduce losses by increasing the operating frequency, the risk of resonant capacitor aging and failure is further increased. Therefore, once the resonant capacitor fails due to aging, the efficient and reliable operation of the LLC resonant converter will be severely threatened. Improvement measures to address the aforementioned reliability issues can be broadly categorized into three types: First, improving the design process to enhance device reliability; second, based on the existing capacitors, monitoring circuit operating conditions, including topology, voltage, temperature, and ripple current, to achieve better robustness; and third, implementing status monitoring of operating capacitors to ensure reliable operation and as a basis for preventative maintenance. Considering feasibility and implementation value, accurate capacitor status monitoring is crucial for the reliable operation of the system.

[0004] Current research on capacitance monitoring primarily focuses on DC-link capacitors, with few methods for monitoring resonant capacitors. Traditional DC-link capacitance monitoring methods can be categorized based on their principles: those based on periodic small-signal ripple, those based on aperiodic large-signal charge-discharge curves, and those based on black-box models. Among these, artificial intelligence methods based on black-box models face limitations in practical applications due to difficulties in acquiring training datasets and uncertainties in actual operating conditions. Secondly, methods based on the principle of periodic small-signal ripple mainly include circuit modeling and signal injection. Establishing an accurate model is the foundation of circuit modeling; however, extracting voltage ripple signals is quite difficult to implement and involves significant interference, making it hard to guarantee accuracy. Furthermore, changes in operating conditions require remodeling, resulting in poor universality. Signal injection methods, to some extent, affect normal system operation, and the introduction of numerous digital filters increases the burden on the control system. In contrast, methods based on aperiodic large-signal charge-discharge curves do not require complex modeling or high-frequency sampling, have relatively simple implementation conditions, and offer good accuracy. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for monitoring the state of the resonant capacitor in a medium-voltage DC LLC resonant converter. This method uses a nonlinear parameter identification algorithm based on the principle of non-periodic large signal to realize the state monitoring of the resonant capacitor in the LLC resonant converter, which is simple and effective.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical method: a method for monitoring the state of the resonant capacitor of a medium-voltage DC LLC resonant converter, comprising:

[0007] Step S1: At the initial moment of shutdown of the LLC resonant converter, turn off the full-bridge switching transistor S. 11 ~S 41 ;

[0008] In step S2, the switch S of the full bridge is turned on after the off state in step S1 has been in the off state for 500 ns. 31 and S 41 This forms a resonant cavity discharge circuit independent of the subsequent stage;

[0009] Step S3: Collect the time-series data of the resonant current in the resonant cavity, and use the following algorithm based on nonlinear parameter identification to estimate the resonant capacitance;

[0010] Step S301, Encoding: Encode the identification parameters initial voltage, initial phase angle, resonant frequency, and attenuation coefficient into a four-element vector;

[0011] Step S302, Initialize the population: Define the upper and lower bounds of the variables according to the physical meaning of the initial voltage, initial phase angle, resonant frequency, and attenuation coefficient, and randomly generate the population;

[0012] Step S303, calculate fitness: Based on the initial voltage, initial phase angle, resonant frequency, and attenuation coefficient of the individual, and according to the RLC discharge estimation model shown in Equation (1), the estimated resonant current sequence is obtained. The difference between the estimated resonant current sequence and the collected resonant current time series data is used to solve the L2 norm, and the reciprocal is taken as the fitness.

[0013]

[0014] In the formula, I Lr I0 is the resonant current, and I0 is the amplitude of the current signal at the initial stage of discharge. ω is the initial phase angle, ω is the resonant frequency, and τ is the attenuation coefficient;

[0015] Step S304, Individual Evolution: First, based on the selection operator, some poor individuals are eliminated and some better individuals are selected by utilizing the fitness of each individual. Then, according to the selection probability, crossover points are randomly set in the individual encoding strings of the pairs and genes are exchanged to form new individuals. Finally, mutation points are randomly selected according to the mutation probability, and a random search is performed on the new individuals at the mutation points.

[0016] Step S305, Iterative optimization: Repeat steps S303 and S304 to search for the optimal resonant current estimate within the set number of iterations. Based on the corresponding resonant frequency and the known resonant inductance value, the resonant capacitance estimate is calculated using the following formula (2):

[0017]

[0018] Furthermore, it also includes step S4: correcting the estimated value of the resonant capacitance based on the influence of ambient temperature.

[0019] Preferably, in step S4, the formula for correcting the estimated value of the resonant capacitance is as follows:

[0020]

[0021] In the formula, α M β M γ M These are all temperature characteristic parameters of the capacitor. For a specific type and model of capacitor, these three temperature characteristic parameters can be measured experimentally; T a T a，min T a,max These represent the actual operating temperature, minimum operating temperature, and maximum operating temperature of the capacitor, respectively.

[0022] The resonant capacitor state monitoring method for medium-voltage DC LLC resonant converters provided by this invention mainly utilizes a nonlinear parameter identification algorithm based on the non-periodic large-signal principle to process the current data collected during the discharge of the resonant cavity of the LLC resonant converter, thereby realizing the state monitoring of the resonant capacitor in the LLC resonant converter. Compared with traditional capacitor state monitoring methods, this method, which estimates the resonant capacitor value based on the discharge curve of finite sampling data, is low-cost and highly reliable. Furthermore, this method uses a basic swarm intelligence algorithm—genetic algorithm—in the nonlinear parameter identification algorithm for parameter estimation, resulting in high estimation accuracy. In summary, the method involved in this invention is simple, reliable, and can effectively realize the state monitoring of the resonant capacitor in LLC resonant converters. Attached Figure Description

[0023] Figure 1 This is a circuit diagram of the medium-voltage DC LLC resonant converter involved in this invention;

[0024] Figure 2 The waveform of the resonant current during the RLC discharge process of this invention is shown.

[0025] Figure 3 This is a schematic diagram of the resonant current data points obtained by the present invention at different sampling frequencies. Detailed Implementation

[0026] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0027] The medium-voltage DC LLC resonant converter involved in this invention comprises four parts: a switching network, a resonant network, a transformer, and a rectifier-filter network. The specific structure is described in detail below. Figure 1 In the diagram, V in R L These represent the input voltage and load resistance of the LLC resonant converter, respectively. A switching transistor Q1 is connected in parallel on the input side of the LLC resonant converter, along with a DC-Link capacitor C. in1 With bleeder resistor R in1 The RC circuit is composed of power switching transistors S. The switching network of the LLC resonant converter includes the power switching transistor S. 11 ~S 41 and body diode D 11 ~D 41 The resonant network includes the resonant capacitor C. r1 Resonant inductor L r1 And excitation inductance L m1 The transformer is T1 (the primary to secondary turns ratio is m:1), and the rectifier filter network includes rectifier diodes D. 51 ~D 81The rectifier bridge and the filter output capacitor C connected in parallel with the rectifier bridge constitute the rectifier bridge. f1 .

[0028] Based on the above medium-voltage DC LLC resonant converter, this invention provides a method for monitoring the state of the resonant capacitor of a medium-voltage DC LLC resonant converter. The specific implementation process of this method is as follows.

[0029] Step S1: At the initial moment of shutdown of the LLC resonant converter, turn off the full-bridge switching transistor S. 11 ~S 41 Considering the specific implementation conditions, the operating mode of the LLC resonant converter at the initial moment of shutdown is uncertain. Since the turn-off process of the switching transistor lasts for about 200ns, from the perspective of safe operation, in order to avoid direct switching of operating modes, all full-bridge switching transistors are turned off as a transition phase. The duration of this phase is set to 500ns. At this time, the energy of the resonant cavity is stored in the resonant inductor and resonant capacitor.

[0030] Step S2: After the shutdown phase of step S1 ends, the switching transistor S of the full bridge of the LLC resonant converter is turned on. 31 and S 41 Resonant capacitor C in1 Resonant inductor L r1 And excitation inductance L m1 and switching transistor S 31 and S 41 The on-resistance forms a closed loop, generating an RLC resonance process. RLC discharge occurs in the resonant cavity, and its discharge waveform is a sine wave with decreasing amplitude, such as... Figure 3 As shown, its discharge estimation model is as follows (1):

[0031]

[0032] In the formula, I Lr I0 is the resonant current, and I0 is the amplitude of the current signal at the initial stage of discharge. ω is the initial phase angle, ω is the resonant frequency, and τ is the attenuation coefficient.

[0033] Step S3: Collect time-series data of the resonant current in the resonant cavity, and estimate the resonant capacitance using a nonlinear parameter identification algorithm.

[0034] This invention acquires timing data of the resonant current during the discharge process at a certain sampling frequency, such as... Figure 3As shown, the data points of the resonant current differ at different sampling frequencies. This invention identifies the capacitance parameters based on the collected data points. Traditional parameter identification algorithms include least squares, augmented least squares, gradient approximation, and maximum likelihood estimation. Their drawbacks include limitation to linear systems, single-point search, reliance on the initial point, and susceptibility to limitations. Nonlinear system identification algorithms include swarm intelligence algorithms and neural networks, which offer advantages such as strong multi-point search / nonlinear fitting capabilities, making them suitable for handling nonlinear systems. Therefore, this invention selects a nonlinear parameter identification algorithm to identify the capacitance parameters from the collected data points. The identified parameters include initial voltage, initial phase angle, resonant frequency, and attenuation coefficient. Considering the limited number of parameters and the well-defined domain of the parameters based on their practical physical meaning, this invention employs a basic swarm intelligence algorithm—genetic algorithm—to efficiently solve for the capacitance parameters, as detailed below.

[0035] Step S301, Encoding. The swarm intelligence algorithm operates on the actual decision variables of the optimization problem. The decision variables are represented as string-structured data through encoding. Here, the identification parameters—initial voltage, initial phase angle, resonant frequency, and attenuation coefficient—are represented as a four-element vector. The initial voltage is set to the amplitude of a sine wave under normal operating conditions, which is related to the operating conditions. The initial phase angle is set to 0. The resonant frequency is obtained based on the initially designed resonant inductance and capacitance values ​​of the resonant cavity. The attenuation coefficient is obtained based on the initially designed resonant inductance and loop resistance values.

[0036] Step S302: Initialize the population. Define the upper and lower bounds of the variables based on the physical meaning of the identification parameters initial voltage, initial phase angle, resonant frequency, and attenuation coefficient, and randomly generate the population.

[0037] Step S303: Calculate fitness. Based on the initial voltage, initial phase angle, resonant frequency, and attenuation coefficient of an individual, substitute them into the RLC discharge estimation model shown in Equation (1) to obtain the estimated resonant current sequence. Based on the time series of the collected data, calculate the L2 norm by subtracting the estimated resonant current sequence from the collected resonant current time series data, and take the reciprocal as the fitness. The more accurate the estimated resonant current parameters are, the greater their fitness will be.

[0038] Step S304, Individual evolution, including selection, crossover, and mutation.

[0039] Selection: This process uses the fitness values ​​of each individual to eliminate some poor individuals and select some better individuals, that is, to select the solutions with relatively more accurate parameter estimates for the next step of crossover and mutation operations.

[0040] Crossover: The crossover operator is implemented by using a single-point crossover method. That is, a crossover point is randomly set in the individual code strings of each pair according to the selection probability. Then, at this point, some genes of the two paired individuals are exchanged, that is, any variable in the two solutions - initial voltage, initial phase angle, resonant frequency, attenuation coefficient - is exchanged to form two new individuals.

[0041] Mutation: This operation randomly selects mutation points according to the mutation probability, and then randomly searches for new individuals at the mutation points.

[0042] Step S305, Iterative optimization: Repeat steps S303 and S304 to search for the optimal resonant current estimate within the set number of iterations. Based on the corresponding resonant frequency and the resonant inductance value determined during the power supply equipment design, the resonant capacitance estimate is calculated using the following formula (2):

[0043]

[0044] Step S4, correct the estimated value of the resonant capacitance C r Since the operating frequency of the resonant capacitor in the LLC resonant converter is determined by a fixed resonant frequency, the influence of switching frequency variation can be ignored during correction, and only the influence of ambient temperature needs to be considered. The estimated capacitance value C obtained in step S304 is calculated using the following formula (3). r Perform correction:

[0045]

[0046] In the formula, α M β M γ M These are all temperature characteristic parameters of the capacitor. For a specific type and model of capacitor, these three temperature characteristic parameters can be measured experimentally; T a T a，min T a,max These represent the actual operating temperature, minimum operating temperature, and maximum operating temperature of the capacitor, respectively.

[0047] The above embodiments are preferred implementations of the present invention. In addition, the present invention can be implemented in other ways. Any obvious substitutions without departing from the concept of the present technical solution are within the protection scope of the present invention.

[0048] To facilitate understanding by those skilled in the art of the improvements of this invention over the prior art, some of the accompanying drawings and descriptions have been simplified, and for clarity, some other elements have been omitted from this application. Those skilled in the art should realize that these omitted elements may also constitute the content of this invention.

Claims

1. A method for monitoring the state of the resonant capacitor in a medium-voltage DC LLC resonant converter, characterized in that, include: Step S1: At the initial moment of shutdown of the LLC resonant converter, turn off the full-bridge switching transistor S. 11 ~S 41 ; Step S2, after the shutdown process in step S1 lasts for 500 ns, turns on the full-bridge switching transistor S. 31 and S 41 This forms a resonant cavity discharge circuit independent of the subsequent stage; Step S3: Collect the time-series data of the resonant current in the resonant cavity, and use the following algorithm based on nonlinear parameter identification to estimate the resonant capacitance; Step S301, Encoding: Encode the identification parameters initial voltage, initial phase angle, resonant frequency, and attenuation coefficient into a four-element vector; Step S302, Initialize the population: Define the upper and lower bounds of the variables according to the physical meaning of the initial voltage, initial phase angle, resonant frequency, and attenuation coefficient, and randomly generate the population; Step S303, calculate fitness: Based on the initial voltage, initial phase angle, resonant frequency, and attenuation coefficient of the individual, and according to the RLC discharge estimation model shown in Equation (1), the estimated resonant current sequence is obtained. The difference between the estimated resonant current sequence and the collected resonant current time series data is used to solve the L2 norm, and the reciprocal is taken as the fitness. In the formula, I Lr I0 is the resonant current, and I0 is the amplitude of the current signal at the initial stage of discharge. ω is the initial phase angle, ω is the resonant frequency, and τ is the attenuation coefficient; Step S304, Individual Evolution: First, based on the selection operator, some poor individuals are eliminated and some better individuals are selected by utilizing the fitness of each individual. Then, according to the selection probability, crossover points are randomly set in the individual encoding strings of the pairs and genes are exchanged to form new individuals. Finally, mutation points are randomly selected according to the mutation probability, and a random search is performed on the new individuals at the mutation points. Step S305, Iterative optimization: Repeat steps S303 and S304 to search for the optimal resonant current estimate within the set number of iterations. Based on the corresponding resonant frequency and the known resonant inductance value, the resonant capacitance estimate is calculated using the following formula (2):

2. The method for monitoring the resonant capacitor state of a medium-voltage DC LLC resonant converter according to claim 1, characterized in that: It also includes step S4, which corrects the estimated value of the resonant capacitance based on the influence of ambient temperature.

3. The method for monitoring the resonant capacitor state of a medium-voltage DC LLC resonant converter according to claim 2, characterized in that: In step S4, the formula for correcting the estimated value of the resonant capacitance is as follows: In the formula, α M β M γ M These are all temperature characteristic parameters of the capacitor. For a specific type and model of capacitor, these three temperature characteristic parameters can be measured experimentally; T a T a，min T a,max These represent the actual operating temperature, minimum operating temperature, and maximum operating temperature of the capacitor, respectively.

Images