A method, device, equipment and medium for trend early warning of a converter valve cooling system

Abstract

The application discloses a trend early warning method, device and equipment of a converter valve cooling system and a medium. The method is to use disturbance recording wave data generated by periodic switching of a main pump to extract characteristic parameters representing system operation states to form a training set and a verification set. Then, a new observation vector selects training set data according to a KNN strategy to dynamically construct a process memory matrix and obtains an estimated vector according to multivariate state estimation. The cosine similarity of the estimated vector and the observation vector is calculated. The cosine similarity statistical value of the verification set is used as a warning threshold. The cosine similarity of the observation vector is compared with the warning threshold, and a trend early warning is issued. The application extracts characteristic parameters by using disturbance recording wave data generated by periodic switching of a main pump, solves the problem that the converter valve cooling system cannot perform fluid parameter trend early warning, and is simple and easy to implement.

Description

Technical Field

[0001] This invention relates to a trend early warning method, device, equipment, and medium for a converter valve cooling system, and particularly to a method for trend early warning based on periodic disturbance waveform data, extracting characteristic parameters, and then using multivariate state estimation technology. This invention belongs to the field of converter valve cooling systems. Background Technology

[0002] In high-voltage and ultra-high-voltage direct current (HVDC) transmission projects, converter valves are key equipment for energy conversion. The thyristor elements inside the converter valves generate significant heat during operation, requiring cooling by circulating cooling water. As the cooling water flows through the converter valves, its temperature rises, carrying away the generated heat. This heat exchange occurs with an outdoor air cooler, lowering the cooling water temperature to a suitable range before it flows back to the converter valves, forming a closed-loop internal circulation system for the cooling water. Summary of the Invention

[0003] The purpose of this invention is to provide a trend early warning method, device, equipment, and medium for a converter valve cooling system. It extracts feature parameters by utilizing the disturbance waveform data of the periodic main pump switching, thus solving the problem that the converter valve cooling system cannot perform trend early warning of fluid parameters. It is simple and easy to implement.

[0004] To achieve the above objectives, the solution of the present invention is:

[0005] A trend early warning method for a converter valve cooling system includes the following steps:

[0006] Step 1: Using the historical data of disturbance recordings generated by the periodic switching of the main circulation pump, extract the feature parameters that characterize the healthy operating state of the system under periodic disturbances to form training set and validation set sequence vectors;

[0007] Step 2: Select training set data from the sequence vectors of the validation set according to the KNN strategy, dynamically construct the process memory matrix, obtain the validation estimation vector based on multivariate state evaluation, calculate the validation cosine similarity between the validation estimation vectors, and finally obtain the statistical threshold of the validation set.

[0008] Step 3: After extracting the feature parameters from the new perturbation recording data, a new observation vector is obtained. Similarly, the process memory matrix is ​​dynamically constructed by selecting the training set data according to the KNN strategy, and a new estimation vector is obtained according to the multivariate state estimation. The observation cosine similarity between the new estimation vector and the new observation vector is calculated.

[0009] Step 4: Compare the observed cosine similarity with the statistical threshold of the validation set. If it is less than the threshold, issue a trend warning.

[0010] The aforementioned disturbance recording data is generated by the control and protection device of the converter valve cooling system when the main circulation pump undergoes periodic switching. This includes, but is not limited to, buffer tank pressure, inlet valve pressure, main pump inlet and outlet pressure difference, and main circulation flow.

[0011] The typical sampling accuracy of the above-mentioned disturbance recording data is 500ms, and the typical time before and after triggering the recording is 1min.

[0012] In step 2 above, the training set and validation set cover the normal operating conditions of the valve cooling system under different three-way valve angles and different main pump motor speeds.

[0013] In step 2 above, the training set and validation set cover the normal operating conditions of the valve cooling system under different system reference pressures.

[0014] The KNN strategy described above involves calculating the Euclidean distance between the new observation vector and each vector in the training set, and then selecting the K closest training set vectors to form a dynamic process memory matrix.

[0015] In step 2 above, the formula for calculating the statistical threshold is Av-3σ, where Av and σ are the mean and standard deviation of the cosine similarity of the validation set, respectively.

[0016] A trend early warning device for a converter valve cooling system includes,

[0017] The sequence vector extraction module is used to extract feature parameters that characterize the healthy operating state of the system under periodic disturbances based on the historical data of disturbance recordings generated by the periodic switching of the main circulation pump, and form training and validation set sequence vectors.

[0018] The statistical threshold acquisition module is used to select training set data from the sequence vectors of the validation set provided by the aforementioned sequence vector extraction module according to the KNN strategy, dynamically construct the process memory matrix, obtain the validation estimation vector according to the multivariate state evaluation, calculate the validation cosine similarity between the validation estimation vector and the validation estimation vector respectively, and finally obtain the statistical threshold of the validation set.

[0019] The observation cosine similarity calculation module is used to extract feature parameters from new perturbation recording data to obtain new observation vectors, select training set data according to the KNN strategy, dynamically construct the process memory matrix, derive new estimation vectors based on multivariate state estimation, and calculate the observation cosine similarity between the new estimation vector and the new observation vector; and,

[0020] The comparison and trend warning module is used to compare the observed cosine similarity with the statistical threshold of the validation set. If it is less than the threshold, a trend warning is issued.

[0021] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor; the processor executes the computer program to implement the steps of the trend warning method for the converter valve cooling system as described above.

[0022] A computer-readable storage medium storing a computer program; when executed by a processor, the computer program implements the steps of the trend warning method for the converter valve cooling system as described above.

[0023] After adopting the above scheme, the present invention uses the disturbance recording data generated by the cycle switching of the main pump to extract the feature parameters characterizing the system operating state to form a training set and a validation set respectively. Then, the new observation vector selects the training set data according to the KNN strategy to dynamically construct the process memory matrix and obtain the estimated vector according to the multivariate state estimation. The cosine similarity between the estimated vector and the observation vector is calculated. Then, the cosine similarity statistics of the validation set are used as the warning threshold. The cosine similarity of the observation vector is compared with the warning threshold and a trend warning is issued.

[0024] The beneficial effects of this invention are as follows:

[0025] (1) The trend warning scheme based on periodic disturbance waveform data solves the problem of lack of fluid mechanical trend warning in the converter valve cooling system and makes up for the deficiency of incomplete warning caused by using SCADA data as the single data source for trend warning;

[0026] (2) By extracting the characteristic parameters of the disturbance waveform data, the problems of large waveform data volume, difficult data analysis, and high computational complexity are solved;

[0027] (3) A dynamic process memory matrix is ​​constructed by the KNN strategy, which improves the accuracy of multivariate state estimation compared with a fixed process memory matrix;

[0028] (4) Cosine similarity is used to assess the similarity between the observed vector and the estimated or validation vector, and the statistical mean and standard deviation of the validation set are used as thresholds to reduce the probability of false alarms. Attached Figure Description

[0029] Figure 1 This is a flowchart of the method of the present invention;

[0030] Figure 2 This is a schematic diagram of feature parameter extraction. Detailed Implementation

[0031] The technical solution and beneficial effects of the present invention will be described in detail below with reference to the accompanying drawings.

[0032] like Figure 1As shown, the present invention provides a trend early warning method for a converter valve cooling system, comprising the following steps:

[0033] Step 1: Using the historical data of disturbance recordings generated by the periodic switching of the main circulation pump, extract the feature parameters that characterize the healthy operating state of the system under periodic disturbances to form training set and validation set sequence vectors;

[0034] Step 2: Select training set data from the sequence vectors of the validation set according to the KNN strategy, dynamically construct the process memory matrix, obtain the validation estimation vector based on multivariate state evaluation, calculate the validation cosine similarity between the validation estimation vectors, and finally obtain the statistical threshold of the validation set.

[0035] Step 3: After extracting the feature parameters from the new perturbation recording data, a new observation vector is obtained. Similarly, the process memory matrix is ​​dynamically constructed by selecting the training set data according to the KNN strategy, and a new estimation vector is obtained according to the multivariate state estimation. The observation cosine similarity between the new estimation vector and the new observation vector is calculated.

[0036] Step 4: Compare the observed cosine similarity with the statistical threshold of the validation set. If it is less than the threshold, issue a trend warning.

[0037] The fluid disturbance waveform data involved in this invention is generated by the control and protection device of the converter valve cooling system when the main circulation pump undergoes periodic switching. The data includes fluid parameters such as buffer tank pressure, inlet valve pressure, main pump inlet and outlet pressure difference, and main circulation flow rate.

[0038] The typical sampling accuracy of the disturbance recording data is 500ms, and the typical time before and after triggering the recording is 1min.

[0039] In step 2, the training set and validation set should cover the normal operating conditions of the valve cooling system under different three-way valve angles (0°~90°) and different main pump motor speeds; the training set and validation set should cover the normal operating conditions of the valve cooling system under different system reference pressures (from the air replenishment value of the buffer tank inlet solenoid valve to the air exhaust stop value of the buffer tank outlet solenoid valve).

[0040] The KNN strategy involved in steps 2 and 3 is as follows: calculate the Euclidean distance between the new observation vector and each vector in the training set, and select the K closest training set vectors to form a dynamic process memory matrix.

[0041] In step 2, the statistical threshold is calculated using the formula Av-3σ, where Av and σ are the mean and standard deviation of the cosine similarity of the validation set, respectively.

[0042] The present invention also provides a trend early warning device for a converter valve cooling system, comprising:

[0043] The sequence vector extraction module is used to extract feature parameters that characterize the healthy operating state of the system under periodic disturbances based on the historical data of disturbance recordings generated by the periodic switching of the main circulation pump, and form training and validation set sequence vectors.

[0044] The statistical threshold acquisition module is used to select training set data from the sequence vectors of the validation set provided by the aforementioned sequence vector extraction module according to the KNN strategy, dynamically construct the process memory matrix, obtain the validation estimation vector according to the multivariate state evaluation, calculate the validation cosine similarity between the validation estimation vector and the validation estimation vector respectively, and finally obtain the statistical threshold of the validation set.

[0045] The observation cosine similarity calculation module is used to extract feature parameters from new perturbation recording data to obtain new observation vectors, select training set data according to the KNN strategy, dynamically construct the process memory matrix, derive new estimation vectors based on multivariate state estimation, and calculate the observation cosine similarity between the new estimation vector and the new observation vector; and,

[0046] The comparison and trend warning module is used to compare the observed cosine similarity with the statistical threshold of the validation set. If it is less than the threshold, a trend warning is issued.

[0047] Example

[0048] See Figure 1 In one embodiment, the main pump switching cycle of a converter station's valve cooling system is set to 10080 minutes, meaning it switches once a week, triggering disturbance waveform data each time it switches. Over a full year, excluding the impact of the valve cooling system's shutdown during annual inspections, a total of 48 valid historical disturbance waveform data points were generated. First, the 48 historical disturbance waveform data samples were normalized so that each parameter in the actual waveform data was mapped to the [0,1] interval.

[0049] like Figure 2 As shown, characteristic parameters were extracted from four important parameters: inlet valve pressure, main pump differential pressure, main circulation flow rate, and buffer tank pressure, resulting in a total of 13 characteristic parameters, among which:

[0050] x1 represents the lowest value of the inlet valve pressure drop during the disturbance process;

[0051] x2 represents the time from when the inlet valve pressure drops to a lower setpoint during the disturbance process until it first recovers to the lower setpoint.

[0052] x3 represents the time from when the inlet valve pressure drops to the ultra-low setpoint during the disturbance process until it finally recovers to the ultra-low setpoint (avoiding the false recovery caused by the pressure spike generated by the check valve action);

[0053] x4 represents the peak value of the inlet valve pressure generated during the disturbance recovery process due to the action of the check valve;

[0054] x5 represents the lowest value of the main pump differential pressure drop during the disturbance process;

[0055] x6 represents the time from when the main pump differential pressure drops to a lower setpoint during the disturbance to when it first recovers to the lower setpoint.

[0056] x7 represents the time from when the main pump drops to the ultra-low setpoint during the disturbance process until it recovers to the ultra-low setpoint for the last time (avoiding the false recovery caused by the pressure spike generated by the check valve action);

[0057] x8 represents the extreme value of the main pump differential pressure peak caused by the check valve action during the disturbance recovery process;

[0058] x9 represents the lowest value of the main circulation flow drop during the disturbance process;

[0059] x 10 This indicates the time from when the main circulation flow drops to a lower setpoint during the disturbance process until it first recovers to the lower setpoint.

[0060] x 11 This indicates the time from when the main circulation flow drops to the ultra-low setpoint during the disturbance process until it first recovers to the ultra-low setpoint;

[0061] x 12 This indicates the time from the start of the disturbance (i.e., the start of main pump switching or the triggering of waveform recording) to the main circulation flow rate dropping to a lower setpoint.

[0062] x 13 This represents the average value of the buffer tank pressure (the reference pressure of the converter valve cooling system) during the entire disturbance recording period.

[0063] Each disturbance waveform generates a vector, namely:

[0064] x(i)=[x1(i)x1(i)…x 13 (i)] T

[0065] The first 30 perturbation recordings can be selected as the training set, denoted as T, and the remaining perturbation recordings can be used as the validation set K.

[0066]

[0067] Similarly, a new observation vector x is generated from the new perturbation recording data, that is:

[0068] x = [x1 x2 …x 13 ] T

[0069] According to the KNN algorithm, the Euclidean distance between the new observation vector x and each vector in the training set is first calculated, i.e.:

[0070]

[0071] Where i = 1, 2, 3, 4…30

[0072] Sort the new observation vector x by Euclidean distance to each vector in the training set, and select the k training set vectors with the smallest distance (nearest neighbor) as the process memory matrix D. Let k = 7, then:

[0073]

[0074] Each new observation vector x can be used to obtain the process matrix D, which is the dynamic process memory matrix.

[0075] Based on the multivariate state estimation method, a new estimation vector x can be obtained. est :

[0076]

[0077] Here are non-linear operators used to replace multiplication operations in ordinary matrix operations.

[0078] To compare the similarity between the new estimated vector and the new observed vector, cosine similarity is calculated:

[0079]

[0080] Similarly, for the 18 validation vectors in the validation set, the cosine similarity between them and the validation estimation vectors is calculated, and their mean Av and standard deviation σ are calculated. Av-3σ is used as the statistical threshold for the cosine similarity of the validation set.

[0081] If satisfied

[0082] cosθ<Αv-3σ

[0083] This indicates that the operating status of the converter valve cooling system has deviated from the normal operating range, and a trend warning signal can be issued to remind maintenance personnel to check the equipment status.

[0084] This invention also provides another trend warning device for a converter valve cooling system, the device including a processor and a memory configured to store a computer program capable of running on the processor; wherein, when the processor is configured to run the computer program, it executes the method steps in the foregoing embodiments.

[0085] Of course, in practical applications, the various components of this data exchange device are coupled together through a bus system. The bus system is used to enable communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus, and a status signal bus.

[0086] In practical applications, the aforementioned processor includes a Field-Programmable Gate Array (FPGA), and the processor can be a Central Processing Unit (CPU) or a Digital Signal Processor (DSP). It is understood that for different devices, the electronic devices used to implement the functions of the aforementioned processor can also be other types, and this embodiment of the invention does not impose specific limitations.

[0087] The aforementioned memory can be volatile memory, such as random-access memory (RAM); or non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); or a combination of the above types of memory, and provides instructions and data to the processor.

[0088] In an exemplary embodiment, the present invention also provides a computer-readable storage medium for storing a computer program.

[0089] Optionally, the computer-readable storage medium can be applied to any of the methods in the embodiments of the present invention, and the computer program causes the computer to execute the corresponding processes implemented by the processor in the various methods of the embodiments of the present invention. For the sake of brevity, these will not be described in detail here.

[0090] In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0091] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solutions based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A trend early warning method for a converter valve cooling system, characterized in that... Includes the following steps: Step 1: Using the historical data of disturbance recordings generated by the periodic switching of the main circulation pump, extract the feature parameters that characterize the healthy operating state of the system under periodic disturbances to form training set and validation set sequence vectors; Step 2: Select training set data from the sequence vectors of the validation set according to the KNN strategy, dynamically construct the process memory matrix, obtain the validation estimation vector based on multivariate state evaluation, calculate the validation cosine similarity between the validation estimation vectors, and finally obtain the statistical threshold of the validation set. Step 3: After extracting the feature parameters from the new perturbation recording data, a new observation vector is obtained. Similarly, the process memory matrix is ​​dynamically constructed by selecting the training set data according to the KNN strategy, and a new estimation vector is obtained according to the multivariate state estimation. The observation cosine similarity between the new estimation vector and the new observation vector is calculated. Step 4: Compare the observed cosine similarity with the statistical threshold of the validation set. If it is less than the threshold, issue a trend warning. The disturbance recording data is generated by the control and protection device of the converter valve cooling system when the main circulation pump undergoes periodic switching. It includes, but is not limited to, buffer tank pressure, inlet valve pressure, main pump inlet and outlet pressure difference, and main circulation flow.

2. The method as described in claim 1, characterized in that: The typical sampling accuracy of the disturbance recording data is 500ms, and the typical time before and after triggering the recording is 1min.

3. The method as described in claim 1, characterized in that: In step 2, the training set and validation set cover the normal operating conditions of the valve cooling system under different three-way valve angles and different main pump motor speeds.

4. The method as described in claim 1, characterized in that: In step 2, the training set and validation set cover the normal operating conditions of the valve cooling system under different system reference pressures.

5. The method as described in claim 1, characterized in that: The KNN strategy involves calculating the Euclidean distance between the new observation vector and each vector in the training set, and then selecting the K closest training set vectors to form a dynamic process memory matrix.

6. The method as described in claim 1, characterized in that: In step 2, the statistical threshold is calculated using the formula Av-3σ, where Av and σ are the mean and standard deviation of the cosine similarity of the validation set, respectively.

7. A trend early warning device for a converter valve cooling system, characterized in that: include, The sequence vector extraction module is used to extract feature parameters that characterize the healthy operating state of the system under periodic disturbances based on the historical data of disturbance recordings generated by the periodic switching of the main circulation pump, and form training and validation set sequence vectors. The statistical threshold acquisition module is used to select training set data from the sequence vectors of the validation set provided by the aforementioned sequence vector extraction module according to the KNN strategy, dynamically construct the process memory matrix, obtain the validation estimation vector according to the multivariate state evaluation, calculate the validation cosine similarity between the validation estimation vector and the validation estimation vector respectively, and finally obtain the statistical threshold of the validation set. The observation cosine similarity calculation module is used to extract feature parameters from new disturbance recording data to obtain new observation vectors, select training set data according to the KNN strategy, dynamically construct the process memory matrix, obtain new estimation vectors based on multivariate state estimation, and calculate the observation cosine similarity between the new estimation vector and the new observation vector. as well as, The comparison and trend warning module is used to compare the observed cosine similarity with the statistical threshold of the validation set. If it is less than the threshold, a trend warning is issued. The disturbance recording data is generated by the control and protection device of the converter valve cooling system when the main circulation pump undergoes periodic switching. It includes, but is not limited to, buffer tank pressure, inlet valve pressure, main pump inlet and outlet pressure difference, and main circulation flow.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor; characterized in that: When the processor executes the computer program, it implements the steps of the trend warning method for the converter valve cooling system as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program; characterized in that: When the computer program is executed by the processor, it implements the steps of the trend warning method for the converter valve cooling system as described in any one of claims 1 to 6.

Images