DAB converter circuit parameter extraction method based on physical information neural network

Abstract

The invention discloses a DAB converter circuit parameter extraction method based on a physical information neural network, and belongs to the technical field of circuit parameter extraction, and the method specifically comprises the following steps: constructing a time domain expression of a DAB converter; establishing a physical relation between the acquired electric quantity information data set and the to-be-extracted parameter data set; establishing a time domain recursive relation based on the physical relation established in the step S2; a physical information neural network fused by a data mechanism and physical information and used for parameter identification of the DAB converter is constructed, the physical information neural network updates circuit parameter data serving as network weight by predicting collected electric quantity information, and meanwhile accurate prediction of the electric quantity information and accurate identification of the parameter data are achieved. By the adoption of the method, a small amount of inductive current and output voltage data of the DAB converter are input, high-precision circuit parameter extraction is achieved under the scene that the collected data precision influenced by noise and different modulation modes are considered, robustness is high, and generalization is good.

Description

Technical Field

[0001] The invention relates to the technical field of circuit parameter extraction, in particular to a DAB converter circuit parameter extraction method based on physical information neural network. Background Art

[0002] Dual-active-bridge (DAB) converters are highly valued for their high power density, electrical isolation, and bidirectional power transmission capabilities, and are widely used in electrified transportation, distributed power generation, and aerospace. Furthermore, control optimization and status monitoring of DAB converters can greatly improve system efficiency, reduce hardware costs, and reduce the risk of failure. To achieve efficient control optimization and accurate status monitoring, accurate extraction of DAB converter parameters is particularly important.

[0003] Traditional converter parameter extraction methods are mainly based on physical information, which rely on additional sensors and complex mathematical calculations. When the system model order is too high, the extremely high model complexity will inevitably lead to a huge amount of calculation. In recent years, data-driven parameter extraction methods with the help of artificial intelligence (AI) tools have been widely used, but they require a large amount of data to construct the potential relationship between input (voltage, current signals) and output (parameter values ​​to be extracted). In data-sparse fields such as power electronics, their practicality is poor.

[0004] In the prior art, there is a method for extracting parameters of a Buck converter (step-down converter) based on a physical information neural network. The method forms a physical information neural network by sequentially connecting a data-driven network composed of a fully connected neural network and a physical information network implemented by the Runge-Kutta method, which can extract the main parameters of the Buck circuit. The disadvantage is that it is only applicable to scenarios with large inductor current ripple, which is contrary to the efficient operation of the converter. The topology and operating state of the Buck converter are relatively simple, and the technical difficulty is relatively low.

[0005] In the prior art, there is a method for extracting the DC side capacitance and AC side inductance of a three-phase inverter by using a physical information neural network. Disadvantages: Its hardware experiments show that the maximum error of the AC side inductance extraction of this method exceeds 14%, and the maximum error of the DC side capacitance extraction exceeds 5%, and the extraction accuracy is too low.

[0006] In the prior art, there is a method that uses the genetic algorithm backpropagation (GA-BP) algorithm to construct the mapping relationship between the terminal voltage and terminal current of the DAB converter and the converter parameters. Disadvantages: It is a purely data-driven method that requires a large amount of data for training to obtain an accurate relationship. It has the disadvantages of overfitting risk, large data requirements, poor generalization, and poor local deployment capabilities.

[0007] At present, there is no parameter extraction method based on physical information neural network that is suitable for complex DC-AC-AC-DC systems (such as DAB converters, etc.), and most of the current neural network-based parameter extraction methods are only applicable to specific converter types with specific control strategies or modulation methods. They have obvious generalization limitations and are difficult to apply to real scenarios. Summary of the invention

[0008] The purpose of the present invention is to provide a DAB converter circuit parameter extraction method based on physical information neural network to solve the above technical problems.

[0009] To achieve the above object, the present invention provides a DAB converter circuit parameter extraction method based on physical information neural network, and the specific steps are as follows: Step S1: constructing a time-domain differential expression of the DAB converter; Step S2: determining a data set of collected electric quantity information and a data set of parameters to be extracted, and establishing a physical connection between the data set of collected electric quantity information and the data set of parameters to be extracted; Step S3: establishing a time domain recursive relationship based on the physical connection established in step S2; Step S4: construct and train a physical information neural network for DAB converter parameter identification that integrates data mechanism and physical information. The physical information neural network updates the circuit parameter data that serves as the weight of the physical information neural network by predicting the collected power information, thereby realizing the prediction of power information and the identification of parameter data at the same time.

[0010] Preferably, when the product of the transformer primary voltage and the transformer secondary voltage is greater than zero and less than zero in the same cycle, the time domain differential expression of the DAB converter is established as follows: ; in, is the inductor current, and They are the primary equivalent voltage of the output capacitor voltage and the primary equivalent voltage of the output voltage respectively; , is the output capacitor voltage, is the primary-to-secondary turns ratio of the transformer; , is the output voltage, Indicates time; is the input voltage; is the inductance value; is the relative state variable of the original secondary switch, when , ;when , , and They are the transformer primary voltage and transformer secondary voltage respectively; , are the equivalent parameter values ​​of inductance and switch resistance, is the parasitic resistance of the inductor, is the on-resistance value of the switch; , is the equivalent parameter value of the capacitor parasitic resistance, is the parasitic resistance value of the capacitor; , is the equivalent parameter value of resistance, is the load resistance value; , is the capacitance equivalent parameter value, is the capacitance value.

[0011] Preferably, step S2 is specifically as follows: The elements in the data set for collecting power information include inductor current and output voltage; The parameter data set to be extracted is ; When the inductor current is used as the input variable, the differential linearization relationship is as follows: ; in, The inductor current is given by The determined nonlinear operator is expressed as follows: ; When the output voltage is used as the input variable, the differential linearization relationship is as follows: ; in, The equivalent output voltage is given by The determined nonlinear operator is expressed as follows: .

[0012] Preferably, step S3 is specifically as follows: Step S31: Assume that at the starting time and the end moment The inductor current states are and , the output voltage states are and , and satisfies At the same time, assuming that Existence within time period There are unobservable intermediate states, the intermediate states are: , , ,in is the time coefficient, , ,…, , the intermediate state point and is an implicit point; Step S32: Taking half a switching cycle as a unit and the inductor current waveform as a basis, the starting time The first turning point of the inductor current waveform is selected, and a random time within the threshold is set at the middle time of the second turning point and the third turning point. The point corresponding to this time is the quasi-midpoint, and the end time is Select the second turning point and the third turning point as the starting point, then the inductor current and equivalent output voltage starting time Corresponding to and The end points correspond to and , starting time and the end moment The corresponding points are explicit points; Step S33: To couple the implicit point and the explicit point, use The implicit Runge-Kutta method is constructed by Unique set of constant parameters , and are relational constants, is the time coefficient constant in step S31; The forward and backward recursive relationships between the inductor current and the equivalent output voltage are established respectively to couple the explicit point and the implicit point; The backward recursive relation expression is as follows: ; The forward recursive relation expression is as follows: .

[0013] Preferably, step S4 is specifically as follows: Step S41: The data driven network uses a fully connected neural network, and the input is the power information of the explicit point corresponding to the starting time , the power information of the explicit point corresponding to the starting time , variable S and the time interval between the starting time and the end time , bias and weight are The output is the power information of the implicit point at the intermediate moment predicted by the data-driven part and ; Step S42: Establish the physical information network part according to the backward recursive relationship expression and the forward recursive relationship expression, and input the power information of the implicit point at the intermediate moment predicted by the data-driven part and , bias and weight are , the power information of the explicit points corresponding to the start time and the end time are derived respectively through the forward recursive relationship and the backward recursive relationship , , and ; Step S43: Connecting the data-driven network and the physical information network in series to form a physical information neural network for DAB converter parameter identification by integrating data mechanism and physical information; Step S44: Through the simulation and experimental platform, the power data of the DAB converter is collected according to step S3 to form a data set of collected power information of the physical information neural network. In the collection stage of the simulation data, examples of adding analog-to-digital conversion noise and sampling noise and examples of using single phase shift modulation and double phase shift modulation are added; Step S45: Determine the loss function and training stop condition to train the physical information neural network for DAB converter parameter identification in step S43, and output the extracted parameter data when the stop condition is reached.

[0014] Preferably, in step S45, the loss function expression is as follows: ; in, is the loss function, and Starting time and the end moment The predicted value of the inductor current, and Starting time and the end moment Output voltage prediction value; The training stop conditions are as follows: When the training cycle reaches the first set number, or twice in a row ; Or the training cycles reach a second set number.

[0015] Therefore, the present invention adopts the above-mentioned DAB converter circuit parameter extraction method based on physical information neural network, which has the following beneficial effects: (1) A DAB converter physical information neural network is constructed by integrating data mechanism and physical information. By inputting a small amount of key data of the inductor current and output voltage of the DAB converter, high-precision circuit parameter extraction is achieved.

[0016] (2) When considering the accuracy of collected data affected by analog-to-digital conversion noise and sampling noise, the accuracy of the extracted results is less affected and has stronger robustness.

[0017] (3) Considering both single phase-shift modulation and double phase-shift modulation, the parameters can be accurately extracted, showing considerable generalization.

[0018] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The present invention is a flow chart of a method for extracting DAB converter circuit parameters based on a physical information neural network; Figure 2 It is an application diagram of an embodiment of the present invention; Figure 3 The circuit topology diagram of the DAB converter used in the present invention; Figure 4 The equivalent circuit diagram of the DAB converter used in the present invention; Figure 5 A schematic diagram of the explicit starting point, end point and implicit point selected for the present invention; Figure 6 A diagram of the physical information neural network structure constructed for the present invention; Figure 7 It is a convergence process diagram of the physical information neural network of the present invention; Figure 8 This is a comparison chart of the experimental, simulation and predicted results of the inductor current of the solution proposed in the present invention.

[0020] Fig. 9 This is a comparison chart of the experimental, simulation and predicted results of the output voltage of the solution proposed in the present invention. DETAILED DESCRIPTION

[0021] In the description of the present invention, it should be noted that the terms "upper", "lower", "inside", "outside", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, or the positions or positional relationships in which the invented product is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In the description of the present invention, it should also be noted that, unless otherwise clearly specified and limited, the terms "setting", "installation", and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0022] The embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0023] like Figure 1-Figure 2 As shown, a DAB converter circuit parameter extraction method based on physical information neural network, the specific steps are as follows: Step S1: Construct the time domain expression of the DAB converter.

[0024] When the product of the transformer primary voltage and the transformer secondary voltage is greater than zero and less than zero in the same cycle, for a general DAB converter, such as Figure 3 As shown, the equivalent Figure 4 As shown, the time domain differential expression can be established as follows: ; in, is the inductor current, and They are the primary equivalent voltage of the output capacitor voltage and the primary equivalent voltage of the output voltage respectively; , is the output capacitor voltage, is the primary-to-secondary turns ratio of the transformer; , is the output voltage, Indicates time. is the input voltage; is the inductance value; is the relative state variable of the original secondary switch, when , ;when , , and are the transformer primary voltage and transformer secondary voltage respectively. , are the equivalent parameter values ​​of inductance and switch resistance, is the parasitic resistance of the inductor, is the on-resistance of the switch. , is the equivalent parameter value of the capacitor parasitic resistance, is the parasitic resistance of the capacitor. , is the equivalent parameter value of resistance, is the load resistance value. , is the capacitance equivalent parameter value, is the capacitance value. In any specific modulation mode (double phase shift modulation mode and single phase shift modulation mode), as long as >0 and When both situations <0 occur simultaneously in one cycle, the time domain expression can be established.

[0025] Step S2: determining a collected electric quantity information data set and a to-be-extracted parameter data set, and establishing a physical connection between the collected electric quantity information data set and the to-be-extracted parameter data set.

[0026] The elements in the collected power information data set include inductor current and output voltage. The parameter data set to be extracted is .

[0027] When the inductor current is used as the input variable, the differential linearization relationship is as follows: ; in, The inductor current is given by The determined nonlinear operator is expressed as follows: ; When the output voltage is used as the input variable, the differential linearization relationship is as follows: ; in, The equivalent output voltage is given by The determined nonlinear operator is expressed as follows: .

[0028] Step S3: Establish a time domain recursive relationship based on the physical connection established in step S2.

[0029] Step S3 is as follows: Step S31: Assume that at the starting time and the end moment The inductor current states are and , the output voltage states are and , and satisfies At the same time, assuming that Existence within time period There are unobservable intermediate states, the intermediate states are: , , ,in is the time coefficient, , ,…, , the intermediate state point and is an implicit point, such as Figure 5 shown.

[0030] Step S32: Taking half a switching cycle as a unit and the inductor current waveform as a basis, the starting time Select the first turning point of the inductor current waveform, the middle moment of the second turning point and the third turning point to set a random time point within the threshold, and the point corresponding to the time is the quasi-midpoint, and the end time Select the second turning point and the third turning point as the starting point, then the inductor current and equivalent output voltage starting time Corresponding to and The end points correspond to and , starting time and the end moment The corresponding points are explicit points, such as Figure 5 shown.

[0031] Step S33: To couple the implicit point and the explicit point, use The implicit Runge-Kutta method is constructed by Unique set of constant parameters , and are relational constants, is the time coefficient constant in step S31. The Runge-Kutta method order of the physical information neural network It is not fixed and can be adjusted according to actual recognition requirements. In this scenario, the 10th-order Runge-Kutta method is selected.

[0032] The forward and backward recursive relationships between the inductor current and the equivalent output voltage are established respectively to couple the explicit point and the implicit point; The backward recursive relation expression is as follows: ;

[0033] The forward recursive relation expression is as follows: .

[0034] The forward and backward recursive relationships lay the foundation for the subsequent establishment of the physical information network part.

[0035] Step S4: construct and train a physical information neural network for DAB converter parameter identification that integrates data mechanism and physical information. The physical information neural network updates the circuit parameter data that serves as the weight of the physical information neural network by predicting the collected power information, thereby achieving accurate prediction of power information and precise identification of parameter data.

[0036] Step S4 is specifically as follows: Step S41: The data-driven network uses a data-driven fully connected neural network, and the input is the power information of the explicit point corresponding to the starting time , the power information of the explicit point corresponding to the starting time , variable S and the time interval between the starting time and the end time , bias and weight are The output is the power information of the implicit point at the intermediate moment predicted by the data-driven part and .

[0037] Step S42: Establish the physical information network part according to the backward recursive relationship expression and the forward recursive relationship expression, and input the power information of the implicit point at the intermediate moment predicted by the data-driven part and , bias and weight are , the power information of the explicit points corresponding to the start time and the end time are derived respectively through the forward recursive relationship and the backward recursive relationship , , and .

[0038] The structure of the data-driven neural network is not fixed, and the number of hidden layers and the number of neurons in each layer can be adjusted according to the actual recognition requirements. In this scenario, the selected data-driven network is a fully connected neural network with 4 hidden layers, each containing 50 neurons. Its activation function is the Leaky Relu function and the tanh function, and its bias and weight are .

[0039] Step S43: Connecting the data-driven network and the physical information network in series forms a physical information neural network for DAB converter parameter recognition by integrating data mechanism and physical information, such as Figure 6 shown.

[0040] Step S44: Through the simulation and experimental platform, the power data of the DAB converter is collected according to step S3 to form a data set of collected power information of the physical information neural network. In the collection stage of simulation data, examples of adding analog-to-digital conversion noise and sampling noise and examples of using single phase-shift modulation and double phase-shift modulation are added. In the experimental data collection stage, this embodiment uses single phase-shift modulation to control the converter.

[0041] Step S45: Determine the loss function and training stop condition to train the physical information neural network for DAB converter parameter identification in step S43, and output the extracted parameter data when the stop condition is reached.

[0042] In step S45, the loss function expression is as follows: ;

[0043] in, is the loss function, , the physical information neural network is constantly updated To reduce the loss function value, as well as It is also constantly updated, affecting the calculation process of the data-driven network and the physical information network, thereby promoting the convergence of the predicted value to the true value, such as Figure 7 As shown, and Starting time and the end moment The predicted value of the inductor current, and Starting time and the end moment Output voltage prediction value; The training stop conditions are as follows: When the training cycle reaches the first set number, which in this embodiment is 1000, or twice in a row ; Or the training cycle reaches a second set number, which is 20,000.

[0044] After reaching the above stop condition, the final The value is the final unknown parameter data set to be extracted. The comparison of experimental, simulation and prediction results is shown in the figure below. Figure 8-Figure 9 In all simulation examples, the average percentage error of the extracted parameters does not exceed 1.5%, and in the experiments conducted on the hardware-in-the-loop platform, the average percentage error of all extracted parameters does not exceed 6.8%, which verifies the effectiveness and accuracy of the method.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A DAB converter circuit parameter extraction method based on physical information neural network, characterized in that: The specific steps are as follows: Step S1: constructing a time-domain differential expression of the DAB converter; Step S2: determining a data set of collected electric quantity information and a data set of parameters to be extracted, and establishing a physical connection between the data set of collected electric quantity information and the data set of parameters to be extracted; Step S3: establishing a time domain recursive relationship based on the physical connection established in step S2; Step S4: construct and train a physical information neural network for DAB converter parameter identification that integrates data mechanism and physical information. The physical information neural network updates the circuit parameter data that serves as the weight of the physical information neural network by predicting the collected power information, thereby realizing the prediction of power information and the identification of parameter data at the same time.

2. The method for extracting DAB converter circuit parameters based on physical information neural network according to claim 1, characterized in that: When the product of the transformer primary voltage and the transformer secondary voltage is greater than zero and less than zero in the same cycle, the time domain differential expression of the DAB converter is established as follows: ； in, is the inductor current, and They are the primary equivalent voltage of the output capacitor voltage and the primary equivalent voltage of the output voltage respectively; , is the output capacitor voltage, is the primary-to-secondary turns ratio of the transformer; , is the output voltage, Indicates time; is the input voltage; is the inductance value; is the relative state variable of the original secondary switch, when , ;when , , and They are the transformer primary voltage and transformer secondary voltage respectively; , are the equivalent parameter values ​​of inductance and switch resistance, is the parasitic resistance of the inductor, is the on-resistance value of the switch; , is the equivalent parameter value of the capacitor parasitic resistance, is the parasitic resistance value of the capacitor; , is the equivalent parameter value of resistance, is the load resistance value; , is the capacitance equivalent parameter value, is the capacitance value.

3. The method for extracting DAB converter circuit parameters based on physical information neural network according to claim 2 is characterized in that: Step S2 is specifically as follows: The elements in the data set for collecting power information include inductor current and output voltage; The parameter data set to be extracted is ; When the inductor current is used as the input variable, the differential linearization relationship is as follows: ； in, The inductor current is given by The determined nonlinear operator is expressed as follows: ； When the output voltage is used as the input variable, the differential linearization relationship is as follows: ； in, The equivalent output voltage is given by The determined nonlinear operator is expressed as follows: 。 4. The method for extracting DAB converter circuit parameters based on physical information neural network according to claim 2 is characterized in that: Step S3 is as follows: Step S31: Assume that at the starting time and the end moment The inductor current states are and , the output voltage states are and , and satisfies At the same time, assuming that Existence within time period There are unobservable intermediate states, the intermediate states are: , , ,in is the time coefficient, , ,…, , the intermediate state point and is an implicit point; Step S32: Taking half a switching cycle as a unit and the inductor current waveform as a basis, the starting time The first turning point of the inductor current waveform is selected, and a random time within the threshold is set at the middle time of the second turning point and the third turning point. The point corresponding to this time is the quasi-midpoint, and the end time is Select the second turning point and the third turning point as the starting point, then the inductor current and equivalent output voltage starting time Corresponding to and The end points correspond to and , starting time and the end moment The corresponding points are explicit points; Step S33: To couple the implicit point and the explicit point, use The implicit Runge-Kutta method is constructed by Unique set of constant parameters , and are relational constants, is the time coefficient constant in step S31; The forward and backward recursive relationships between the inductor current and the equivalent output voltage are established respectively to couple the explicit point and the implicit point; The backward recursive relation expression is as follows: ； The forward recursive relation expression is as follows: 。 5. The method for extracting DAB converter circuit parameters based on physical information neural network according to claim 2, characterized in that: Step S4 is specifically as follows: Step S41: The data driven network uses a fully connected neural network, and the input is the power information of the explicit point corresponding to the starting time , the power information of the explicit point corresponding to the starting time , variable S and the time interval between the starting time and the end time , bias and weight are The output is the power information of the implicit point at the intermediate moment predicted by the data-driven part and ; Step S42: Establish a physical information network based on the backward recursive relationship expression and the forward recursive relationship expression, and input the power information of the implicit point at the intermediate moment predicted by the data-driven network. and , bias and weight are , the power information of the explicit points corresponding to the start time and the end time are derived respectively through the forward recursive relationship and the backward recursive relationship , , and ; Step S43: Connecting the data-driven network and the physical information network in series to form a physical information neural network for DAB converter parameter identification by integrating data mechanism and physical information; Step S44: Through the simulation and experimental platform, the power data of the DAB converter is collected according to step S3 to form a data set of collected power information of the physical information neural network. In the collection stage of the simulation data, examples of adding analog-to-digital conversion noise and sampling noise and examples of using single phase shift modulation and double phase shift modulation are added; Step S45: Determine the loss function and training stop condition to train the physical information neural network for DAB converter parameter identification in step S43, and output the extracted parameter data when the stop condition is reached.

6. The method for extracting DAB converter circuit parameters based on physical information neural network according to claim 5, characterized in that: In step S45, the loss function expression is as follows: ； in, is the loss function, and Starting time and the end moment The predicted value of the inductor current, and Starting time and the end moment Output voltage prediction value; The training stop conditions are as follows: When the training cycle reaches the first set number, or twice in a row ; Or the training cycles reach a second set number.

Images