Method for identifying and predicting commutation failure of high-voltage direct-current transmission

Abstract

A high-voltage direct-current transmission commutation failure identification and prediction method comprises the following steps: step 1, continuously detecting the voltage of an inverter side commutation bus, and when detecting that the voltage changes suddenly, carrying out data acquisition and data preprocessing to obtain original electrical quantity data; 2, calculating and converting the first segment of data into samples available for an ELM classifier, and inputting the samples into a decision module; 3, taking the first section of sample transmitted by the feature extraction module as the input of a first commutation failure identification model, judging whether the first commutation failure has occurred or not, and if not, transmitting a result to an output module; 4, if the first commutation failure occurs, sending a signal to the feature extraction module, performing second-section data calculation processing, inputting the second-section data into a subsequent commutation failure prediction model of the decision-making module for subsequent commutation failure prediction, and inputting a prediction result into the output module; and step 5, outputting final commutation failure information by an output module.

Description

Technical Field

[0001] The present invention relates to the field of power systems and automation technology, and in particular to a data-driven identification and prediction model and method for high-voltage direct current transmission commutation failure. The present invention is a divisional application of the invention patent with application number 202210224288.5. Background Art

[0002] The converter valves in the HVDC transmission system are very likely to fail in commutation due to AC system faults. If this causes subsequent commutation failures, it will cause unipolar or bipolar locking of the converter station, resulting in power surplus at the sending end and power deficiency at the receiving end, posing a huge threat to the safety and stability of the power grid.

[0003] The first commutation failure of a high-voltage direct current transmission system is generally difficult to avoid, but its impact on the power grid is limited and can be recovered by itself. Subsequent commutation failures that occur after the first commutation failure will cause multiple shocks to the power grid, threatening the safety of the power grid. In the prior art, the patent document with authorization announcement number CN110441658A discloses a high-voltage direct current commutation failure judgment method that takes into account the change of DC current. It calculates the shutdown angle expression under a single-phase grounding fault and the shutdown angle expression under a three-phase short-circuit fault by establishing a shutdown angle model under a three-phase grounding fault and a shutdown angle model under a single-phase grounding fault. When the shutdown angle under a three-phase short-circuit fault is less than or equal to the minimum shutdown angle of the inverter commutation failure, it is judged as a commutation failure; otherwise, it is judged as a commutation success. When a single-phase grounding fault occurs, the shutdown angle under the single-phase grounding fault is compared with the minimum shutdown angle of the inverter when the commutation failure occurs under the DC standard test system. When the shutdown angle under the single-phase grounding fault is less than or equal to the minimum shutdown angle of the inverter when the commutation failure occurs, it is judged as a commutation failure; otherwise, it is judged as a commutation success.

[0004] The above-mentioned prior art can well realize the judgment of commutation failure by comparing the size of the turn-off angle under fault and the minimum turn-off angle. However, this prior art cannot predict the subsequent commutation failure with greater harm. Summary of the invention

[0005] The purpose of the present invention is to identify the first commutation failure and predict the subsequent commutation failure, so as to solve the technical problems that the existing commutation failure identification and prediction methods are insufficient in applicability, poor in accuracy and slow in prediction speed.

[0006] A model for identifying commutation failure of high-voltage direct current transmission includes a data acquisition module, a feature extraction module, a decision module, and an output module;

[0007] The data acquisition module is used to collect and pre-process the original data;

[0008] The feature extraction module is used to extract features from the original data and classify them, and calculate the proposed feature data to convert it into a sample set that can be used by the classifier;

[0009] The decision module is used to identify the first commutation failure and predict subsequent commutation failures;

[0010] The output module is used to output the prediction results.

[0011] The output end of the data acquisition module is connected to the input end of the feature extraction module, the output end of the feature extraction module is connected to the input end of the decision module, and the output end of the decision module is connected to the input end of the output module;

[0012] The feature extraction module includes a first feature calculation module and a second feature calculation module, and the decision module includes a first commutation failure recognition module and a subsequent commutation failure prediction module;

[0013] The output end of the data acquisition module is connected to the input end of the first feature calculation module and the second feature calculation module, the output end of the first feature calculation unit is connected to the input end of the first commutation failure identification module, the output end of the first commutation failure identification module is connected to the output module and the input end of the second feature calculation module, the output end of the second feature calculation module is connected to the input end of the subsequent commutation failure prediction module, and the output end of the subsequent commutation failure prediction module is connected to the input end of the output module.

[0014] The first feature calculation module is used to convert the first segment original feature quantity into a feature quantity usable by the classifier, the second feature calculation module is used to convert the second segment original feature quantity into a feature quantity usable by the classifier, the first commutation failure identification module is used to identify the first commutation failure, and the subsequent commutation failure prediction module is used to predict subsequent commutation failures.

[0015] The data acquisition module includes a data acquisition module and a preliminary calculation module. The output end of the data acquisition module is connected to the input end of the preliminary calculation module. The output end of the preliminary calculation module is connected to the first feature calculation module and the input end of the second feature calculation module.

[0016] Regarding the data acquisition module, the data acquisition module collects real-time operation data, including: inverter-side commutation bus voltage, DC current, advance trigger angle and shutdown angle parameters. The collected data is preliminarily processed to obtain the effective value of the commutation bus voltage and the instantaneous value of each phase, the effective value of the DC current, the trigger angle command value, and the actual value of the shutdown angle.

[0017] Regarding the feature extraction module, the present invention implements two-stage binary classification based on ELM to identify and predict commutation failure. The required two-stage electrical quantity data is extracted from the data output by the data acquisition module. The first stage extracts the effective value of the inverter side commutation bus voltage and the effective value of the DC current after the fault as the original sample set of the first stage ELM classifier, which can be expressed as:

[0018]

[0019] Where, X 1-original is the original sample set collected in the first segment, is the bth original electrical quantity of the ath original sample, m is the number of samples in the original sample set, and n is the number of original electrical quantities contained in each sample.

[0020] The second extraction time is the time from the fault to the time when the shutdown angle is restored to the reference value. The instantaneous value extraction time of the inverter side commutation bus phase voltage is the fault start time. The collected data includes: DC current effective value, commutation bus voltage effective value and each phase instantaneous value, trigger angle command value. As the original sample set of the second ELM classifier, it can be expressed as

[0021]

[0022] Where, X 2-original is the original sample set collected in the second segment, is the bth original electrical quantity of the ath original sample, m is the number of samples in the original sample set, and n is the number of original electrical quantities contained in each sample.

[0023] The original feature quantity is converted into a feature quantity, and the feature quantity used for the first segment is:

[0024] I 1-max : Maximum current, U 1-min : minimum voltage;

[0025]

[0026] Where, X 1-data To calculate the sample set of the first segment, is the bth original electrical quantity of the ath sample, m is the number of samples in the sample set, and n is the number of feature quantities contained in each sample.

[0027] The feature quantities used for the second segment are:

[0028] I 2-max : Maximum current, U 2-max : Maximum voltage, α 2-max : Maximum trigger angle, I 2-min : Minimum current, U2-min : Minimum voltage, α 2-min : Minimum trigger angle, I 2-average : Average current value, U 2-averag : Average voltage, α 2-average : Average value of firing angle, THD u : Total voltage harmonic distortion rate;

[0029] △T: Sampling time interval

[0030]

[0031] Where, X 2-data To calculate the sample set of the second segment, is the bth original electrical quantity of the ath sample, m is the number of samples in the sample set, and n is the number of feature quantities contained in each sample.

[0032] The first sample set is used to identify the first commutation failure, and the second sample set is used to predict subsequent commutation failures. In addition, the calculation and processing of the second sample set data must be started according to the decision module. If the decision module determines that the first commutation failure occurs, the processing of the second data is started and the subsequent process continues, otherwise it is not started.

[0033] Regarding the decision module, the decision module is divided into two parts, namely, the first commutation failure recognition and the subsequent commutation failure prediction, among which the subsequent commutation failure prediction is realized on the basis of the first commutation failure recognition. First, the first segment of the sample set in the feature extraction module is input into the pre-trained first commutation failure model to realize the first commutation failure recognition. If the model output result is that no commutation failure occurs, the recognition result is output to the output module; if the result is that the first commutation failure occurs, a signal is sent to the feature extraction module to start processing the second segment of data. Then, the converted sample set is input into the pre-trained subsequent commutation failure prediction model in the decision module to realize the subsequent commutation failure prediction. Finally, the model prediction result is output to the output module.

[0034] When the feature extraction module and the decision module work together, the following steps are included:

[0035] Step 1: The feature extraction module extracts two pieces of electrical feature quantity data required for commutation failure identification and prediction from the original electrical quantity data;

[0036] Step 2: The first segment of data is converted into a sample that can be used by the classifier by the first feature calculation module and input into the first commutation failure identification module in the decision module to perform first commutation failure identification;

[0037] Step 3: judging whether the commutation has failed based on the output result of the first commutation failure identification module; if the output result is that no commutation failure has occurred, outputting the identification result to the output module;

[0038] Step 4: If the output result is that the first commutation failure occurs, a signal is sent to the second feature calculation module, and the second feature calculation module calculates the second segment of data and converts it into a sample that can be used by the classifier and inputs it to the subsequent commutation failure prediction module in the decision module to perform subsequent commutation failure prediction;

[0039] Step 5: The output module outputs the model prediction results.

[0040] An ELM network model for commutation identification and prediction is established by using the following steps:

[0041] Step 1. Train the first commutation failure recognition model;

[0042] Step 1 includes the following steps:

[0043] 1) Select the infinitely differentiable function g1(x) as the activation function of the hidden layer nodes, and randomly set the weight ω1 between the input layer and the hidden layer and the threshold b1 of the hidden layer nodes;

[0044] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0045] 3) Calculate the hidden layer output matrix H1;

[0046] 4) Calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0047] 5) Calculate the output weight β1 and obtain the model output T1=H1β1;

[0048] The above steps complete the first commutation failure recognition model training;

[0049] Step 2. Training the subsequent commutation failure prediction model;

[0050] Step 2 includes the following steps:

[0051] 1) Select the infinitely differentiable function g2(x) as the activation function of the hidden layer nodes, and randomly set the weight ω2 between the input layer and the hidden layer and the threshold b2 of the hidden layer nodes;

[0052] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0053] 3) Calculate the hidden layer output matrix H2;

[0054] 4) Calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ;

[0055] 5) Calculate the output weight β2 and obtain the model output T2 = H2β2;

[0056] The above steps complete the subsequent commutation failure prediction model training;

[0057] In step 1.3), the matrix H1 is:

[0058]

[0059] Among them, r is the number of samples, l is the number of hidden layer nodes, F 1j ,ω 1i , b 1i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., r; i=1, 2, ..., l);

[0060] In step 1.4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0061] In step 1.5), the calculation formula of β1 is as follows:

[0062] β1=H1 + T1.

[0063] In step 2.3), the matrix H2 is:

[0064]

[0065] Among them, R is the number of samples, L is the number of hidden layer nodes, and F 2j ,ω 2i , b 2i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., R; i=1, 2, ..., L);

[0066] In step 2.4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ;

[0067] In step 2.5), the calculation formula of β2 is as follows:

[0068] β2=H2 + T2.

[0069] In the first commutation failure identification model, after obtaining the classification result T1, it is identified whether commutation failure has occurred under each fault based on the electrical characteristic quantity under the fault;

[0070] In the subsequent commutation failure prediction model, after the classification result T2 is obtained, it is predicted whether each fault will lead to subsequent commutation failure based on the electrical characteristic quantity under the fault.

[0071] A method for identifying and predicting commutation failure of high voltage direct current transmission comprises the following steps:

[0072] Step 1: Continuously detect the inverter-side commutation bus voltage. When a sudden change in voltage is detected, data acquisition and data preprocessing are performed to obtain the original electrical quantity data.

[0073] Step 2: Extract two segments of electrical feature quantities required for commutation failure identification and prediction from the original electrical quantity data, convert the first segment of data into samples that can be used by the ELM classifier and input them into the decision module;

[0074] Step 3: Use the first sample sent by the feature extraction module as the input of the first commutation failure recognition model to determine whether the first commutation failure has occurred. If not, send the result to the output module;

[0075] Step 4: If the first commutation failure occurs, a signal is sent to the feature extraction module, the second segment of data is calculated and processed, and the data is input into the subsequent commutation failure prediction model of the decision module to predict the subsequent commutation failure and the prediction result is input into the output module;

[0076] Step 5: The output module outputs the final commutation failure information.

[0077] The ELM network model is used as the first commutation failure recognition model. When the first commutation failure recognition model is trained, the following steps are adopted:

[0078] 1) Select the infinitely differentiable function g1(x) as the activation function of the hidden layer nodes, and randomly set the weight ω1 between the input layer and the hidden layer and the threshold b1 of the hidden layer nodes;

[0079] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0080] 3) Calculate the hidden layer output matrix H1;

[0081] 4) Calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0082] 5) Calculate the output weight β1 and obtain the model output T1=H1β1;

[0083] The above steps complete the first commutation failure recognition model training.

[0084] In step 3), the matrix H1 is:

[0085]

[0086] Among them, r is the number of samples, l is the number of hidden layer nodes, F 1j ,ω 1i 、b 1i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., r; i=1, 2, ..., l);

[0087] In step 4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0088] In step 5), the calculation formula of β1 is as follows:

[0089] β1=H1 + T1.

[0090] The ELM network model is used as a subsequent commutation failure prediction model. When training the subsequent commutation failure prediction model, the following steps are adopted:

[0091] 1) Select the infinitely differentiable function g2(x) as the activation function of the hidden layer nodes, and randomly set the weight ω2 between the input layer and the hidden layer and the threshold b2 of the hidden layer nodes;

[0092] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0093] 3) Calculate the hidden layer output matrix H2;

[0094] 4) Calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ;

[0095] 5) Calculate the output weight β2 and obtain the model output T2=H2β2; The above steps are used to complete the subsequent commutation failure recognition model training.

[0096] The above steps complete the subsequent commutation failure prediction model training.

[0097] In step 3), the matrix H2 is:

[0098]

[0099] Among them, R is the number of samples, L is the number of hidden layer nodes, and F 2j ,ω 2i 、b 2i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., R; i=1, 2, ..., L);

[0100] In step 4), the Moor-Penrose generalized inverse H2 of the output matrix H2 is calculated using the singular value decomposition method + ;

[0101] In step 5), the calculation formula of β2 is as follows:

[0102] β2=H2 + T2.

[0103] Compared with the prior art, the present invention has the following technical effects:

[0104] The present invention is based on a data-driven ELM classification recognition and prediction model, and considers a variety of commutation failure influencing factors as feature quantities, making up for the shortcomings of the previous quantification and calculation methods based mainly on current and voltage factors. This model can identify the first commutation failure and predict subsequent commutation failures, solving the problems of insufficient applicability, poor accuracy and slow prediction speed of existing commutation failure identification and prediction methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0106] Figure 1 This is a structural diagram of a commutation failure identification and prediction model of the present invention;

[0107] Figure 2 It is a schematic diagram of the commutation failure identification and prediction model flow of the present invention;

[0108] Figure 3 The invention provides an IEEE39 node modification system including direct current transmission. DETAILED DESCRIPTION

[0109] like Figure 1 As shown, a model for identifying commutation failure of high-voltage direct current transmission includes a data acquisition module 1, a feature extraction module 2, a decision module 3, and an output module 4;

[0110] The data acquisition module 1 is used to collect and preprocess the original data;

[0111] The feature extraction module 2 is used to extract features from the original data and classify them, and calculate the proposed feature data to convert them into a sample set that can be used by the classifier;

[0112] Decision module 3 is used to identify the first commutation failure and predict subsequent commutation failures;

[0113] The output module 4 is used to output the prediction results.

[0114] The output end of the data acquisition module 1 is connected to the input end of the feature extraction module 2, the output end of the feature extraction module 2 is connected to the input end of the decision module 3, and the output end of the decision module 3 is connected to the input end of the output module 4;

[0115] The feature extraction module 2 includes a first feature calculation module 2-1 and a second feature calculation module 2-2, and the decision module 3 includes a first commutation failure recognition module 3-1 and a subsequent commutation failure prediction module 3-2;

[0116] The output end of the data acquisition module 1 is connected to the input end of the first feature calculation module 2-1 and the second feature calculation module 2-2, the output end of the first feature calculation unit 2-1 is connected to the input end of the first commutation failure identification module 3-1, the output end of the first commutation failure identification module 3-1 is connected to the output module 4 and the input end of the second feature calculation module 2-2, the output end of the second feature calculation module 2-2 is connected to the input end of the subsequent commutation failure prediction module 3-2, and the output end of the subsequent commutation failure prediction module 3-2 is connected to the input end of the output module 4.

[0117] The first feature calculation module 2-1 is used to convert the first segment of original feature quantities into feature quantities that can be used by the classifier, the second feature calculation module 2-2 is used to convert the second segment of original feature quantities into feature quantities that can be used by the classifier, the first commutation failure identification module 3-1 is used to identify the first commutation failure, and the subsequent commutation failure prediction module 3-2 is used to predict subsequent commutation failures.

[0118] The data acquisition module 1 includes a data acquisition module 1-1 and a preliminary calculation module 1-2. The output end of the data acquisition module 1-1 is connected to the input end of the preliminary calculation module 1-2. The output end of the preliminary calculation module 1-2 is connected to the input end of the first feature calculation module 2-1 and the second feature calculation module 2-2.

[0119] When the feature extraction module 2 and the decision module 3 work together, the following steps are included:

[0120] Step 1: The feature extraction module 2 extracts two pieces of electrical feature quantity data required for commutation failure identification and prediction from the original electrical quantity data;

[0121] Step 2: The first segment of data is converted into a sample that can be used by the classifier by the first feature calculation module 2-1 and input into the first commutation failure identification module 3-1 in the decision module 3 to identify the first commutation failure;

[0122] Step 3: judging whether the commutation has failed based on the output result of the first commutation failure identification module 3-1; if the output result is that no commutation failure has occurred, outputting the identification result to the output module 4;

[0123] Step 4: If the output result is that the first commutation failure occurs, a signal is sent to the second feature calculation module 2-2, and the second feature calculation module 2-2 calculates the second segment of data and converts it into a sample that can be used by the classifier and inputs it to the subsequent commutation failure prediction module 3-2 in the decision module 3 to predict the subsequent commutation failure;

[0124] Step 5: The output module 4 outputs the model prediction results.

[0125] The present invention also includes an ELM network model for phase change identification and prediction,

[0126] Its purpose is to solve the technical problems of insufficient applicability, poor accuracy and slow prediction speed of the existing network models for identifying and predicting HVDC transmission commutation failures.

[0127] It takes the following steps when it is built:

[0128] Step 1. Train the first commutation failure recognition model;

[0129] Step 1 includes the following steps:

[0130] 1) Select the infinitely differentiable function g1(x) as the activation function of the hidden layer nodes, and randomly set the weight ω1 between the input layer and the hidden layer and the threshold b1 of the hidden layer nodes;

[0131] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0132] 3) Calculate the hidden layer output matrix H1;

[0133] 4) Calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0134] 5) Calculate the output weight β1 and obtain the model output T1=H1β1;

[0135] The above steps complete the first commutation failure recognition model training;

[0136] Step 2. Training the subsequent commutation failure prediction model;

[0137] Step 2 includes the following steps:

[0138] 1) Select the infinitely differentiable function g2(x) as the activation function of the hidden layer nodes, and randomly set the weight ω2 between the input layer and the hidden layer and the threshold b2 of the hidden layer nodes;

[0139] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0140] 3) Calculate the hidden layer output matrix H2;

[0141] 4) Calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ;

[0142] 5) Calculate the output weight β2 and obtain the model output T2 = H2β2;

[0143] The following steps are used to complete the subsequent commutation failure prediction model training;

[0144] In step 1.3), the matrix H1 is:

[0145]

[0146] Among them, r is the number of samples, l is the number of hidden layer nodes, F 1j ,ω 1i 、b 1i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., r; i=1, 2, ..., l);

[0147] In step 1.4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0148] In step 1.5), the calculation formula of β1 is as follows:

[0149] β1=H1 + T1.

[0150] In step 2.3), the matrix H2 is:

[0151]

[0152] Among them, R is the number of samples, L is the number of hidden layer nodes, and F 2j ,ω 2i , b 2i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., R; i=1, 2, ..., L);

[0153] In step 2.4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ;

[0154] In step 2.5), the calculation formula of β2 is as follows:

[0155] β2=H2 + T2.

[0156] In the first commutation failure identification model, after obtaining the classification result T1, it is identified whether commutation failure has occurred under each fault based on the electrical characteristic quantity under the fault;

[0157] In the subsequent commutation failure prediction model, after the classification result T2 is obtained, it is predicted whether each fault will lead to subsequent commutation failure based on the electrical characteristic quantity under the fault.

[0158] The present invention also includes a method for identifying and predicting high voltage direct current transmission commutation failure.

[0159] Its purpose is to solve the technical problems of insufficient applicability, poor accuracy and slow prediction speed of existing HVDC transmission commutation failure identification and prediction methods;

[0160] The following steps are involved:

[0161] Step 1: Continuously detect the inverter-side commutation bus voltage. When a sudden change in voltage is detected, data acquisition and data preprocessing are performed to obtain the original electrical quantity data.

[0162] Step 2: Extract two segments of electrical feature quantities required for commutation failure identification and prediction from the original electrical quantity data, convert the first segment of data into samples that can be used by the ELM classifier and input them into the decision module;

[0163] Step 3: Use the first sample sent by the feature extraction module as the input of the first commutation failure recognition model to determine whether the first commutation failure has occurred. If not, send the result to the output module;

[0164] Step 4: If the first commutation failure occurs, a signal is sent to the feature extraction module, the second segment of data is calculated and processed, and the data is input into the subsequent commutation failure prediction model of the decision module to predict the subsequent commutation failure and the prediction result is input into the output module;

[0165] Step 5: The output module transmits the final commutation failure information.

[0166] The ELM network model is used as the first commutation failure recognition model. When the first commutation failure recognition model is trained, the following steps are adopted:

[0167] 1) Select the infinitely differentiable function g1(x) as the activation function of the hidden layer nodes, and randomly set the weight ω1 between the input layer and the hidden layer and the threshold b1 of the hidden layer nodes;

[0168] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0169] 3) Calculate the hidden layer output matrix H1;

[0170] 4) Calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0171] 5) Calculate the output weight β1 and obtain the model output T1=H1β1;

[0172] The above steps complete the first commutation failure recognition model training.

[0173] In step 3), the matrix H1 is:

[0174]

[0175] Among them, r is the number of samples, l is the number of hidden layer nodes, F 1j ,ω 1i 、b 1i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., r; i=1, 2, ..., l);

[0176] In step 4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ;

[0177] In step 5), the calculation formula of β1 is as follows:

[0178] β1=H1 + T1.

[0179] The ELM network model is used as a subsequent commutation failure prediction model. When training the subsequent commutation failure prediction model, the following steps are adopted:

[0180] 1) Select the infinitely differentiable function g2(x) as the activation function of the hidden layer nodes, and randomly set the weight ω2 between the input layer and the hidden layer and the threshold b2 of the hidden layer nodes;

[0181] 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes;

[0182] 3) Calculate the hidden layer output matrix H2;

[0183] 4) Calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ;

[0184] 5) Calculate the output weight β2 and obtain the model output T2 = H2β2;

[0185] The above steps complete the subsequent commutation failure prediction model training.

[0186] In step 3), the matrix H2 is:

[0187]

[0188] Among them, R is the number of samples, L is the number of hidden layer nodes, and F 2j ,ω 2i 、b 2i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., R; i=1, 2, ..., L);

[0189] In step 4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ;

[0190] In step 5), the calculation formula of β2 is as follows:

[0191] β2=H2 + T2.

[0192] In order to facilitate those skilled in the art to better understand the present invention, further explanations and descriptions are as follows:

[0193] A model for identifying commutation failure of high-voltage direct current transmission includes a data acquisition module, a feature extraction module, a decision module, and an output module. The data acquisition module performs raw data acquisition and data preprocessing; the feature extraction module extracts features from the raw data and classifies them, and at the same time calculates the proposed feature data to convert them into a sample set that can be used by the classifier; the decision module identifies the first commutation failure and predicts subsequent commutation failures through an ELM network trained in advance with simulation data; and the output module outputs the prediction results. The output end of the data acquisition module is connected to the input end of the feature extraction module, the output end of the feature extraction module is connected to the input end of the decision module, and the output end of the decision module is connected to the input end of the output module. In addition, the feature extraction module also has an input end from the decision module.

[0194] Specifically, the functions of the above four modules are as follows:

[0195] 1. Data acquisition module

[0196] The real-time operation data is collected, and the sampling time interval can be 0.0002s. The data includes: inverter side commutation bus voltage, DC current, advance trigger angle and shutdown angle parameters. The collected data is preliminarily processed to obtain the effective value of the commutation bus voltage and the instantaneous value of each phase, the effective value of the DC current, the trigger angle command value, and the actual value of the shutdown angle.

[0197] 2. Feature extraction module

[0198] The present invention implements two-stage binary classification based on ELM to identify and predict commutation failure. The required two-stage electrical quantity data are extracted from the data output by the data acquisition module. The first stage extracts the effective value of the inverter side commutation bus voltage and the effective value of the DC current after the fault (which can be 0.03s) as the original sample set of the first stage ELM classifier, which can be expressed as:

[0199]

[0200] Where, X 1-original is the original sample set collected in the first segment, is the bth original electrical quantity of the ath original sample, m is the number of samples in the original sample set, and n is the number of original electrical quantities contained in each sample.

[0201] The second extraction time is the time from the fault (which can be 0.03s) to the time when the turn-off angle is restored to the reference value. The instantaneous value extraction time of the inverter side commutation bus phase voltage is the fault start time. The collected data includes: DC current effective value, commutation bus voltage effective value and each phase instantaneous value, trigger angle command value. As the original sample set of the second ELM classifier, it can be expressed as

[0202]

[0203] In the formula, X 2-original is the original sample set collected in the second segment, is the bth original electrical quantity of the ath original sample, m is the number of samples in the original sample set, and n is the number of original electrical quantities contained in each sample.

[0204] The original feature quantity is converted into a feature quantity, and the feature quantity used for the first segment is:

[0205] I 1-max : Maximum current; U 1-min : minimum voltage;

[0206]

[0207] In the formula, X 1-data To calculate the sample set of the first segment, is the bth original electrical quantity of the ath sample, m is the number of samples in the sample set, and n is the number of feature quantities contained in each sample.

[0208] The feature quantities used for the second segment are:

[0209] I 2-max : Maximum current; U 2-max : Maximum voltage; α 2-max : Maximum value of the trigger angle; I 2-min : Minimum current value; U 2-min : Minimum voltage; α 2-min : Minimum trigger angle; I 2-average : Average current value; U 2-averag : voltage average value; α 2-average : Average value of firing angle; THD u : total voltage harmonic distortion rate; △T: sampling time interval;

[0210]

[0211] In the formula, X 2-data To calculate the sample set of the second segment, is the bth original electrical quantity of the ath sample, m is the number of samples in the sample set, and n is the number of feature quantities contained in each sample.

[0212] The first sample set is used to identify the first commutation failure, and the second sample set is used to predict subsequent commutation failures. In addition, the calculation and processing of the second sample set data must be started according to the decision module. If the decision module determines that the first commutation failure occurs, the processing of the second data is started and the subsequent process continues, otherwise it is not started.

[0213] 3. Decision-making module

[0214] The decision module is divided into two parts, namely, the first commutation failure recognition and the subsequent commutation failure prediction, among which the subsequent commutation failure prediction is realized on the basis of the first commutation failure recognition. First, the first sample set in the feature extraction module is input into the pre-trained first commutation failure model to realize the first commutation failure recognition. If the model output result is that no commutation failure occurs, the recognition result is output to the output module; if the result is that the first commutation failure occurs, a signal is sent to the feature extraction module to start processing the second segment of data. Then, the converted sample set is input into the pre-trained subsequent commutation failure prediction model in the decision module to realize the subsequent commutation failure prediction. Finally, the model prediction result is output to the output module.

[0215] The present invention realizes commutation failure recognition and prediction based on the ELM two-stage two-classification algorithm, which requires two different ELM network models to be trained in advance. The training process is as follows:

[0216] The training model is trained using simulation data, and the data set is generated by setting faults on the busbars of different nodes in the inverter-side AC system. The relevant parameters are shown in Table 1.

[0217] Table 1 Fault setting related parameters

[0218]

[0219] Corresponding to the commutation failure information, two sample sets are labeled for training the model. In the first sample set, the sample label without the first commutation failure is 0, and the label with the first commutation failure is 1. In the second sample set, the sample label without subsequent commutation failure is 2, and the label with subsequent commutation failure is 3.

[0220] The first commutation failure recognition model is trained. The specific training steps are:

[0221] (1) Select an infinitely differentiable function g1(x) as the activation function of the hidden layer nodes, and randomly set the weight ω1 between the input layer and the hidden layer and the threshold b1 of the hidden layer nodes.

[0222] The activation functions of the ELM algorithm used are: S-shaped function (Sig), sine function (Sin), triangular basis function (Triangular basis), hard threshold function (Hardlim) and radial basis function (Radial basis).

[0223] The sample set used for the first segment classification is used as the input of the ELM algorithm, and the number of hidden layer nodes is selected as 10-60. The accuracy of the training set and test set of different activation functions is compared. Different activation functions have different accuracy rates, and the highest accuracy rate is selected as the activation function of the first commutation failure recognition model.

[0224] (2) Select the function with the highest accuracy as the activation function. By changing the number of hidden layer nodes of the ELM algorithm (the number of hidden layer nodes is generally less than the number of training sets), the accuracy with different numbers of hidden layer nodes is compared, and the optimal number of hidden layer nodes is selected.

[0225] (3) Calculate the hidden layer output matrix H1:

[0226]

[0227] Among them, r is the number of samples, l is the number of hidden layer nodes, F 1j ,ω 1i 、b 1i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., r; i=1, 2, ..., l);

[0228] (4) Use the singular value decomposition method to calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + .

[0229] (5) Calculate the output weight β1 and obtain the model output T1=H1β1. The calculation formula of β is as follows:

[0230] β1=H1 + T1

[0231] (6) Complete the first commutation failure identification model training.

[0232] The subsequent commutation failure prediction model is trained. The specific training steps are:

[0233] (1) Select the infinitely differentiable function g2(x) as the activation function of the hidden layer nodes, and randomly set the weight ω2 between the input layer and the hidden layer and the threshold b2 of the hidden layer nodes.

[0234] The activation functions of the ELM algorithm used are: S-shaped function (Sig), sine function (Sin), triangular basis function (Triangular basis), hard threshold function (Hardlim) and radial basis function (Radial basis).

[0235] The sample set used for the second stage classification is used as the input of the ELM algorithm, and the number of hidden layer nodes is selected as 10-60. The accuracy of the training set and test set of different activation functions is compared. Different activation functions have different accuracy rates, and the highest accuracy rate is selected as the activation function of the first commutation failure recognition model.

[0236] (2) Select the function with the highest accuracy as the activation function. By changing the number of hidden layer nodes of the ELM algorithm (the number of hidden layer nodes is generally less than the number of training sets), the accuracy with different numbers of hidden layer nodes is compared, and the optimal number of hidden layer nodes is selected.

[0237] (3) Calculate the hidden layer output matrix H2:

[0238]

[0239] Among them, R is the number of samples, L is the number of hidden layer nodes, and F 2j ,ω 2i , b 2i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., R; i=1, 2, ..., L);

[0240] (4) Use the singular value decomposition method to calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + .

[0241] (5) Calculate the output weight β2 and obtain the model output T2=H2β2. The calculation formula of β is as follows:

[0242] β2=H2 + T2

[0243] (6) Complete the subsequent commutation failure prediction model training.

[0244] For a given input data F r 、F R , the classification results T1, T2 can be expressed as:

[0245] T1=H1(F 11 ,…F 1r ,ω 11 ,…ω 1l ,b 11 ,…b 1l )

[0246] T2=H2(F 21 ,…F 2R ,ω 21 ,…ω 2L ,b 21 ,…b 2L )

[0247] In the first commutation failure identification model, after the classification result T1 is obtained, it is identified whether a commutation failure has occurred under each fault based on the electrical characteristic quantity under the fault.

[0248] In the subsequent commutation failure prediction model, after the classification result T2 is obtained, it is predicted whether each fault will lead to subsequent commutation failure based on the electrical characteristic quantity under the fault.

[0249] Next, the effect of the commutation failure identification and prediction model in the present invention is evaluated, and the classification accuracy (CA) is used as the performance evaluation index of the classifier. The higher the accuracy of the two-segment classification results, the better the effect of the segmented classification model proposed by the present invention.

[0250] Classification accuracy (CA) is a relatively traditional evaluation standard and has been widely used. The calculation formula is as follows:

[0251]

[0252] Among them, for a given test set, TP and TN are the number of samples judged as positive and negative respectively, and P and N are the total number of samples of positive and negative respectively.

[0253] 4. Output module:

[0254] Output the decision results of the two ELM classifiers.

[0255] A data-driven method for identifying and predicting commutation failure in high-voltage direct current transmission has the following process:

[0256] First, the inverter-side commutation bus voltage is continuously detected. When a sudden change in voltage is detected, the model is started to enter the data acquisition module for data acquisition and data preprocessing to obtain the original electrical quantity data; then, two sections of electrical feature quantities required for commutation failure identification and prediction are extracted from the original electrical quantity data, and the first section of data is calculated and converted into samples available to the ELM classifier and input into the decision module; then, the first section of samples transmitted by the feature extraction module is used as the input of the first commutation failure identification model to determine whether the first commutation failure has occurred. If not, the result is transmitted to the output module. If it has occurred, a signal is sent to the feature extraction module to calculate and process the second section of data and input it into the subsequent commutation failure prediction model of the decision module for subsequent commutation failure prediction and the prediction result is input into the output module; finally, the output module transmits the final commutation failure information.

[0257] Simulation Model:

[0258] In the PSCAD / EMDTC electromagnetic transient simulation software, four DC circuits are added to the IEEE 39-node standard system to form the following Figure 3 The simulation model is shown in Figure 1. Among them, the LCC-HVDC system adopts the CIGRE standard model, the system model is a single-pole HVDC system, the DC rated voltage is 500kV, the DC rated power is 1000MW, the inverter side AC rated voltage is 230kV, and the inverter side shutdown angle reference value is 18°.

Claims

1. A method for identifying and predicting commutation failure in high voltage direct current transmission, characterized in that: The following steps are involved: Step 1: Continuously detect the inverter-side commutation bus voltage. When a sudden change in voltage is detected, data acquisition and data preprocessing are performed to obtain the original electrical quantity data. Step 2: Extract two segments of electrical feature quantities required for commutation failure identification and prediction from the original electrical quantity data, convert the first segment of data into samples that can be used by the ELM classifier and input them into the decision module; Step 3: Use the first sample sent by the feature extraction module as the input of the first commutation failure recognition model to determine whether the first commutation failure has occurred. If not, send the result to the output module; Step 4: If the first commutation failure occurs, a signal is sent to the feature extraction module, the second segment of data is calculated and processed, and the data is input into the subsequent commutation failure prediction model of the decision module to predict the subsequent commutation failure and the prediction result is input into the output module; Step 5: The output module transmits the final commutation failure information.

2. The method according to claim 1, characterized in that: The ELM network model is used as the first commutation failure recognition model. When the first commutation failure recognition model is trained, the following steps are adopted: 1) Select the infinitely differentiable function g1(x) as the activation function of the hidden layer nodes, and randomly set the weight ω1 between the input layer and the hidden layer and the threshold b1 of the hidden layer nodes; 2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes; 3) Calculate the hidden layer output matrix H1; 4) Calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ; 5) Calculate the output weight β1 and obtain the model output T1=H1β1; The above steps complete the first commutation failure recognition model training.

3. The method according to claim 2, characterized in that In step 3), the matrix H1 is: Among them, r is the number of samples, l is the number of hidden layer nodes, F 1j ,ω 1i 、b 1i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., r; i=1, 2, ..., l); In step 4), the singular value decomposition method is used to calculate the Moor-Penrose generalized inverse H1 of the output matrix H1 + ; In step 5), the calculation formula of β1 is as follows: β1=H1 + T1。 4. The method according to claim 2, characterized in that: The ELM network model is used as a subsequent commutation failure prediction model. When training the subsequent commutation failure prediction model, the following steps are adopted: S1) Select the infinitely differentiable function g2(x) as the activation function of the hidden layer nodes, and randomly set the weight ω2 between the input layer and the hidden layer and the threshold b2 of the hidden layer nodes; S2) Select the function with the highest accuracy as the activation function, and compare the accuracy with different numbers of hidden layer nodes by changing the number of hidden layer nodes of the ELM algorithm, so as to select the optimal number of hidden layer nodes; S3) calculating the hidden layer output matrix H2; S4) Calculate the Moor-Penrose generalized inverse H2 of the output matrix H2 + ; S5) Calculate the output weight β2 to obtain the model output T2=H2β2; The subsequent commutation failure recognition model training is completed through the above steps; The above steps complete the subsequent commutation failure prediction model training.

5. The method according to claim 4, characterized in that In step S3), the matrix H2 is: Among them, R is the number of samples, L is the number of hidden layer nodes, and F 2j ,ω 2i 、b 2i are the input vector, input weight vector and threshold vector of the hidden layer neurons respectively (j=1, 2, ..., R; i=1, 2, ..., L); In step S4), the Moor-Penrose generalized inverse H2 of the output matrix H2 is calculated using the singular value decomposition method + ; In step S5), the calculation formula of β2 is as follows: β2=H2 + T2。 6. The method according to claim 1, characterized in that The decision module is used to identify the first commutation failure and predict the subsequent commutation failure; the feature extraction module is used to extract and classify features from the original data, and calculate the proposed feature data to convert it into a sample set that can be used by the classifier.

Images