Wavelet neural network-based upfc device fault prediction method and system

Abstract

The application relates to a UPFC device fault prediction method based on a wavelet neural network, and comprises the following steps: defining a fault variable of a unified power flow controller; classifying the fault variable, and defining a data acquisition structure for device state monitoring and fault prediction; acquiring signal data of each node of the unified power flow controller, classifying and coding the signal data after filtering and A / D conversion; constructing a device fault prediction model by adopting a wavelet neural network; inputting real-time signal data into the trained device fault prediction model to obtain a device state prediction result output by the device fault prediction model; judging whether a device fault exists according to the device state prediction result; and issuing a fault early warning signal for the predicted device fault. The application realizes multi-dimensional and all-around UPFC online state monitoring, fault diagnosis and fault early warning, predicts specific fault information of the UPFC device in advance, effectively protects the operation safety of the device, and avoids damage to other power electronic elements caused by the UPFC device fault.

Description

Technical Field

[0001] This invention belongs to the field of power engineering, specifically relating to a method and system for predicting UPFC equipment faults based on wavelet neural networks. Background Technology

[0002] The modern power system is a complex, high-dimensional, and highly nonlinear system. It is a power network formed by the interconnection and mutual influence of various power equipment, including generation, transmission, and distribution. Currently, it is developing towards higher voltage, larger capacity, larger scale, and longer distances. With the development of my country's power system construction, the interconnection of power grids in different regions has become an inevitable trend. The current power system's regulation and control of transmission lines is limited to relay protection and reclosing. However, many power system accidents and operational problems occur on transmission lines, making them a "bottleneck" for the safe and economical operation of the power system. The emergence of Unified Power Flow Controller (UPFC) technology provides a very important control tool for solving this problem.

[0003] As a third-generation flexible AC transmission system (FACTS) device, UPFC primarily achieves power flow regulation in terms of power system stability, rationally controlling active and reactive power, improving the overall transmission capacity of the power grid, and optimizing power flow distribution. Dynamically, UPFC improves system voltage stability by dynamically supporting the voltage at the connection point through rapid reactive power compensation. Simultaneously, UPFC can improve system damping and enhance power angle stability. This new technology has broad application prospects in my country. The main function of UPFC is to control the transmission power of lines and bus voltage. Research on utilizing FACTS components, especially UPFC, to significantly improve the controllability of power flow and voltage under existing equipment conditions, reducing the reserve capacity of interconnected systems, is suitable for the situation in my country's power construction where large-scale equipment upgrades are not feasible due to budget constraints. It also allows for the use of new technologies to improve existing power transmission levels and the safe and reliable operation of the power system, possessing both theoretical and practical significance.

[0004] Upstream Synchronous Power Controller (UPFC) is a novel power flow control device combining parallel-compensated static synchronous compensators (SSCs) and series-compensated static synchronous series compensators (SSCs). The UPFC device has a complex structure, consisting of a parallel-side transformer, a series-side transformer, a series-side converter, a parallel-side converter, a DC switch, and an AC connection bus. When a UPFC device fails, severe inrush currents and voltages often occur, easily damaging the high-cost internal power electronic modules. Although the internal protection devices of the UPFC can act immediately to isolate the faulty equipment and reduce the impact of the UPFC failure on the system, many sensitive power electronic components are damaged during overvoltage or overcurrent faults, thus affecting the control function of the UPFC device and consequently impacting the safe and stable operation of the power grid.

[0005] Adopting effective methods to quickly and accurately detect the operating status of UPFC equipment, timely predict and locate the fault type of UPFC equipment, predict the degree of impact of equipment status, and reduce the damage of faults to the system has significant engineering application value and practical significance for improving the safety and reliability of equipment and systems.

[0006] Existing UPFC fault early warning methods have the following problems:

[0007] (1) The number of fault types diagnosed is limited, and only a few fault types can be warned. However, the actual operating status of UPFC is complex, resulting in a wide variety of equipment faults and a lack of comprehensive equipment status warning methods.

[0008] (2) The fault signal data comes from the modeling and simulation results. The equipment operating parameters are related to the UPFC system model. The simulation model is an equivalent modeling data parameter, which is different from the actual operating network parameters of the power system. The simulation data has the problem of distortion.

[0009] (3) UPFC equipment lacks a long-term and effective online monitoring method, and the internal protection system of UPFC does not have the functions of fault early warning and location at all time scales. Summary of the Invention

[0010] The purpose of this invention is to address the aforementioned problems by providing a UPFC (Upstream Power Controller) equipment fault prediction method based on wavelet neural networks. This method categorizes UPFC equipment faults into two main types: system-wide faults and valve-level equipment faults. A data acquisition structure for fault prediction is defined for each UPFC equipment fault. A wavelet neural network is used to construct a fault prediction model. The input layer neurons of this model correspond to the variables of the data acquisition structure, and the hidden layer contains neurons that correspond one-to-one with UPFC equipment faults. Wavelet functions are used as the activation functions of these neurons to improve prediction accuracy. This method also offers better fault tolerance and robustness against noisy data, exhibits adaptability and learning capabilities, and ensures real-time accuracy of fault prediction. Furthermore, it enables multi-dimensional and comprehensive online status monitoring, fault diagnosis, and fault early warning for UPFC devices, predicting specific fault information in advance to avoid damage to other power electronic components caused by UPFC equipment faults, thereby improving the safety and stability of power grid operation.

[0011] The technical solution of this invention is a UPFC equipment fault prediction method based on wavelet neural networks, comprising the following steps:

[0012] Step 1: Define the fault variables for the unified power flow controller;

[0013] Step 2: Classify the data acquisition variables and define the data acquisition structure for equipment status monitoring and fault prediction based on their classification;

[0014] Step 3: Obtain the signal data of each node of the unified power flow controller, and perform filtering, A / D conversion and classification encoding on the signal data in sequence;

[0015] Step 4: Construct a device fault prediction model using a wavelet neural network, and train the device fault prediction model using a historical fault signal dataset;

[0016] Step 5: Input the real-time signal data obtained in Step 3 into the trained equipment fault prediction model to obtain the equipment status prediction result output by the equipment fault prediction model;

[0017] Step 6: Based on the equipment status prediction results obtained in Step 5, determine whether there is any equipment failure;

[0018] Step 7: Based on the fault judgment results of Step 6, issue a fault warning signal for the predicted equipment fault.

[0019] Furthermore, step 1 also includes defining fault classification variables for the unified power flow controller: overall system fault S and valve-level device fault V.

[0020] Preferably, the overall system fault S includes the following fault variables:

[0021] Parallel-side AC overvoltage fault S1, parallel-side AC undervoltage fault S2, parallel-side AC overcurrent fault S3, series-side AC overvoltage fault S4, and series-side AC overcurrent fault S5;

[0022] The valve-level equipment fault V includes the following fault variables:

[0023] DC capacitor overvoltage fault V1, power device bridge arm shoot-through fault V2, power device bridge arm open circuit fault V3, module overheating fault V4, drive circuit fault V5.

[0024] Furthermore, step 2 also includes defining data acquisition classification variables: system-level overall data SD and valve-level data VD.

[0025] The data acquisition variables include the grid-side head current Ih, grid-side tail current Ie, parallel-side AC outlet current Ip, series-side AC outlet current Is, grid-side head voltage Uh, grid-side tail voltage Ue, parallel transformer secondary phase voltage Up2, series transformer secondary phase voltage Us2, parallel-side upper arm current Ipu, parallel-side lower arm current Ipd, series-side upper arm current Isu, series-side lower arm current Isd, parallel-side unit first submodule capacitor voltage Up1, and series-side unit first submodule capacitor voltage Us1.

[0026] In step 3, the voltage and current sample values ​​of each node of the unified power flow controller are obtained at the same time. After filtering, the signal data of the node is represented as discrete sample values ​​f(t1) according to the Nyquist sampling theorem. , ..., where f represents the sampled value of the continuous signal, t represents time, and t1 represents the sampling time of a certain sampling point. This indicates the sampling step size of the connection signal.

[0027] The data structure consisting of f(Ih), f(Ie), f(Ip), f(Is), f(Uh), f(Ue), f(Up2), and f(Us2) is used as the system-level overall data of UPFC, and the data structure consisting of f(Ipu), f(Ipd), f(Isu), f(Isd), f(Up1), and f(Us1) is used as the valve-level data of UPFC.

[0028] In step 4, the equipment fault prediction model includes an input layer, 6 hidden layers and an output layer. The input layer contains 14 neurons, each hidden layer contains 10 neurons, and the output layer contains 11 neurons.

[0029] in For the input layer's first i One input sample, For the output layer's first j One output value, To connect input layer nodes i and hidden layer nodes k The weights, To connect hidden layer nodes k and output layer nodes j The weights; , Hidden layer nodes k The scaling and translation dimensions,

[0030] .

[0031] Preferably, in step 4, the hidden layer uses a wavelet function as the neuron activation function, expressed as:

[0032]

[0033]

[0034] In the formula Describe the wavelet basis functions. Let X represent the derivative of the wavelet basis function, and X be the input of the neuron.

[0035] Furthermore, in step 4, the expression for the output of the equipment failure prediction model is:

[0036]

[0037] In the formula y This is the output of the wavelet neural network. For the sigmoid function, These are Morlet wavelet basis functions. , Hidden layer nodes k The scaling and translation scales are M, where M is the number of hidden layer nodes and N is the number of input layer nodes.

[0038] Preferably, in step 4, each neuron in the output layer of the equipment fault prediction model corresponds to a different fault variable:

[0039]

[0040] In the formula Represents the first output layer j The output value of each neuron SRN This indicates that the unified power flow controller is operating normally.

[0041] Compared with the prior art, the beneficial effects of the present invention include:

[0042] 1) This invention categorizes unified power flow controller (UPFC) faults into two main types: system-wide faults and valve-level equipment faults. These two types are further subdivided into 10 categories of equipment faults. A data acquisition structure for fault prediction is defined, and signal data from UPFC nodes is collected. After filtering and A / D conversion, the data is classified and encoded to establish a dataset for fault prediction. A wavelet neural network is used to construct an equipment fault prediction model. The hidden layer of this equipment fault prediction model contains 10 neurons, which correspond one-to-one with the 10 categories of UPFC equipment faults. This enables multi-dimensional and comprehensive online status monitoring, fault diagnosis, and fault early warning of UPFC, predicting specific fault information of the UPFC device in advance, effectively protecting the safe operation of the device, and avoiding damage to other power electronic components caused by UPFC equipment faults.

[0043] 2) This invention designs a data acquisition structure containing 14 data acquisition variables for system-wide faults and valve-level equipment faults. These variables correspond one-to-one with the 14 neurons in the input layer of the equipment fault prediction model. The data acquisition variables for system-wide faults include the voltage and current at the beginning and end of the grid, the AC output current on the parallel and series sides, and the phase voltage on the secondary side of the parallel and series transformers. The data acquisition variables for valve-level equipment faults include the current of the upper and lower bridge arms on the parallel side, the current of the upper and lower bridge arms on the series side, and the capacitor voltage of the unit submodule on the parallel and series sides. This structure can quickly and accurately locate the fault location, facilitating operation and maintenance personnel to take targeted measures and ensuring the safe and stable operation of the power grid.

[0044] 3) This invention employs a wavelet neural network as the fault prediction model for UPFC equipment. Based on wavelet functions, the wavelet neural network, through the time-frequency localization characteristics of wavelet transform, can more accurately approximate complex nonlinear functional relationships. This results in higher prediction accuracy and precision when handling complex operating conditions and dynamic changes in UPFC equipment fault prediction. The wavelet neural network also exhibits strong fault tolerance and robustness when processing noisy or incomplete data. Furthermore, the wavelet neural network combines the self-learning function of neural networks with the multi-resolution analysis characteristics of wavelet transform, giving the fault prediction model powerful adaptability and learning capabilities. During training, it can automatically adjust the parameters of the wavelet function to better adapt to the actual operating conditions and fault characteristics of the UPFC equipment, thereby improving the accuracy of fault prediction.

[0045] 4) This invention constructs a monitoring system based on a wavelet neural network for predicting UPFC equipment faults. It realizes online monitoring of UPFC equipment, judges the operating status of UPFC, predicts specific fault information of UPFC in advance, and issues fault warning signals in advance for the predicted UPFC equipment faults. This can reduce the damage to other power electronic components caused by equipment faults and improve the safety and stability of power grid operation. Attached Figure Description

[0046] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0047] Figure 1 This is a flowchart illustrating the UPFC equipment fault prediction method according to an embodiment of the present invention.

[0048] Figure 2 This is a system architecture diagram of the UPFC equipment fault prediction system according to an embodiment of the present invention. Detailed Implementation

[0049] like Figure 1 As shown, the UPFC equipment fault prediction method based on wavelet neural network includes the following steps:

[0050] Step 1: Define the fault variables for the unified power flow controller: S represents the overall system fault, and V represents the valve-level equipment fault.

[0051] In this embodiment, the overall system fault S includes the following fault variables:

[0052] Parallel-side AC overvoltage fault S1, parallel-side AC undervoltage fault S2, parallel-side AC overcurrent fault S3, series-side AC overvoltage fault S4, and series-side AC overcurrent fault S5;

[0053] In this embodiment, the valve-level equipment fault V includes the following fault variables:

[0054] DC capacitor overvoltage fault V1, power device bridge arm shoot-through fault V2, power device bridge arm open circuit fault V3, module overheating fault V4, drive circuit fault V5.

[0055] Step 2: Define the data acquisition structure type, classify the acquired data, and quickly locate the fault location and fault type.

[0056] The collected data is defined in two categories: system-level overall data (SD) and valve-level data (VD). The definitions of the data collection variables are shown in Table 1.

[0057] Table 1. Definition of Data Collection Variables

[0058]

[0059] Step 3: Obtain the signal data of each node of the unified power flow controller, and perform filtering, A / D conversion and classification encoding on the signal data in sequence.

[0060] Since AC sampling acquires instantaneous values, the effective values ​​of AC voltage and current are obtained using a discretization algorithm:

[0061] The effective voltage value is:

[0062]

[0063] In the formula u It is a continuous voltage signal. t Indicates time.

[0064] The effective value of the current is:

[0065]

[0066] In the formula i ( t () indicates a continuous current signal.

[0067] If a period of data collection nThe time interval between two adjacent samples is [number] times. U m This represents the instantaneous value of the sampled voltage during the (m-1)th time interval;

[0068] Number of samples within signal period T:

[0069]

[0070] Generally, the time interval between two adjacent samples is taken to be equal, that is... It is a constant;

[0071] The effective value of the voltage can be calculated using the following formula:

[0072]

[0073] Similarly, for alternating current, assuming one cycle of sampling is n times, the instantaneous value of the sampled current in the (m-1)th time interval is I. m The effective value of the current can be calculated using the following formula:

[0074]

[0075] The voltage and current sample values ​​of each node of the unified power flow controller are obtained at the same time. After filtering, the signal data of the node is represented as discrete sample values ​​f(t1) according to the Nyquist sampling theorem. , ..., where f represents the sampled value of the continuous signal, t represents time, and t1 represents the sampling time of a certain sampling point. This indicates the sampling step size of the connection signal.

[0076] The data structure consisting of f(Ih), f(Ie), f(Ip), f(Is), f(Uh), f(Ue), f(Up2), and f(Us2) is used as the system-level overall data of UPFC, and the data structure consisting of f(Ipu), f(Ipd), f(Isu), f(Isd), f(Up1), and f(Us1) is used as the valve-level data of UPFC.

[0077] The signal processing device in this embodiment, which uses a DSP signal processing chip, filters the signal and then isolates and amplifies it before sending it to the A / D conversion and timing control module under the control of a multiplexer. Under the software control of the power frequency phase-locked square wave output circuit, the current and voltage at the same moment within the same cycle are instantaneously acquired. In each relative cycle, the instantaneous values ​​of one channel of three-phase current and three-phase voltage can be acquired. After calculation by the software algorithm, the effective values ​​of the voltage, current and other data of the circuit under test can be obtained.

[0078] Step 4: Construct a UPFC equipment fault prediction model using a wavelet neural network, and train the UPFC equipment fault prediction model using a historical fault signal dataset.

[0079] In this embodiment, the wavelet neural network has 14 neurons in the input layer, 10 neurons in the hidden layer, 6 hidden layers, and 11 neurons in the output layer. Let i be the i-th input sample of the input layer, i = 1, 2, 3, ..., m. Let j be the j-th output value of the output layer, where j = 1, 2, 3, ..., n. The weights connecting input layer node i and hidden layer node k are... The weights connecting the hidden layer node k and the output layer node j; and These represent the scaling and translation scales of the hidden layer node k, respectively.

[0080] Each neuron in the output layer of the wavelet neural network corresponds to a different fault variable:

[0081]

[0082] In the formula Represents the first output layer j The output value of each neuron SRN This indicates that the unified power flow controller is operating normally.

[0083] The hidden layers of a wavelet neural network use wavelet functions as neuron activation functions, expressed as follows:

[0084]

[0085]

[0086] In the formula Describe the wavelet basis functions. Let X represent the derivative of the wavelet basis function, and X be the input of the neuron.

[0087] The output layer function uses the Sigmoid function, whose expression is:

[0088]

[0089] The output of the wavelet neural network can be expressed as:

[0090]

[0091] In the formula For the sigmoid function, Here, M represents the number of hidden layer nodes, and N represents the number of input layer nodes.

[0092] The training process of the wavelet network algorithm, input samples The input samples are fed into each hidden layer through the input layer. Perform a weighted summation, and input the weighted summation result into the wavelet function. The process involves scaling and translation to complete a nonlinear mapping. After wavelet function transformation, the results are input into the next hidden layer for another nonlinear mapping. Finally, the final result data is input into the output layer, where the data is weighted and summed. The output layer function then calculates the network output.

[0093] Step 5: Input the real-time signal data obtained in Step 3 into the trained UPFC equipment fault prediction model to obtain the equipment status prediction result output by the UPFC equipment fault prediction model.

[0094] Step 6: Based on the equipment status prediction results obtained in Step 5, determine whether there is any equipment failure;

[0095] During the scanning cycle, based on the equipment status prediction results of the UPFC equipment fault prediction model, if the operating parameters are normal, the UPFC system is considered to be operating normally; if the UPFC system operating parameters are abnormal, abnormal parameter equipment information is sent to the UPFC maintenance personnel.

[0096] Based on the real-time status prediction results of the equipment output by the UPFC equipment fault prediction model, the fault types of the UPFC device are determined as follows: ① Normal operation; ② Overall system faults: parallel side AC overvoltage fault, parallel side AC undervoltage fault, parallel side AC overcurrent fault, series side AC overvoltage fault, series side AC overcurrent fault; ③ Valve-level equipment faults: DC capacitor overvoltage fault, power device bridge arm shoot-through fault, power device bridge arm open circuit fault, module overheating fault, drive circuit fault.

[0097] Step 7: Based on the fault judgment results of Step 6, issue a fault warning signal for the predicted equipment fault.

[0098] UPFC fault warnings are divided into: UPFC system overall operation fault warnings, which send UPFC system overall parameter fault information to operators within the scanning period if the overall system parameters are faulty, triggering a UPFC disconnection command; and valve-level equipment fault warnings, which send a UPFC disconnection command to operators if valve-level parameters are determined to be faulty within the scanning period.

[0099] Compared with existing UPFC fault early warning methods, the UPFC equipment fault prediction method based on wavelet neural networks in this embodiment has the following unique advantages:

[0100] (1) Obtain the normal and fault operating parameters of each component of the actual power grid system UPFC, analyze and process the actual power grid data, and obtain accurate and reliable data sources.

[0101] (2) Provide early warning of UPFC fault operation status in actual power grid system 10 minutes in advance, predict and judge UPFC operation status in advance, and provide strong guarantee for UPFC to be connected to the power grid for safe and stable operation.

[0102] (3) Considering the actual operating status of UPFC equipment in the power grid system, the collected data will be classified to quickly and accurately locate the fault location. It has the ability to analyze the internal fault types of UPFC in multiple dimensions, providing effective tools and methods for operation and maintenance personnel to judge the fault type of UPFC.

[0103] (4) Real-time collection, analysis, processing and judgment of relevant parameters data of UPFC operation are used to form an online equipment monitoring system, predict the specific fault information of the system device in advance, effectively protect the safe operation of the device, reduce the damage to other power electronic components caused by equipment failure, and have great economic benefits.

[0104] like Figure 2 As shown, the UPFC equipment fault prediction system of the above method includes a data acquisition and processing module, a fault training module, and an online monitoring module.

[0105] The data acquisition and processing module includes an AC / DC signal acquisition device, a filtering device, a signal processing device, and a data storage module. The signal processing device consists of a DSP signal processing chip.

[0106] The AC / DC signal acquisition device is used to acquire AC / DC signals from each node of the UPFC and transmit the acquired signal data to the filtering device. The filtering device performs noise filtering on the acquired AC / DC signals and transmits the filtered signal data to the signal processing device. The signal processing device performs A / D conversion and data classification encoding on the filtered signal and transmits the processed signal data to the data storage module. The data storage module stores the processed information data.

[0107] The fault training module includes a training module and a testing module.

[0108] The data storage module transmits the stored historical power grid signal data to the training module, which then trains and analyzes the fault data. The training module transmits the trained fault judgment model data to the test module, and the signal processing device transmits the processed real-time UPFC signal data to the test module. The test module forms a fault test model, inputs the processed UPFC signal information to the test module, and outputs the test results.

[0109] The online monitoring module includes a status monitoring module, a fault early warning system, and a display module.

[0110] The testing module transmits the test result data to the status monitoring module, which generates an online monitoring status signal for the UPFC. The status monitoring module then transmits the UPFC status signal data to the fault early warning unit, which issues a UPFC equipment fault early warning message. The fault early warning unit transmits the UPFC equipment fault early warning and online status result data to the display module, which displays the UPFC equipment status information.

[0111] In this embodiment, the AC / DC signal acquisition device consists of a current transformer, a voltage transformer, and a pressure transmitter. The pressure transmitter converts the current and voltage of the secondary circuit into a 0-500mV AC signal through the isolation of the secondary current transformer and voltage transformer.

[0112] The AC / DC signal acquisition device acquires the voltage and current sampling values ​​of each node of the UPFC device at the same time. After being filtered by the filtering device, the data is input to the signal processing device for A / D conversion, data classification and encoding, and the processed UPFC device signal is stored in the data storage module.

[0113] The training module consists of a training FPGA chip. The training module reads historical fault signals from the data storage module, analyzes them through software algorithms, constructs a fault detection model for the UPFC device based on a wavelet neural network, and uses the operating status data of the UPFC device for training.

[0114] In this embodiment, the data is obtained from the data storage module at the UPFC failure time t. f Centered on, t f -10min to t f The waveform of the data collected over +10 minutes was identified as a fault waveform. The normal operating time t of the UPFC was then obtained. n Centered on, t n -10min to t n The waveform of the acquired data over a period of 10 minutes was used to randomly select 600 sets of standard data from the historical operation database to form the normal operating set of UPFC.

[0115] The test module consists of a test FPGA chip. The training module transmits the trained UPFC fault detection model to the test module, and the signal processing device transmits the real-time UPFC processing signals to the test module. The test module obtains the output results by analyzing the input signals. .

[0116] The status monitoring module, composed of an FPGA chip, analyzes the status of the UPFC device. The test module inputs its output results to the status monitoring module, which then analyzes and processes the UPFC device status data to form a judgment result.

[0117] The fault early warning unit consists of a signal distribution system, and the display module consists of display software and a display screen. The status monitoring module transmits the processing results to the fault early warning unit. When a UPFC device fault early warning occurs, the fault early warning unit promptly sends information such as abnormal operation and fault to the operation and management personnel, notifying them immediately. At the same time, it sends a command to the UPFC protection device to cut off the faulty UPFC device. The display module receives the UPFC device status analysis data and displays the UPFC device status on the display screen in real time.

[0118] The implementation results show that the UPFC equipment fault prediction system of the present invention realizes online monitoring of UPFC devices, provides early warning of UPFC fault operation status in actual power grid systems 10 minutes in advance, predicts and judges the UPFC operation status in advance, and provides strong guarantee for the safe and stable operation of UPFC connected to the power grid.

Claims

1. A UPFC device fault prediction method based on a wavelet neural network, characterized in that, Includes the following steps: Step 1: Define the fault variables for the unified power flow controller; Define the fault classification variables for the unified power flow controller: overall system fault S and valve-level equipment fault V; The overall system fault S includes the following fault variables: Parallel-side AC overvoltage fault S1, parallel-side AC undervoltage fault S2, parallel-side AC overcurrent fault S3, series-side AC overvoltage fault S4, and series-side AC overcurrent fault S5; The valve-level equipment fault V includes the following fault variables: DC capacitor overvoltage fault V1, power device bridge arm shoot-through fault V2, power device bridge arm open circuit fault V3, module overheating fault V4, drive circuit fault V5; Step 2: Classify the data acquisition variables and define the data acquisition structure for equipment status monitoring and fault prediction based on their classification; Define data acquisition categorical variables: system-level overall data SD and valve-level data VD; The data acquisition variables include the grid-side head current Ih, grid-side tail current Ie, parallel-side AC outlet current Ip, series-side AC outlet current Is, grid-side head voltage Uh, grid-side tail voltage Ue, parallel transformer secondary phase voltage Up2, series transformer secondary phase voltage Us2, parallel-side upper arm current Ipu, parallel-side lower arm current Ipd, series-side upper arm current Isu, series-side lower arm current Isd, parallel-side unit first submodule capacitor voltage Up1, and series-side unit first submodule capacitor voltage Us1. Step 3: Obtain the signal data of each node of the unified power flow controller, and perform filtering, A / D conversion and classification encoding on the signal data in sequence; Step 4: Construct a device fault prediction model using a wavelet neural network, and train the device fault prediction model using a historical fault signal dataset; The equipment fault prediction model includes an input layer, six hidden layers, and an output layer. The input layer contains 14 neurons, each hidden layer contains 10 neurons, and the output layer contains 11 neurons. in For the input layer's first i One input sample, For the output layer's first j One output value, To connect input layer nodes i and hidden layer nodes k The weights, To connect hidden layer nodes k and output layer nodes j The weights; , Hidden layer nodes k The scaling and translation dimensions, ； The hidden layer uses wavelet functions as neuron activation functions, expressed as follows: ； ； In the formula Describe the wavelet basis functions. Let x represent the derivative of the wavelet basis function, where x is the input to the neuron; The expression for the output of the equipment failure prediction model is: ； In the formula y This is the output of the wavelet neural network. To connect input layer nodes i and hidden layer nodes k The weights, To connect hidden layer nodes k and output layer nodes j The weights; For the sigmoid function, For Morlet wavelet basis functions, , Hidden layer nodes k The scaling and translation scales are M, where M is the number of hidden layer nodes and N is the number of input layer nodes; Each neuron in the output layer of the equipment fault prediction model corresponds to a different fault variable: ； In the formula Represents the first output layer j The output value of each neuron SRN This indicates that the unified power flow controller is operating normally; Step 5: Input the real-time signal data obtained in Step 3 into the trained equipment fault prediction model to obtain the equipment status prediction result output by the equipment fault prediction model; Step 6: Based on the equipment status prediction results obtained in Step 5, determine whether there is any equipment failure; Step 7: Based on the fault judgment results of Step 6, issue a fault warning signal for the predicted equipment fault.

2. The UPFC equipment fault prediction method based on wavelet neural network according to claim 1, characterized in that, In step 3, the voltage and current sample values ​​of each node of the unified power flow controller are obtained at the same time. After filtering, the signal data of the node is represented as discrete sample values ​​f(t1) according to the Nyquist sampling theorem. , ..., where f represents the sampled value of the continuous signal, t represents time, and t1 represents the sampling time of a certain sampling point. Indicates the sampling step size of the connection signal; The data structure consisting of f(Ih), f(Ie), f(Ip), f(Is), f(Uh), f(Ue), f(Up2), and f(Us2) is used as the system-level overall data of UPFC, and the data structure consisting of f(Ipu), f(Ipd), f(Isu), f(Isd), f(Up1), and f(Us1) is used as the valve-level data of UPFC.

Images