A method for online parameter identification of thyristor converter valves based on BP neural network

Abstract

This invention discloses an online parameter identification method for thyristor converter valves based on a BP neural network. A simulation model of a six-pulse thyristor converter valve is built based on the design parameters of the converter valve in actual engineering. The aging process of the thyristor stage is simulated by changing the values ​​of the damping capacitance and the turn-off equivalent resistance in the simulation model. Data from various sampling moments within one cycle of the simulated aging process of the thyristor stage in the simulation model are selected to create an input-output sample dataset. The BP neural network model is used to train the sample dataset offline until the required identification accuracy is achieved. The neural network weights are then used as the initial values ​​for online learning, resulting in a trained BP neural network model. The trained BP neural network model is then used for parameter monitoring of the thyristor stage in an actual engineering converter valve. Measured voltage and current data are used as inputs to solve for the parameter identification results of the thyristor stage damping capacitance and turn-off equivalent resistance.

Description

Technical Field

[0001] This invention relates to the field of high voltage direct current transmission technology, specifically to a method for online parameter identification of thyristor converter valves based on a BP neural network. Background Technology

[0002] As a core component of traditional high-voltage direct current (HVDC) transmission, the thyristor converter valve is a crucial functional unit for AC-DC power conversion. Its long-term reliable operation is essential for the safe and stable operation of the HVDC transmission system. Considering that converter valves are frequently subjected to external environmental interference or prolonged operational stress, it is necessary to monitor their aging process and operating status in a timely manner. The thyristor-level resistance-capacitance (RC) parameters are one of the key indicators for evaluating their electrical performance. To enable real-time assessment of the thyristor-level aging condition, online identification of the thyristor-level RRC parameters in the converter valve is required.

[0003] Currently, existing thyristor converter valve monitoring systems in China can only measure parameters such as the impedance of the damping circuit and the static equalization circuit offline during planned annual maintenance, and test whether the trigger monitoring board is functioning properly. They cannot achieve real-time online monitoring of parameters such as the damping circuit and the static equalization circuit connected in parallel across the thyristor. Furthermore, the self-test function of the trigger monitoring board is not perfect, and it cannot detect and replace components in time during the early stages of converter valve aging, thus leading to more serious failure risks. Summary of the Invention

[0004] This invention provides an online parameter identification method for thyristor converter valves based on a BP neural network. The method achieves efficient parameter identification and status monitoring through a model, filling the gap in real-time monitoring of thyristor converter valves in existing engineering projects. Specifically, it is implemented through the following technical solution:

[0005] Step 1: Based on the design parameters of the converter valve in the actual project, build a simulation model of the six-pulse thyristor converter valve in the simulation software;

[0006] Step 2: Change the values ​​of the damping capacitor and turn-off equivalent resistance in the simulation model to simulate the aging process of the thyristor stage. The damping capacitor and turn-off equivalent resistance of each thyristor stage are the parameters to be identified.

[0007] Step 3: Select data from each sampling moment of the simulated aging process of the thyristor stage in the simulation model within one cycle, and create an input / output sample dataset for the BP neural network;

[0008] Step 4: Use the BP neural network model to train the sample dataset offline. After achieving the required recognition accuracy, use the neural network weights obtained from the offline training as the initial values ​​for the online learning of the neural network to obtain the trained BP neural network model.

[0009] Step 5: Use the trained BP neural network model to monitor the parameters of the thyristor stage of the converter valve in actual engineering. Use the measured voltage and current data as input to solve for the parameter identification results of the thyristor stage damping capacitance and turn-off equivalent resistance.

[0010] From a data-driven perspective, the complex internal structure of multi-stage thyristors in a converter valve can be ignored. By using existing historical aging data for modeling, the key electrical parameters of the thyristor stage can be quickly identified and monitored in real time, providing theoretical support and data support for subsequent fault diagnosis, operation and maintenance of the converter valve.

[0011] Based on the above plan, further steps include:

[0012] The design parameters of the converter valve in step one include the smoothing reactor, saturation reactor, commutation equivalent inductance, thyristor equivalent turn-on resistance, thyristor equivalent turn-off resistance, and damping circuit parameters.

[0013] Based on the above plan, further steps include:

[0014] In step two, assuming the number of thyristors connected in series in each single valve is n, then the electrical parameters to be identified are n damping capacitance values ​​and n equivalent turn-off resistance values, for a total of 2n parameters. In creating the sample dataset, to fully characterize various typical degradation states of the thyristor stage, the more datasets the better. The damping capacitors are aged at 5% intervals, and the equivalent turn-off resistances are aged at 10% intervals. The aging range for both the damping capacitors and the equivalent turn-off resistances is [0, 40%]. Therefore, the sample dataset contains 9... n ×5 n Group, corresponding to 9 n ×5 n Aging status of thyristor level parameters.

[0015] Based on the above plan, further steps include:

[0016] In step three, the input sample dataset consists of the total voltage and loop current of the single valve arm of the thyristor when the converter valve is working under each aging state; the output sample dataset consists of the damping capacitance value and turn-off equivalent resistance value of each thyristor stage under each aging state, i.e., the parameters to be identified.

[0017] Based on the above plan, further steps include:

[0018] In step three, the sampling step size is 2×10. -4 s, the input sample dataset has a total of 200 input values, and the sample input dataset is a 9 n ×5 n A matrix of 200 rows and 200 columns; the output dataset is a 9-row matrix. n×5 n A matrix of 2n rows and 2n columns.

[0019] Based on the above plan, further steps include:

[0020] In step four, the weights of the neural network are adjusted online using the gradient descent method, and the system is trained and adjusted online so that the output of the neural network is close to the actual value.

[0021] Based on the above plan, further steps include:

[0022] In step four, let the loss function be Loss:

[0023]

[0024] x i The parameters output by the neural network. The actual parameter values ​​in the sample data are used to characterize the identification error of the neural network model. The effect of weight adjustment is judged based on the value of Loss. The smaller the value of Loss, the better the identification effect.

[0025] Based on the above plan, further steps include:

[0026] In step four, the BP neural network is a three-layer nonlinear neural network, including an input layer, a hidden layer, and an output layer; wherein, the input layer contains 200 input neurons, the output layer contains 2n neurons, and the number of neurons in the hidden layer is determined by a combination of empirical formulas and actual experiments.

[0027] Based on the above plan, further steps include:

[0028] In step four, the number of neurons p in the hidden layer of the BP neural network takes the value range formed by p1, p2, and p3:

[0029]

[0030] p2 = 2N + 1

[0031]

[0032] In the formula, N is the number of neurons in the input layer; q is the number of neurons in the output layer.

[0033] Based on the above plan, further steps include:

[0034] In step five, the sampling step size of the measured voltage and current data should be consistent with the step size of the simulation sample dataset to keep the input layer structure of the neural network unchanged.

[0035] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0036] 1. This invention, from a data-driven perspective, can ignore the complex internal structure of multi-stage thyristors connected in series in the converter valve. It utilizes existing historical aging data for modeling, enabling rapid identification and real-time status monitoring of key electrical parameters at the thyristor level. This provides theoretical assurance and data support for subsequent fault diagnosis, operation, and maintenance of the converter valve.

[0037] 2. It eliminates the need for detailed modeling and precise solution of the six-pulse thyristor converter valve, greatly simplifying the calculation process and reducing the computational difficulty.

[0038] 3. Once the model is successfully trained, there is no need to retrain the network in the subsequent identification process. The existing network can be used directly for parameter identification or state monitoring. Even if hundreds of sample data need to be identified at the same time or multi-dimensional electrical parameters need to be identified, it only takes a few minutes.

[0039] 4. It can identify thyristor-level electrical parameters in real time without the need for additional sensors or other measuring devices. It is fast, low-cost, and easy to identify, filling the gap in real-time monitoring of thyristor converter valves in existing projects. It has great engineering value and research significance. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0041] Figure 1 This is a flowchart of the method in the embodiment;

[0042] Figure 2 The simulation model diagram of the six-pulse thyristor-based converter valve is shown in the embodiment.

[0043] Figure 3 This is a simulation waveform of the thyristor-stage single-valve voltage over one cycle in the example embodiment;

[0044] Figure 4 The simulation waveform of the thyristor-stage single valve current over one cycle is shown in the example.

[0045] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. The realization of the object of the present invention, its functional characteristics and advantages will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the present invention clearer, the accompanying drawings of the embodiments of the present invention will be used in conjunction with the following description. Figures 1 to 4 The technical solutions in the embodiments of the present invention are clearly and completely described herein. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0047] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other embodiments, well-known structures, circuits, materials, or methods have not been specifically described in order to avoid obscuring the invention.

[0048] Those skilled in the art will understand that all or part of the steps in the above-described facts and methods can be implemented by a program instructing related hardware. The program, or the program described herein, can be stored in a computer-readable storage medium. When executed, the program includes the following steps: [The text then describes the corresponding method steps.] The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0049] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0050] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied 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.

[0051] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0052] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0053] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0054] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0055] Example:

[0056] like Figure 1 As shown in this embodiment, a method for online parameter identification of a thyristor converter valve based on a BP neural network includes the following specific steps:

[0057] A. Based on the design parameters of the converter valve in actual engineering, including the smoothing reactor, saturation reactor, commutation equivalent inductance, thyristor equivalent turn-on and turn-off resistances, damping circuit parameters, etc., a simulation model of the six-pulse thyristor converter valve is built in simulation software, such as... Figure 2 As shown.

[0058] B. Creating a sample dataset: Assuming n thyristors are connected in series in each single valve, the electrical parameters to be identified are n damping capacitance values ​​and n equivalent turn-off resistance values, for a total of 2n parameters. When creating the sample dataset, the more datasets available, the better, to fully characterize various typical degradation states of the thyristor stage. Here, the sample dataset is created by aging the damping capacitor at 5% intervals and the equivalent turn-off resistance at 10% intervals. The aging range for both the damping capacitor and the equivalent turn-off resistance is [0, 40%]. Therefore, the final sample dataset contains 9... n ×5 n Group, corresponding to 9 n ×5 n Aging status of thyristor level parameters.

[0059] C. Create a sample input dataset, selecting the thyristor single-valve voltage and single-valve current at various sampling times within one period (0.02s) as the input to the subsequent neural network, such as... Figures 3-4 As shown, a sampling step size of 2×10 is selected here. -4 If s, then a total of 200 input values ​​are fed into the BP neural network. The final sample input dataset is a 9 n ×5 n A matrix of 200 rows and 200 columns; the output dataset is a 9-row matrix. n ×5 n A matrix of 2n rows and 2n columns.

[0060] D. Establish a BP neural network model and perform offline training on the collected thyristor current and voltage data. The neural network is a three-layer nonlinear neural network, including an input layer, a hidden layer, and an output layer. The number of neurons in each layer is set as follows: the input layer has 200 input neurons, the output layer has 2n neurons, and the number of neurons in the hidden layer is determined by a combination of empirical formulas and actual experiments. The range of values ​​is mainly calculated using the following formulas:

[0061]

[0062] p2 = 2N + 1

[0063]

[0064] In the formula, N is the number of neurons in the input layer; q is the number of neurons in the output layer; and the number of hidden layer neurons p takes the value range formed by p1, p2, and p3.

[0065] E. Use offline sample data to adjust the weights of the neural network through gradient descent with varying learning rate. When the loss function Loss converges to its minimum value, use the weights obtained from offline training as the initial weights for online monitoring of the neural network.

[0066]

[0067] F. Apply the trained 3-layer BP neural network model to the parameter monitoring of the thyristor stage of the converter valve in actual engineering, using measured voltage and current data as input. Note that the sampling step size of the measured data should be the same as the step size set during simulation (2×10). - 4 The input and output layers have the same variable dimensions, and the final solution yields the parameter identification results of the damping capacitors and turn-off equivalent resistances of each thyristor stage.

[0068] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for online parameter identification of a thyristor converter valve based on a BP neural network, characterized in that, Includes the following steps: Step 1: Based on the design parameters of the converter valve in the actual project, build a simulation model of the six-pulse thyristor converter valve in the simulation software; The design parameters of the converter valve include the smoothing reactor, saturation reactor, commutation equivalent inductance, thyristor equivalent turn-on resistance, thyristor equivalent turn-off resistance, and damping circuit parameters. Step 2: Change the values ​​of the damping capacitor and turn-off equivalent resistance in the simulation model to simulate the aging process of the thyristor stage. The damping capacitor and turn-off equivalent resistance of each thyristor stage are the parameters to be identified. Assuming n is the number of thyristors connected in series in each single valve, then the electrical parameters to be identified are n damping capacitance values ​​and n equivalent turn-off resistance values, for a total of 2n parameters. In the prepared sample dataset, the damping capacitance is aged at 5% intervals, and the equivalent turn-off resistance is aged at 10% intervals. The aging range for both the damping capacitance and the equivalent turn-off resistance is [0, 40%]. Therefore, the sample dataset contains 9... n ×5 n Group, corresponding to 9 n ×5 n Aging status of thyristor stage parameters; Step 3: Select data from each sampling moment of the simulated aging process of the thyristor stage in the simulation model within one cycle, and create an input / output sample dataset for the BP neural network; The input sample dataset consists of the total voltage and loop current of a single thyristor valve arm when the converter valve is operating under each aging condition; the output sample dataset consists of the damping capacitance and turn-off equivalent resistance values ​​of each thyristor stage under each aging condition, i.e., the parameters to be identified; the sampling step size is 2×10. -4 s, the input sample dataset has a total of 200 input values, and the sample input dataset is a 9 n ×5 n A matrix with 200 rows and 200 columns; The output dataset is a 9 n ×5 n A matrix of 2n rows and 2n columns; Step 4: Use the BP neural network model to train the sample dataset offline. After achieving the required recognition accuracy, use the neural network weights obtained from the offline training as the initial values ​​for the online learning of the neural network to obtain the trained BP neural network model. Step 5: Use the trained BP neural network model to monitor the parameters of the thyristor stage of the converter valve in actual engineering. Use the measured voltage and current data as input to solve for the parameter identification results of the thyristor stage damping capacitance and turn-off equivalent resistance.

2. The method for online parameter identification of a thyristor converter valve based on a BP neural network according to claim 1, characterized in that, In step four, the weights of the neural network are adjusted online using the gradient descent method, and the system is trained and adjusted online so that the output of the neural network is close to the actual value.

3. The method for online parameter identification of a thyristor converter valve based on a BP neural network according to claim 2, characterized in that, In step four, let the loss function be Loss: x i The parameters output by the neural network. The actual parameter values ​​in the sample data are used to characterize the identification error of the neural network model. The effect of weight adjustment is judged based on the value of Loss. The smaller the value of Loss, the better the identification effect.

4. The method for online parameter identification of a thyristor converter valve based on a BP neural network according to claim 1, characterized in that, In step four, the BP neural network is a three-layer nonlinear neural network, including an input layer, a hidden layer, and an output layer; wherein the input layer contains 200 input neurons and the output layer contains 2n neurons.

5. The method for online parameter identification of a thyristor converter valve based on a BP neural network according to claim 4, characterized in that, In step four, the number of neurons p in the hidden layer of the BP neural network takes the value range formed by p1, p2, and p3: p2 = 2N + 1 In the formula, N is the number of neurons in the input layer; q represents the number of output layer elements.

6. The method for online parameter identification of a thyristor converter valve based on a BP neural network according to claim 1, characterized in that, In step five, the sampling step size of the measured voltage and current data should be consistent with the step size of the simulation sample dataset to keep the input layer structure of the neural network unchanged.

Images