Method for detecting and judging commutation failure of commutation converter valve

Abstract

The invention provides a commutation converter valve commutation failure detection and determination method, and the method comprises the steps: building a fuzzy neural network model, and taking obtained historical data as a training data set to train the built fuzzy neural network model; and after the training is completed, detecting the bus voltage, and judging whether the commutation of the commutation converter valve fails or not according to the trained fuzzy neural network model and a detection result. The method can effectively reduce the damage of commutation failure to the power grid, and improves the friendliness of the DC power transmission system to the AC power grid.

Description

Technical Field

[0001] The present invention belongs to the technical field of power transmission, and in particular relates to a method for detecting and determining commutation failure of a commutation converter valve. Background Art

[0002] At present, more and more renewable energy power generation bases are being built in my country, and the problem of renewable energy power consumption needs to be solved urgently. High-voltage direct current transmission technology is an important technical means for the transmission of renewable energy. LCC (line commutated converter) uses thyristors and relies on grid voltage to achieve phase change. It is simple to implement and low-cost. Therefore, most of the converter valves in existing direct current transmission projects use LCC, but the problem of phase change failure of LCC has not been better solved so far.

[0003] HCC (Hybrid commutated converter) can effectively avoid the occurrence of commutation failure. LCC commutation failure is mostly caused by AC grid voltage fluctuation or drop. The essence is that the arc extinction angle is too small and the thyristor is not completely turned off, resulting in reverse phase and commutation failure. Commutation failure can cause problems such as reduced DC voltage, reduced transmission power, increased current, and DC bias of the converter transformer. When the system is multi-terminal DC feed, it may even cause chain commutation failure, seriously affecting the safe and stable operation of the receiving end grid. HCC is based on IGCT power devices and has active shutdown capability without relying on grid commutation. Therefore, it can effectively avoid commutation failure and reduce the risk of commutation failure.

[0004] In the actual operation of HCC, if commutation failure occurs, based on its active shutdown capability, if timely detection and judgment can be made and intervention control can be performed, the converter valve will quickly recover to the normal working point, reducing the impact of commutation failure on the power grid system. However, in the prior art, there is no technical solution for timely judging commutation failure.

[0005] Therefore, it is necessary to design a method for detecting and determining commutation failure of a commutation valve to solve the above technical problems. Summary of the invention

[0006] In view of the above technical problems, the present invention provides a method for detecting and determining commutation failure of a commutation valve, wherein the method comprises:

[0007] Establish a fuzzy neural network model, and use the acquired historical data as a training data set to train the established fuzzy neural network model;

[0008] After the training is completed, the bus voltage is detected, and based on the trained fuzzy neural network model and the detection result, it is determined whether the commutation valve fails to commutate.

[0009] Furthermore, the establishment of the fuzzy neural network model includes:

[0010] The fuzzy neural network model is established with the AC side voltage and the DC side voltage at both ends of the phase-changing converter valve as input variables.

[0011] Furthermore, the conditions satisfied for the training to be completed include:

[0012] When the output error of the fuzzy neural network model meets the error requirement, the training is considered qualified and the fuzzy neural network model completes the training.

[0013] Furthermore, determining whether the commutation valve fails in commutation includes:

[0014] Regularly collect the AC side voltage and DC side voltage at both ends of the phase-changing valve;

[0015] The collected voltage signal, namely the voltage variable, is sent to the fuzzy neural network model for judgment.

[0016] Furthermore, determining whether the commutation valve fails in commutation also includes:

[0017] Fuzzify the input voltage variable to make it meet the input requirements of the fuzzy neural network model;

[0018] Calculate and determine the credibility of the fuzzified input variables for the fuzzy rules;

[0019] Normalize and calculate the credibility of fuzzy rules;

[0020] According to the fuzzy rules and combined with the membership function, the output corresponding to each node is calculated;

[0021] The total output variables are summarized and calculated to obtain a determination signal of whether commutation failure occurs.

[0022] Furthermore, the structure of the fuzzy neural network model includes:

[0023] The fuzzification layer is used to fuzzify the input variables, namely, the AC side voltage and the DC side voltage at both ends of the commutation valve;

[0024] The strength release layer of the fuzzy rules is used to calculate the credibility of the fuzzy rules based on historical data and formulate corresponding fuzzy rules, wherein the fuzzy rules have orders of magnitude differences;

[0025] Normalization layer, used to calculate the normalized credibility of each fuzzy rule to ensure the equality of each fuzzy rule at different magnitudes;

[0026] The output layer of fuzzy rule calculation is used to determine the output corresponding to each input variable according to the fuzzy rules and the membership function;

[0027] The total output layer is used to calculate the total output of all input variables as a decision signal to determine whether commutation failure occurs.

[0028] Furthermore, the fuzzification process is determined by the following formula:

[0029]

[0030] Where, u1 and u2 are the AC side voltage and DC side voltage of the two ends of the commutation valve, respectively. i , N i is a fuzzy set, is the fuzzy set M i , N i The membership function of is the membership function of u1, is the membership function of u2, i is the number of nodes, and n is a positive integer.

[0031] Furthermore, the credibility of the fuzzy rule is determined by the following formula:

[0032]

[0033] In the formula, and w i is the credibility value of the fuzzy rule.

[0034] Furthermore, the normalized credibility of the fuzzy rule is calculated by the following formula:

[0035]

[0036] In the formula, is the output of the normalization layer, which is the normalized credibility value of the fuzzy rule

[0037] Furthermore, the output corresponding to each input variable is calculated by the following formula:

[0038]

[0039] In the formula, {p i ,q i ,r i} is the parameter set of the output layer nodes for calculating fuzzy rules, f i is the output function of node i, The output of the output layer for calculating fuzzy rules.

[0040] Furthermore, the total output is calculated by the following formula:

[0041]

[0042] In the formula, Indicates the total output.

[0043] Furthermore, the historical data is historical data of commutation failure.

[0044] On the other hand, the present invention also provides a phase change valve commutation failure detection and determination system, wherein the system comprises:

[0045] Establishing a module for establishing a fuzzy neural network model, and using the acquired historical data as a training data set to train the established fuzzy neural network model;

[0046] The determination module is used to detect the bus voltage after the training is completed, and to determine whether the commutation valve fails to commutate based on the trained fuzzy neural network model.

[0047] Furthermore, the establishment of the fuzzy neural network model includes:

[0048] The fuzzy neural network model is established with the AC side voltage and the DC side voltage at both ends of the phase-changing converter valve as input variables.

[0049] The present invention provides a method for detecting and determining commutation failure of a phase-changing converter valve, which can be applied to determine whether the commutation of an HCC converter valve has failed, and can effectively reduce the harm caused by the commutation failure to the power grid, and improve the friendliness of the DC transmission system to the AC power grid.

[0050] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A flow chart of a method for detecting and determining commutation failure of a commutation valve according to an embodiment of the present invention is shown;

[0053] Figure 2 A schematic diagram of the structure of a fuzzy neural network model according to an embodiment of the present invention is shown;

[0054] Figure 3The topology of a double-terminal direct current transmission system according to an embodiment of the present invention is shown;

[0055] Figure 4 A schematic structural diagram of a phase-changing valve commutation failure detection and determination system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] The HCC converter valve uses IGCT power devices as switching devices, which can operate at a small arc extinction angle and require rapid identification and recovery after a commutation failure occurs. According to the State Grid Corporation's regulations on locking converter valves due to UHV commutation failure, it is required to lock the converter valve when three consecutive commutation failures cause power fluctuations.

[0058] Taking into account that commutation failure is mostly caused by bus voltage drop, and in order to improve the accuracy and speed of HCC commutation failure detection and judgment, the present invention proposes a HCC converter valve commutation failure detection and judgment method based on fuzzy neural network self-learning algorithm (i.e., a commutation failure detection and judgment method for commutation converter valves), which is applied to the judgment of whether the HCC converter valve commutation has failed (i.e., it can also be called: a commutation failure detection and judgment method applied to HCC converter valves), which can effectively reduce the harm caused by commutation failure to the power grid and improve the friendliness of the DC transmission system to the AC power grid.

[0059] A description is given below (it should be noted that the phase-changing valve of the present invention, illustratively, refers to the HCC valve).

[0060] The present invention provides a method for detecting and determining commutation failure of a commutation valve, the method comprising:

[0061] like Figure 1 As shown, a fuzzy neural network algorithm model is established, and the historical data (historical commutation failure data) is used as a training data set to train the fuzzy neural network model to extract the bus voltage characteristics of the commutation failure and form a commutation failure characteristic memory.

[0062] Then determine whether the output of the fuzzy neural network model meets the error requirement. If not, continue with the previous step. If so, periodically detect the HCC AC and DC voltages and calculate them through the fuzzy logic model (i.e., the fuzzy neural network model below). After the calculation is completed, determine whether it meets the phase change failure characteristics. If not, continue with the previous step. If it meets, that is, when the voltage characteristics meet the phase change failure characteristics, it is determined that a phase change failure has occurred.

[0063] The above process can be summarized into the following steps S1 and S2:

[0064] Step S1: Establish a fuzzy neural network model, and use the acquired historical data as a training data set to train the established fuzzy neural network model;

[0065] Step S2: After the training is completed, the bus voltage is detected, and based on the trained fuzzy neural network model and the detection result, it is determined whether the commutation valve fails to commutate.

[0066] The above steps are described in detail below.

[0067] For step S1:

[0068] Step S11: First, a fuzzy neural network model needs to be established. The specific execution steps include:

[0069] The fuzzy neural network model is established by taking the AC side voltage and the DC side voltage at both ends of the phase-changing converter valve as input variables;

[0070] Step S12: After the fuzzy neural network model is established, the acquired (ie collected) historical data is used as a training data set to train the established fuzzy neural network model, wherein the historical data is historical data of LCC commutation failure.

[0071] The training is completed if the following conditions are met: the training is qualified when the output error of the fuzzy neural network model meets the error requirement, and the fuzzy neural network model completes the training. The fuzzy neural network model after training can be used as a judgment tool for commutation failure.

[0072] The structure of the fuzzy neural network model in the above step S1 has the following structure:

[0073] like Figure 2 The structure of the fuzzy neural network model includes:

[0074] Layer 1: Fuzzy layer

[0075] The fuzzification layer is used to fuzzify the input variables u1 and u2, that is, the AC side voltage and the DC side voltage at both ends of the phase-changing valve, so that they meet the input requirements of the fuzzy neural network model. The fuzzification process is determined by the following formula:

[0076]

[0077] Where, u1 and u2 are the AC side voltage and DC side voltage of the two ends of the commutation valve respectively, M i , N i is a fuzzy set, is the fuzzy set M i , N i The membership function of and is the membership function, and i is the number of nodes.

[0078] Layer 2: Fuzzy rules strength release layer

[0079] The strength release layer of the fuzzy rule is used to calculate the credibility of the fuzzy rule based on the historical data of commutation failure, and formulate the corresponding fuzzy rule. The fuzzy rule completely covers all voltage drop ranges, and the fuzzy rule has data magnitude differences. The output of each node in this layer represents the credibility of the fuzzy rule. The credibility of the fuzzy rule is determined by the following formula:

[0080]

[0081] In the formula, and w i is the credibility value of the fuzzy rule.

[0082] Layer 3: Normalization layer

[0083] Since the fuzzy rules covering all voltage drop ranges have different data magnitudes, in order to ensure the equality of each fuzzy rule at different magnitudes, the normalized credibility of each fuzzy rule can be calculated by using the normalization layer to ensure the equality of each fuzzy rule at different magnitudes; the normalized credibility of the fuzzy rule is calculated by the following formula:

[0084]

[0085] In the formula, is the output of the third layer (normalization layer) of the fuzzy neural network model, that is, the normalized credibility value of the fuzzy rule. is the normalized credibility of the fuzzy rule, and n is a positive integer.

[0086] Layer 4: Output layer for calculating fuzzy rules

[0087] The output layer of the fuzzy rule calculation is used to calculate the output corresponding to each input variable according to the fuzzy rule and the membership function. The output corresponding to each input variable is calculated by the following formula:

[0088]

[0089] In the formula, {p i ,q i ,r i} is the parameter set of the fourth layer, that is, the output layer node for calculating fuzzy rules, where p i ,q i ,r i All of them are used to calculate the coefficients mentioned above, which can be selected as needed, f i is the output function of node i, The output of the output layer for calculating fuzzy rules.

[0090] Layer 5: Total output layer, used to calculate the total output of all input variables (signals) as a determination signal for determining whether commutation failure occurs, wherein calculating the total output of all input variables (signals) specifically includes: summarizing the outputs corresponding to all input variables to obtain the total output. The total output is calculated by the following formula:

[0091]

[0092] In the formula, represents the total output, and Figure 2 In the figure (U represents the nodes of the second layer, V represents the nodes of the third layer), y represents the output variable corresponding to the total output.

[0093] For step S1, the whole process is to collect historical commutation failure data to form the training data set and evaluation data set of the fuzzy neural network model in the HCC commutation failure detection and judgment method. Initialize the fuzzy inference system FIS (fuzzy inference system, FIS), use the anfis function in MATLAB to train and evaluate the initialized fuzzy inference system FIS. When the fuzzy neural network output meets the error requirement, the fuzzy neural network model training is considered completed and can be used for actual HCC commutation failure detection and judgment.

[0094] For step S2:

[0095] After the above fuzzy neural network model is established, it is determined whether the commutation valve fails to commutate. The following is an example to illustrate the specific execution method:

[0096] A typical implementation case is Figure 3 As shown, Figure 3It is a double-terminal DC transmission system, which includes two HCCs (i.e. Figure 3 HCC1 and HCC2 in the figure), one end of HCC1 is connected to three-phase alternating current (A, B, C three-phase alternating current); one end of HCC2 is connected to three-phase alternating current (A, B, C three-phase alternating current).

[0097] Among them, HCC1 and HCC2 are connected through two lines, one of which includes a DCB (DC circuit breaker) connected to the other end of HCC1 and a DCB connected to the other end of HCC2, and in one of the lines, the DCB connected to HCC1 and the DCB connected to HCC2 are connected through Pos (positive line).

[0098] The other line also includes a DCB connected to the other end of HCC1 and a DCB connected to HCC2. In the other line, the DCB connected to the other end of HCC1 and the DCB connected to HCC2 are connected via Neg (negative line). The DC voltage level of the double-terminal DC transmission system is ±800kV. The switch device in the converter valve adopts an IGCT device with active shutdown capability, and the DCB adopts a hybrid DC circuit breaker.

[0099] Therefore, the specific steps of step S2 include:

[0100] Step S21: regularly collecting the AC side voltage and the DC side voltage at both ends of the phase-changing valve;

[0101] Step S22: sending the collected voltage signal, ie, the voltage variable, to the fuzzy neural network model for calculation and judgment.

[0102] Step S23: fuzzifying the input voltage variable to make it meet the input requirements of the fuzzy neural network model;

[0103] Step S24: Calculate and determine the credibility of the fuzzified input variables for the fuzzy rules;

[0104] Step S25: performing normalization calculation on the credibility of the fuzzy rule;

[0105] Step S26: Calculate the output corresponding to each node in the fourth layer according to the fuzzy rules and in combination with the membership function;

[0106] Step S27: Summarize and calculate the total output variable to obtain a determination signal of whether commutation failure occurs.

[0107] On the other hand, Figure 4 As shown, the present invention also provides a phase change valve phase change failure detection and determination system, wherein the system comprises:

[0108] Establishing a module for establishing a fuzzy neural network model, and using the acquired historical data as a training data set to train the established fuzzy neural network model;

[0109] The determination module is used to detect the bus voltage after the training is completed, and to determine whether the commutation valve fails to commutate based on the trained fuzzy neural network model.

[0110] The other functions and implementation methods of each module in the phase-changing valve switching failure detection and judgment system of the present invention correspond to the functions and implementation methods of each step of the phase-changing valve switching failure detection and judgment method mentioned above in the present invention, and therefore, they will not be repeated here.

[0111] The above is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for detecting and determining commutation failure of a commutation valve, wherein: The method comprises: Establish a fuzzy neural network model, and use the acquired historical data as a training data set to train the established fuzzy neural network model; After the training is completed, the bus voltage is detected, and based on the trained fuzzy neural network model and the detection result, it is determined whether the commutation valve fails to commutate.

2. A method for detecting and determining commutation failure of a commutation valve according to claim 1, wherein: The fuzzy neural network model is established, comprising: The fuzzy neural network model is established with the AC side voltage and the DC side voltage at both ends of the phase-changing converter valve as input variables.

3. A method for detecting and determining commutation failure of a commutation valve according to claim 1 or 2, wherein: The conditions satisfied for the training to be completed include: When the output error of the fuzzy neural network model meets the error requirement, the training is considered qualified and the fuzzy neural network model completes the training.

4. A method for detecting and determining commutation failure of a commutation valve according to claim 1 or 2, wherein: Determine whether the commutation valve fails to commutate, including: Regularly collect the AC side voltage and DC side voltage at both ends of the phase-changing valve; The collected voltage signal, namely the voltage variable, is sent to the fuzzy neural network model for judgment.

5. A method for detecting and determining commutation failure of a commutation valve according to claim 4, wherein: Determining whether the commutation valve fails in commutation also includes: Fuzzify the input voltage variable to make it meet the input requirements of the fuzzy neural network model; Calculate and determine the credibility of the fuzzified input variables for the fuzzy rules; Normalize and calculate the credibility of fuzzy rules; According to the fuzzy rules and combined with the membership function, the output corresponding to each node is calculated; The total output variables are summarized and calculated to obtain a determination signal of whether commutation failure occurs.

6. A method for detecting and determining commutation failure of a commutation valve according to claim 1, 2 or 5, wherein: The structure of the fuzzy neural network model includes: The fuzzification layer is used to fuzzify the input variables, namely, the AC side voltage and the DC side voltage at both ends of the commutation valve; The strength release layer of the fuzzy rules is used to calculate the credibility of the fuzzy rules based on historical data and formulate corresponding fuzzy rules, wherein the fuzzy rules have orders of magnitude differences; Normalization layer, used to calculate the normalized credibility of each fuzzy rule to ensure the equality of each fuzzy rule at different magnitudes; The output layer of fuzzy rule calculation is used to determine the output corresponding to each input variable according to the fuzzy rules and the membership function; The total output layer is used to calculate the total output of all input variables as a decision signal to determine whether commutation failure occurs.

7. A method for detecting and determining commutation failure of a commutation valve according to claim 6, wherein: The fuzzification process is determined by the following formula: Where, u1 and u2 are the AC side voltage and DC side voltage of the two ends of the commutation valve, respectively. i , N i is a fuzzy set, is the fuzzy set M i , N i The membership function of is the membership function of u1, is the membership function of u2, i is the number of nodes, and n is a positive integer.

8. A method for detecting and determining commutation failure of a commutation valve according to claim 7, wherein: The credibility of the fuzzy rule is determined by the following formula: In the formula, and w i is the credibility value of the fuzzy rule.

9. A method for detecting and determining commutation failure of a commutation valve according to claim 8, wherein: The normalized credibility of the fuzzy rule is calculated by the following formula: In the formula, is the output of the normalization layer, which is the normalized credibility value of the fuzzy rule 10. A method for detecting and determining commutation failure of a commutation valve according to claim 9, wherein: The output corresponding to each input variable is calculated using the following formula: In the formula, {p i ,q i ,r i } is the parameter set of the output layer nodes for calculating fuzzy rules, f i is the output function of node i, The output of the output layer for calculating fuzzy rules.

11. A method for detecting and determining commutation failure of a commutation valve according to claim 10, wherein: The total output is calculated by the following formula: In the formula, Indicates the total output.

12. A method for detecting and determining commutation failure of a commutation valve according to any one of claims 1, 2, 5, 7-9, wherein: The historical data is historical data of commutation failure.

13. A phase-changing valve commutation failure detection and determination system, wherein: The system comprises: Establishing a module for establishing a fuzzy neural network model, and using the acquired historical data as a training data set to train the established fuzzy neural network model; The determination module is used to detect the bus voltage after the training is completed, and to determine whether the commutation valve fails to commutate based on the trained fuzzy neural network model.

14. A phase-changing valve commutation failure detection and determination system according to claim 13, wherein: The fuzzy neural network model is established, comprising: The fuzzy neural network model is established with the AC side voltage and the DC side voltage at both ends of the phase-changing converter valve as input variables.

Images