An oil-immersed transformer oil chromatographic monitoring point intelligent optimization layout method based on gas distribution simulation

Abstract

The application discloses an oil-immersed transformer oil chromatographic monitoring point intelligent optimization layout method based on gas distribution simulation, relates to the technical field of transformer optimization design, and first fuses multiple load rates and multiple fault gas production conditions to construct complete gas three-dimensional distribution data sets, so that the spatial difference of gas concentration caused by the non-uniformity of the oil flow field inside the transformer can be quantified comprehensively. The candidate space is limited to the layout area which can obtain the fault characteristic gas concentration information inside the transformer and meet the requirements of insulation distance, sampling or detection. A hierarchical priority quantitative evaluation criterion is adopted, multi-dimensional evaluation of layout point representation ability is performed, and the common law of gas distribution under multiple conditions is fully tapped. A particle swarm optimization algorithm is used to carry out global intelligent optimization, so as to make up for the short board that the existing gas monitoring point layout cannot comprehensively reflect the spatial distribution of the gas in the transformer. Finally, the verified optimal monitoring point three-dimensional coordinates are output, which are used for the structural design in the production stage of the transformer and the structural modification in the operation repair stage of the transformer.

Description

Technical Field

[0001] This application relates to the field of transformer optimization design technology, specifically to an intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation. Background Technology

[0002] Oil chromatography is a crucial method for capturing early fault characteristic gases inside oil-immersed transformers, widely used in transformer insulation condition assessment and fault early warning. With the development of online monitoring technology, in addition to oil chromatography, spectral detection, photoacoustic detection, and other gas detection methods are increasingly being applied to fault characteristic gas monitoring in oil-immersed transformers. Although different detection methods differ in their detection principles and device forms, they all require obtaining the concentration information of fault characteristic gases at specific locations within the transformer. Monitoring points can be the sampling locations of oil or gases, or the installation locations of gas sensors, optical probes, or other detection units. Therefore, the deployment location of monitoring points is a critical design parameter during transformer manufacturing and operational maintenance.

[0003] Currently, the placement of transformer oil chromatography monitoring points is mainly based on experience or simple rules, typically placing sampling points on the side walls of the tank or in cooling pipes—locations easily accessible for sampling. However, the internal structure of oil-immersed transformers is complex and their volume is large. Forced oil circulation and thermal convection result in a significantly non-uniform oil flow field, leading to highly uneven diffusion and distribution of fault-specific gases within the oil. Existing experience-based fixed monitoring point placement methods cannot accurately reflect the three-dimensional spatial distribution characteristics of gases within the transformer, exhibiting an inherent limitation that local monitoring results cannot characterize the overall condition.

[0004] Existing technologies have conducted research on the determination of monitoring points for dissolved gases in transformer oil, analysis of monitoring data, and reliability of monitoring devices. Chinese patent application CN117309687A discloses a method for determining monitoring points for transformer fault information. This method considers the influence of fault gas diffusion patterns on the selection of monitoring points, but it mainly relies on the diffusion path and diffusion time between preset fault points and candidate monitoring points for point selection. Chinese patent application CN121899317A discloses a method for visual analysis of dynamic thresholds in transformer oil chromatography. This method mainly focuses on threshold adjustment, anomaly identification, and early warning analysis of monitoring data, but does not address the optimization of the physical layout of oil chromatography monitoring points inside the transformer or in the oil circuit system.

[0005] In summary, the current bottlenecks in the layout technology of oil chromatography monitoring points for oil-immersed transformers are mainly reflected in two aspects: First, existing layout schemes rely on experience or simple statistics, lacking a quantitative understanding of the spatial distribution patterns of gases. Second, existing simulation studies mostly focus on phenomenon revelation or distribution prediction, failing to integrate with intelligent optimization algorithms to form a systematic layout optimization method. Therefore, how to achieve optimized layout design of oil chromatography monitoring points for oil-immersed transformers is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] In order to solve the above-mentioned technical problems, this application proposes the following technical solution: In a first aspect, embodiments of this application provide an intelligent optimization layout method for oil chromatography monitoring points in oil-immersed transformers based on gas distribution simulation, including: A three-dimensional spatial distribution dataset of fault characteristic gases inside a transformer under multiple operating conditions is obtained. The multiple operating conditions include various load rates and multiple typical fault gas generation locations. The three-dimensional spatial distribution dataset includes the fault characteristic gas concentration values ​​at each spatial location in the transformer's internal oil calculation domain under each operating condition. A candidate space for monitoring points is defined and discretized into a set of candidate points. The candidate space includes a deployable area inside the transformer where fault characteristic gas concentration information can be obtained, and which meets the requirements for insulation distance, sampling, or detection. The deployable area includes the inner area of ​​the top cover and side wall of the oil tank, the oil space near the core and windings, and the area adjacent to the usable interface connected to the oil circuit, excluding areas occupied by solid structures, structural interference areas, and other prohibited areas. The set of candidate points consists of three-dimensional coordinate points generated by spatial discretizing the deployable area according to a preset selection interval. A quantitative evaluation criterion for the monitoring point scheme is constructed, and the evaluation criterion adopts a priority-based hierarchical screening rule; The optimal monitoring point layout scheme is searched using the particle swarm optimization algorithm, and the obtained optimal monitoring point layout scheme is verified. The three-dimensional coordinates of each monitoring point in the optimal monitoring point layout scheme are output.

[0007] In one possible implementation, acquiring the three-dimensional spatial distribution dataset of fault characteristic gases inside the transformer under multiple operating conditions includes: The target oil-immersed transformer was selected, and a three-dimensional simulation model of the transformer was built based on the actual structure and size parameters of the equipment. The model retains key structures that affect the oil flow field and gas transport, such as windings, core, oil pipes, and cooler interfaces, while simplifies structures that have less impact on the oil flow field and fault characteristic gas transport, such as fastening components. Configure multiple typical operating conditions and aggregate the three-dimensional spatial distribution data of fault characteristic gases corresponding to all operating conditions to form a three-dimensional spatial distribution dataset.

[0008] In one possible implementation, the configuration of multiple typical operating conditions and the aggregation of all operating condition data into a three-dimensional spatially distributed dataset includes: Set the load rate level to The number of faulty gas-producing locations is A typical operating condition is determined by the pairwise combination of load factor and fault gas generation location. Therefore, the total number of typical operating conditions is... The product of the number of load factor ranges and the number of fault gas generation locations: in: Includes various operating conditions such as no-load, light-load, half-load, and heavy-load. Distributed in windings, cores, leads, bushings, and other typical fault locations; The constant gas production rate of the fault characteristic gas is set as follows: The simulation duration is set to Based on the established three-dimensional simulation model of the transformer, the transport process of fault characteristic gas under various operating conditions is solved, and the concentration of fault characteristic gas at each grid node in the computational domain of the transformer internal oil is obtained in steady state. Let the total number of grid nodes in the oil computational domain be . , No. The three-dimensional coordinates of each grid node are ,in, ; The three-dimensional spatial distribution data corresponding to all working conditions are aggregated to form a dataset: in: For the first The data subset corresponding to each working condition Indicates the first The first working condition Fault characteristic gas concentration at each grid node This is the transformer operating condition sequence number, and its value range is: .

[0009] In one possible implementation, defining the candidate space of monitoring points and discretizing it into a set of candidate points includes: The transformer is divided into a deployable area and a prohibited area. The deployable area includes the area inside the transformer where fault characteristic gas concentration information can be obtained and where insulation distance, sampling or detection requirements are met. This includes the area inside the top cover and side walls of the tank, the oil space near the core and windings, and the area adjacent to available interfaces connected to the oil circuit.

[0010] The restricted areas include the bushing riser installation interference area, the cooler pipe connection interference area, the unsuitable arrangement area corresponding to the external support shielding, the area occupied by solid structures such as iron core and winding, and other unsuitable arrangement areas determined according to the insulation safety, sampling, detection and maintenance requirements of the monitoring device. No candidate measurement points are generated within the restricted areas.

[0011] The monitoring points described in this application are used to characterize the location of fault characteristic gas concentration information, and do not limit the main body of the detection device to be installed at the location of the monitoring point. For monitoring points located in the oil space inside the transformer, the fault characteristic gas information at the corresponding location can be obtained through sampling channels, detection channels, optical detection channels, or sensor probes connected to that location; the main body of the detection device can be set inside or outside the oil tank according to the actual installation conditions.

[0012] The deployable region is discretized according to a preset spacing. Uniform discretization generates a set of candidate measurement points: in: The total number of candidate points, each candidate point has unique three-dimensional spatial coordinates; The number of monitoring points to be optimized is set as follows: ,in Monitoring point plan Represented as from the candidate point set Selected from The combination of points, namely: in, Indicating the monitoring point scheme The first in One monitoring point, Its three-dimensional spatial coordinates are as follows , .

[0013] For any two monitoring points and The Euclidean distance between the two is: The monitoring point scheme satisfies the constraint that the Euclidean distance between any two monitoring points is not less than a preset minimum spacing. .

[0014] In one possible implementation, the quantitative evaluation criteria for the monitoring point construction scheme, wherein the evaluation criteria adopt a priority-based hierarchical screening rule, include: Two evaluation metrics were constructed: threshold trigger probability and minimum dilution ratio of the scheme. Number of monitoring points Under the same conditions, the merits of the schemes are determined based on the threshold trigger probability and the minimum dilution ratio of the scheme.

[0015] Among them, the first priority is to compare the threshold trigger probability of each monitoring point scheme, and the monitoring point scheme with a higher threshold trigger probability is better than the monitoring point scheme with a lower threshold trigger probability. When the threshold trigger probabilities of two monitoring point schemes are the same, the second priority compares the minimum dilution ratio of the schemes, and the monitoring point scheme with the larger minimum dilution ratio is superior to the monitoring point scheme with the smaller minimum dilution ratio.

[0016] In one possible implementation, the construction of the two evaluation metrics, threshold trigger probability and minimum dilution ratio, includes: Let the preset alarm threshold be For the first For various working conditions, when the monitoring point scheme At least one monitoring point corresponds to a fault characteristic gas concentration greater than or equal to the preset alarm threshold. At that time, determine the monitoring point scheme. In the The alarm was successfully triggered under the following conditions; Statistical monitoring point scheme In all Number of working conditions that successfully triggered alarms in various working conditions Construct the threshold trigger probability: in, The range of values ​​is The larger the value, the stronger the ability of the monitoring point scheme to successfully detect fault characteristic gases and trigger alarms under multiple operating conditions.

[0017] For the Fault origin under various working conditions and plan The first in monitoring points Construct a single-point dilution ratio for: in: For monitoring points In the Characteristic gas concentration under certain operating conditions The source of the fault Characteristic gas concentration in situ; The scheme is determined based on the single-point dilution ratio. The minimum dilution ratio for this scheme is: in, Monitoring point scheme The monitoring point index in the middle, ; For operating condition index, ; Indicates the first Solution under various working conditions The Middle The single-point dilution ratio corresponding to each monitoring point. This represents the minimum single-point dilution ratio of the scheme under all operating conditions and all monitoring point combinations. It is used to characterize the degree of fidelity of the scheme in terms of the concentration of fault characteristic gas under the most unfavorable conditions. The larger the value, the smaller the gas dilution under the most unfavorable operating conditions.

[0018] In one possible implementation, a particle swarm optimization algorithm is used to search for an optimal monitoring point layout scheme. Then, the obtained optimal monitoring point layout scheme is determined and verified, and the three-dimensional coordinates of each monitoring point in the optimal layout scheme are output, including: Each particle represents a scheme containing a preset number of monitoring points. By iteratively updating the position and velocity of the particles and applying monitoring point spacing constraints and no-displacement zone constraints, the scheme is optimized generation by generation until convergence, thus obtaining the globally optimal monitoring point scheme. The obtained global optimal monitoring point scheme is compared with the preset comparison point layout scheme, the performance evaluation index is calculated, and the three-dimensional coordinates of each monitoring point in the global optimal monitoring point scheme are output.

[0019] In one possible implementation, each particle represents a scheme containing a preset number of monitoring points. The position and velocity of the particles are iteratively updated, and monitoring point spacing constraints and forbidden zone constraints are applied. The process is optimized generation by generation until convergence, obtaining the globally optimal monitoring point scheme. This includes: Set the population size as , A positive integer, the position vector of each particle. Indicates a set containing The plan includes monitoring points, totaling [number] points. Dimension, that is: In the candidate point set Random selection Each particle has an initial position consisting of _ points, and the distance between any two points is not less than _ points. Initialize the velocity vector of each particle. It is a zero vector; For each particle, its position vector is... Decoded as The coordinates of each monitoring point are calculated, and the threshold trigger probability of the monitoring point scheme corresponding to the particle is determined. and minimum dilution ratio of the scheme .in: This is used to characterize the proportion of operating conditions in which the monitoring point scheme successfully triggers an alarm across all operating conditions. This is used to characterize the fidelity of the monitoring point scheme in terms of the concentration of fault characteristic gases under the most unfavorable conditions across all operating modes.

[0020] According to priority comparison ,exist Compare only when they are the same Priority-based stratified selection rules determine the ranking of each particle in the population; Update the individual historical best position of each particle. and the global optimal position of the population ; According to the above and Update the velocity and position of each particle; After each position update, the coordinates of each particle are checked. If the coordinates of any monitoring point fall into the forbidden region or are not in the candidate point set, the check is performed. If the Euclidean distance between any two monitoring points is less than 1, then map it to the nearest valid candidate point; if the Euclidean distance between any two monitoring points is less than 1, then map it to the nearest valid candidate point. Then, the coordinates are adjusted by random perturbation until the constraint conditions are met; Two iteration termination conditions are set: first, the number of iterations reaches the maximum number of iterations. Second, the global optimal position of the population. The priority ranking of the corresponding monitoring point schemes has not improved for 20 consecutive generations. When any of the above conditions are met, the iteration terminates, and the current globally optimal position is output. Corresponding monitoring point scheme .

[0021] In one possible implementation, the statement based on the and The velocity and position of each particle are updated, including: In the formula: and Particles In the The position vector and velocity vector of the step; As an inertial weight, it is preset to decrease linearly from a larger value to a smaller value; and As a learning factor, and for Independent random numbers within the interval.

[0022] In one possible implementation, the step of comparing the obtained globally optimal monitoring point scheme with a preset comparative point layout scheme, calculating performance evaluation indicators, and outputting the three-dimensional coordinates of each monitoring point in the globally optimal monitoring point scheme includes: Selecting the globally optimal monitoring point scheme Pre-set comparison point layout scheme with the same number of monitoring points The globally optimal monitoring point scheme Comparison with the preset point layout scheme Substitute into the above Calculate the threshold trigger probability for each scheme from the three-dimensional spatial distribution dataset corresponding to the various working conditions. and minimum dilution ratio of the scheme ; Calculate the The increase in trigger probability and the increase in minimum dilution ratio of the solution: Output the globally optimal monitoring point scheme Description of the three-dimensional coordinates and location of each monitoring point.

[0023] In this embodiment, a three-dimensional spatial distribution dataset of fault characteristic gases is first constructed by integrating multiple load rates and multiple fault gas-generating conditions, thereby quantifying the concentration differences of fault characteristic gases caused by the spatial non-uniformity of the oil flow field inside the transformer. The candidate space of monitoring points covers the deployable area inside the transformer where fault characteristic gas concentration information can be obtained, excluding areas occupied by solid structures, structural interference areas, and other unsuitable areas, thus balancing the engineering feasibility of monitoring points and their coverage of the spatial distribution of fault characteristic gases. By discretizing the deployable area to form a candidate point set with unified three-dimensional spatial coordinates, a hierarchical evaluation criterion is adopted, with threshold trigger probability as the first priority and minimum dilution ratio of the scheme as the second priority, to comprehensively evaluate the multi-condition fault detection capability of different monitoring point layout schemes. On this basis, a particle swarm optimization algorithm is used to globally optimize the combination of monitoring points, and the optimization results are verified using multi-condition gas distribution data. Finally, the three-dimensional coordinates and position description of the optimal monitoring point are output, realizing the transformation of fault characteristic gas monitoring point layout from empirical layout to intelligent optimization layout based on gas transport mechanism and multi-condition data. This can be used for structural design in the transformer production stage and structural modification in the operation and maintenance stage. Attached Figure Description

[0024] Figure 1A flowchart illustrating an intelligent optimization layout method for oil chromatography monitoring points in an oil-immersed transformer based on gas distribution simulation, provided for an embodiment of this application. Figure 2 This is a schematic diagram of a three-dimensional model of a converter transformer provided in an embodiment of this application; Figure 3 A schematic diagram of the three-dimensional spatial distribution of the four monitoring point layout schemes provided in the embodiments of this application inside the transformer; Figure 4 A comparison chart of threshold trigger probabilities for four monitoring point layout schemes provided in the embodiments of this application; Figure 5 The first operating condition (load rate 0, gas production point 5) provided in this application embodiment is used to connect the gas production point to each monitoring point. Component mass fraction variation curve; Figure 6 Under the second operating condition (load rate 25%, gas production point 9) provided in the embodiments of this application, the distance from the gas production point to each monitoring point is as follows: Component mass fraction variation curve; Figure 7 The third operating condition (load rate 55%, gas production point 3) provided in the embodiments of this application, from the gas production point to each monitoring point Component mass fraction variation curve; Figure 8 The fourth operating condition (load rate 85%, gas production point 10) provided in this application embodiment, from the gas production point to each monitoring point Component mass fraction variation curve. Detailed Implementation

[0025] The present solution will now be described in conjunction with the accompanying drawings and specific embodiments.

[0026] See Figure 1 The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation provided in this embodiment includes: S101, Obtain the three-dimensional spatial distribution dataset of characteristic gases in transformer internal faults under multiple operating conditions.

[0027] This embodiment uses an oil-immersed converter transformer as an example to detail the specific implementation process of the intelligent optimization layout method for monitoring gas fault characteristics of oil-immersed transformers based on gas distribution simulation proposed in this application. The specific model, rated capacity, and structural dimensions of the converter transformer can be determined according to the actual application, and this embodiment does not limit this. Based on the selected target transformer, a three-dimensional simulation model of the transformer is built according to its actual structural characteristics and dimensional parameters. In this embodiment, a three-dimensional model of the transformer's internal structure is constructed using three-dimensional modeling software. The model retains structures that affect the oil flow field and gas transport, such as windings, core, oil pipes, and cooler interfaces, while simplifying structures with less impact on the oil flow field and fault characteristic gas transport, such as fastening components. Figure 2 This is a schematic diagram of the three-dimensional model of the converter transformer in this embodiment.

[0028] It can be configured with multiple typical operating conditions. This embodiment sets the number of load rate thresholds. The value is 4, corresponding to load rates of 0%, 25%, 55%, and 85% respectively; let the number of fault gas-generating locations be... The fault gas-generating locations are 10, including the upper clamp of the winding of column 2, the lead wire of the valve-side winding, the bottom end of the valve-side bushing, the upper clamp of the winding of column 1, the connecting wire of the valve-side winding, the bottom of the screen of column 2, the connecting wire of the mesh-side winding, the outside of the upper coil of column 1, the inside of the upper coil of column 1, and the outside of the upper coil of column 2. Combining each load rate with each fault gas-generating location in pairs forms 40 typical operating conditions, calculated using the following formula: It should be noted that the number of load factor taps, load factor values, number of fault gas generation locations, and specific settings of fault gas generation locations can be adjusted according to the structural form of the target transformer, historical fault types, and monitoring requirements, and are not limited to the operating conditions listed in this embodiment.

[0029] This embodiment selects acetylene as the target fault characteristic gas for illustrative purposes. Acetylene can be used to characterize internal discharge faults in transformers. In other embodiments, hydrogen, methane, ethylene, ethane, carbon monoxide, carbon dioxide, or one or more other fault characteristic gases that can reflect the internal fault state of an oil-immersed transformer can also be selected.

[0030] Set acetylene to a constant gas production rate 440 Continuous gas production, simulation duration Set to 60 The transport process of acetylene in the transformer internal oil was solved based on the component transport equation, and the acetylene concentration at each grid node in the computational domain of the oil was obtained at steady state. The concentration units were unified as follows: In this embodiment, the total number of grid nodes in the oil computational domain is... Approximately 4.3 million, the The three-dimensional coordinates of each grid node are represented as follows: .

[0031] All operating condition data are aggregated into a dataset, denoted as . Each group The expression is: in, The expression represents the first All working conditions The three-dimensional coordinates of each grid node and its corresponding characteristic gas concentration value. Let be the characteristic gas concentration corresponding to the g-th node under the j-th operating condition.

[0032] S102, define the candidate space of monitoring points and discretize it into a set of candidate points.

[0033] In this embodiment, the center point of the transformer tank bottom surface is used as the origin of the three-dimensional coordinate system. The candidate space for monitoring points includes areas inside the transformer where fault characteristic gas concentration information can be obtained, and which meet the insulation distance, sampling, or detection requirements of the monitoring device.

[0034] Specifically, the deployable areas include the inner area of ​​the tank top cover and side walls, the oil space near the core and windings, and the area adjacent to available interfaces connected to the oil circuit. The prohibited deployment areas include the interference zone for the bushing riser installation, the interference zone for cooler pipe connections, unsuitable deployment areas corresponding to external support obstructions, areas occupied by solid structures such as the core and windings, and other unsuitable deployment areas determined according to the installation, sampling, detection, and maintenance requirements of the monitoring device. No candidate measurement points are generated within the prohibited deployment areas.

[0035] In this embodiment, the candidate monitoring points are used to characterize the location for obtaining fault characteristic gas concentration information, and it is not limited that the main body of the detection device must be set at the corresponding three-dimensional coordinate position. For candidate monitoring points located in the oil space inside the transformer, sampling pipelines, detection channels, optical probes, gas sensing probes, or other structures capable of obtaining fault characteristic gas information at the corresponding location can be used to achieve monitoring.

[0036] The preset discretization interval is The deployable region is spatially discretized according to the preset discretization interval to generate a set of candidate monitoring points: in, This represents the total number of candidate monitoring points, with each candidate monitoring point having unique three-dimensional spatial coordinates. .

[0037] The number of monitoring points to be optimized is set as follows: This embodiment will carry out the following: Equal to 4 and There are two sets of optimization calculations, equal to 7, corresponding to the final optimal 4-point scheme and the optimal 7-point scheme, respectively. One set of monitoring point schemes. Represented as from the candidate point set Selected from The combination of points, i.e. ,in, Indicating the monitoring point scheme The first in One monitoring point, Its three-dimensional spatial coordinates are as follows , .

[0038] Define any two points and The Euclidean distance between them is: The monitoring point scheme satisfies the constraint that the Euclidean distance between any two monitoring points is not less than a preset minimum spacing. In this embodiment Set it to 500mm.

[0039] S103, Construct quantitative evaluation criteria for monitoring point schemes, wherein the evaluation criteria adopt a priority-based hierarchical screening rule.

[0040] Two evaluation metrics were constructed: threshold trigger probability and minimum dilution ratio of the scheme. Let the preset alarm threshold be For the first For various working conditions, when the monitoring point scheme At least one monitoring point corresponds to a fault characteristic gas concentration greater than or equal to the preset alarm threshold. At that time, determine the monitoring point scheme. In the The alarm was successfully triggered under the following conditions; the unit of the fault characteristic gas concentration at the monitoring point is... ; Statistical monitoring point scheme In all Number of working conditions that successfully triggered alarms in various working conditions And construct the threshold trigger probability: in: This represents the total number of operating conditions. For the plan The number of operating conditions that successfully triggered the alarm. The range of values ​​is The higher the value, the stronger the overall monitoring reliability of the scheme.

[0041] For detection devices that directly output gas concentration, the gas concentration can be directly compared with the alarm threshold. For devices that output absorbance, transmitted light intensity, acoustic response, electrical signals, or other detection signals, the detection signals can be converted into the concentration or equivalent concentration characterization value of the target fault characteristic gas through a pre-established calibration relationship.

[0042] In this embodiment Set the alarm threshold to 40. It is 0.3 For a set of solutions Count how many of the 40 operating conditions have at least one monitoring point with an acetylene concentration greater than or equal to 100%. , recorded as ,calculate .

[0043] For the Fault origin under various working conditions and plan The first in monitoring points Construct a single-point dilution ratio for: in: For monitoring points In the Characteristic gas concentration under certain operating conditions The source of the fault Characteristic gas concentration in situ; The scheme is determined based on the single-point dilution ratio. The minimum dilution ratio for this scheme is: in, Monitoring point scheme The monitoring point index in the middle, ; For operating condition index, ; Indicates the first Solution under various working conditions The Middle The single-point dilution ratio corresponding to each monitoring point. This represents the minimum single-point dilution ratio of the scheme across all operating conditions and all monitoring point combinations. It characterizes the fidelity of the scheme in preserving the concentration of fault-specific gases under the most unfavorable conditions. A higher value indicates a smaller gas dilution rate under the most unfavorable conditions. In this embodiment, for 40 operating conditions and the scheme... For each monitoring point, calculate the dilution ratio at all individual points, and take the minimum value as the dilution ratio. .

[0044] Based on the threshold trigger probability and the minimum dilution ratio of the scheme, a stratified selection of the best scheme is performed. Specifically: under the condition that the number of monitoring points N is the same, the scheme... Superior Solution The conditions are: Scheme Threshold trigger probability The trigger probability is greater than the threshold of scheme S2. ,or equal And the minimum dilution ratio of scheme S1 Larger than the plan Minimum dilution ratio of the scheme That is, the threshold trigger probability is compared first, and if the trigger probabilities are the same, the minimum dilution ratio of the scheme is compared. This rule fully meets the engineering requirements of ensuring that faults are not missed in the field of transformer energized operation.

[0045] S104. The particle swarm optimization algorithm is used to search for the optimal monitoring point layout scheme, and the obtained optimal monitoring point layout scheme is verified. The three-dimensional coordinates of each monitoring point in the optimal monitoring point layout scheme are output.

[0046] Set population size =100. Position vector of each particle. Indicates a set containing The plan includes monitoring points, totaling [number] points. Dimension, that is: In the candidate point set Random selection Each particle has an initial position consisting of _ points, and the distance between any two points is not less than _ points. ,in Set the value to 500mm. Initialize the velocity vector for each particle. It is a zero vector.

[0047] For each particle, its position vector is... Decoded as The coordinates of each monitoring point are calculated, and the threshold trigger probability of the monitoring point scheme corresponding to the particle is determined. and minimum dilution ratio of the scheme .in, This is used to characterize the proportion of operating conditions in which the monitoring point scheme successfully triggers an alarm across all operating conditions. This is used to characterize the fidelity of the monitoring point scheme in preserving the concentration of fault characteristic gases under the most unfavorable conditions across all operating scenarios. It is compared according to priority. ,exist Compare only when they are the same Priority-based stratified selection rules determine the ranking of each particle in the population; Update the individual historical best position of each particle. (Even the optimal position in the score hierarchy), and the globally optimal position of the population. .

[0048] Update the velocity and position of each particle according to the following formula: In the formula: and Particles In the The position vector and velocity vector of the step; As an inertia weight, it decreases linearly from 0.9 to 0.4; All are set to 1.5; for Independent random numbers within the interval.

[0049] After each position update, the coordinates of each particle are checked: if the coordinates of any monitoring point fall into the forbidden region or are not in the candidate point set... If the Euclidean distance between any two monitoring points is less than 1, then map it to the nearest valid candidate point; if the Euclidean distance between any two monitoring points is less than 1, then map it to the nearest valid candidate point. Then, the coordinates are adjusted by random perturbation until the constraints are met.

[0050] Set the maximum number of iterations. The iteration count is set to 200, or terminates early if the global optimum shows no improvement within 20 consecutive iterations. After the iteration ends, the monitoring point scheme corresponding to the global optimum position is output. .for The answer is 4, which gives the optimal 4-point solution. (Referred to as Plan #1). For The answer is 7, which gives the optimal 7-point solution. (Referred to as Scheme #3). Table 1 provides... The three-dimensional coordinates and location descriptions of each monitoring point are given in Table 2. The three-dimensional coordinates and location descriptions of each monitoring point are provided. Depending on the actual gas detection method used, each coordinate location can correspond to the oil sample or gas sampling interface, the gas sensor installation location, the location of the optical probe or optical detection channel, and the information acquisition location of other gas detection units.

[0051] Figure 3 The diagram shows the three-dimensional spatial distribution of four monitoring point layout schemes inside the transformer.

[0052] The optimal solution output above (Option #1) and (Scheme #3), together with the existing engineering experience-based point placement scheme (denoted as the original scheme) and another 4-point manual experience scheme (denoted as Scheme #2, as a comparison example), are substituted into the 40 sets of working condition datasets in step S101 to calculate the threshold trigger probability of each scheme under different load rates. and its average value under all operating conditions, the results are as follows Figure 4 As shown.

[0053] Figure 4 In the analysis, the original scheme is the existing 4-point sidewall monitoring scheme, with an average threshold trigger probability of 42.5% across all operating conditions; Scheme #2 is another 4-point scheme based on human experience, with an average threshold trigger probability of 72.5% across all operating conditions; Scheme #1 is the 4-point scheme optimized in this application, with an average threshold trigger probability of 82.5% across all operating conditions. Given that the number of monitoring points is always 4, the threshold trigger probability of Scheme #1 is higher than that of the original scheme and Scheme #2, indicating that optimizing the monitoring point layout based on the three-dimensional spatial distribution data of fault characteristic gases under multiple operating conditions can improve the operating condition coverage capability of the monitoring point combination.

[0054] Figure 4 In the middle, scheme #3 is a further increase in the number of monitoring points in this application. The optimized scheme obtained under the given conditions has an average threshold trigger probability of 90.0% across all operating conditions, demonstrating that the method described in this application can be used to optimize the layout of monitoring points under different preset monitoring point numbers. Since the number of monitoring points in scheme #3 differs from the aforementioned four schemes, scheme #3 is not used alone as a comparative scheme with the same number of points to demonstrate the effectiveness of the monitoring point spatial layout optimization.

[0055] Figures 5-8 The demonstration further showcased the path from the fault gas production point to each monitoring point under four typical operating conditions for Scheme #3. Curve showing the change in component mass fraction. Figures 5-8 The component mass fraction is used to show the spatial variation trend of the fault characteristic gas along the path from the fault gas generation point to the monitoring point, and is not directly used as the comparison quantity of the alarm threshold in step S103.

[0056] Figures 5-8The horizontal axis of the curve represents the relative distance from the monitoring point to the fault gas generation point. This distance passes through solid structures such as windings and the core. The transformer interior contains solid structures such as windings and the core. These solid structures occupy spaces without insulating oil, preventing the fault characteristic gases from dissolving, diffusing, and being transported within them; therefore, no corresponding gas concentration data exists. When the path connecting the gas generation point and the monitoring point crosses the area occupied by these solid structures, the concentration curve will be interrupted within the relative distance range corresponding to the solid structure. The spatial location of the truncated segment corresponds to the area occupied by solid components inside the transformer. The continuous curve segments before and after the truncation correspond to the effective oil flow path segments on both sides of the solid structure, and the continuous curve segments are all located within the area where the oil phase medium can flow.

[0057] For the Fault gas generation point under various operating conditions With the monitoring points Let the straight-line distance between the two be . The straight-line distance from any point on the path to the fault gas-producing point is The relative distance from the monitoring point to the fault gas production point Represented as: in, The range of values ​​is ,when When, it corresponds to the fault gas production point; when At that time, the corresponding monitoring point; when At that time, it corresponds to the middle position on the path between the fault gas production point and the monitoring point.

[0058] Table 1 Optimal 4-point scheme Monitoring point coordinates Table 2 Optimal 7-point scheme Monitoring point coordinates Calculate the The trigger probability enhancement rate and the minimum dilution ratio enhancement rate are calculated using the following formulas: in, Based on existing engineering experience in site selection and manual experience in site selection.

[0059] Finally, output the optimal solution. and The three-dimensional coordinates and location descriptions of each monitoring point are shown in Tables 1 and 2.

[0060] It should be noted that this embodiment uses acetylene as the target fault characteristic gas and a concentration threshold as the alarm criterion for illustration. This is merely an example of the specific implementation process of this application and does not constitute a limitation on the type of target gas or the detection method. In other embodiments, one or more fault characteristic gases can be selected according to monitoring requirements, and gas chromatography, spectroscopy, photoacoustics, electrochemistry, semiconductor gas sensing, or other gas detection methods can be employed. For different detection methods, the optical, electrical, acoustic, or other response signals output by the detection device can be converted into gas concentration or equivalent concentration characterization values ​​through calibration relationships. The evaluation criteria and intelligent optimization algorithm described in this application are used to determine the monitoring point layout for the corresponding detection method. Multiple different detection methods can be used simultaneously, and it is not limited to a single oil chromatography monitoring method.

[0061] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, the simultaneous existence of A and B, or the existence of B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects have an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0062] The above description is merely a specific embodiment of this application. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application. The protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for intelligent optimization of oil chromatography monitoring point layout in oil-immersed transformers based on gas distribution simulation, characterized in that, include: A three-dimensional spatial distribution dataset of fault characteristic gases inside a transformer under multiple operating conditions is obtained. The multiple operating conditions include various load rates and multiple typical fault gas generation locations. The three-dimensional spatial distribution dataset includes the fault characteristic gas concentration values ​​at each spatial location in the transformer's internal oil calculation domain under each operating condition. A candidate space for monitoring points is defined and discretized into a set of candidate points. The candidate space for monitoring points includes a deployable region that meets preset conditions. The set of candidate points consists of three-dimensional coordinate points generated by spatial discretizing the deployable region according to a preset selection interval. A quantitative evaluation criterion for the monitoring point scheme is constructed, and the evaluation criterion adopts a priority-based hierarchical screening rule; The optimal monitoring point layout scheme is constructed using the particle swarm optimization algorithm, and the obtained optimal monitoring point layout scheme is verified. The three-dimensional coordinates of each monitoring point in the optimal monitoring point layout scheme are output.

2. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 1, characterized in that, The acquisition of the three-dimensional spatial distribution dataset of characteristic gases in transformer internal faults under multiple operating conditions includes: A target oil-immersed transformer was selected, and a three-dimensional simulation model of the transformer was built based on the actual structure and size parameters of the equipment. The model retains the structures that have an impact on the oil flow field and gas transport, and simplifies the structures that have a smaller impact on the oil flow field and fault characteristic gas transport. Configure multiple typical operating conditions and aggregate all operating condition data to form a three-dimensional spatially distributed dataset.

3. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 2, characterized in that, The configuration includes multiple typical operating conditions, and all operating condition data are aggregated to form a three-dimensional spatially distributed dataset, including: Set the load rate level to The number of faulty gas-producing locations is The total operating condition number is formed by the pairwise combination of load rate and fault gas generation location. : in: Includes various operating conditions such as no-load, light-load, half-load, and heavy-load. Distributed in windings, cores, leads, bushings, and other typical fault locations; The constant gas production rate of the fault characteristic gas is set as follows: The simulation duration is set to Solve for the concentration of characteristic gases at each grid node in the computational domain of the transformer's internal oil after steady-state operation; Let the total number of grid nodes in the oil computational domain be . , No. The three-dimensional coordinates of each grid node are ,in, ; The three-dimensional spatial distribution data corresponding to all operating conditions are aggregated to form a total operating condition data set: in: For the first The data subset corresponding to each working condition Indicates the first The first working condition Fault characteristic gas concentration at each grid node This is the transformer operating condition sequence number, and its value range is... .

4. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 1, characterized in that, The definition of the candidate space for monitoring points and its discretization into a set of candidate points includes: The system is divided into deployable and prohibited areas. The deployable area is the area inside the transformer where fault characteristic gas concentration information can be obtained and the insulation distance, sampling, or detection requirements are met. The prohibited area includes the bushing riser installation interference area, the cooler pipe connection interference area, the external support shielding area, the area occupied by solid structures such as the iron core and windings, as well as other unsuitable deployment areas determined according to the insulation safety, sampling, detection, and maintenance requirements of the monitoring device. No candidate measurement points are generated in the prohibited area. The deployable region is discretized according to a preset spacing. Uniform discretization generates a set of candidate measurement points: in: The total number of candidate points, each candidate point has unique three-dimensional spatial coordinates; The number of monitoring points to be optimized is set as follows: ,in Monitoring point plan Represented as from the candidate point set Selected from The combination of points, namely: in, Indicating the monitoring point scheme The first in One monitoring point, Its three-dimensional spatial coordinates are as follows , ; For any two monitoring points and The Euclidean distance between the two is: The monitoring point scheme satisfies the constraint that the Euclidean distance between any two monitoring points is not less than a preset minimum spacing. .

5. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 1, characterized in that, The quantitative evaluation criteria for the monitoring point construction scheme, wherein the evaluation criteria adopt a priority-based hierarchical screening rule, include: Two evaluation metrics were constructed: threshold trigger probability and minimum dilution ratio of the scheme. Number of monitoring points Under the same conditions, the merits of the schemes are determined based on the threshold trigger probability and the minimum dilution ratio of the scheme. The threshold trigger probability is compared first, and the scheme with the highest trigger probability is the optimal scheme. If the trigger probabilities are equal, then the schemes with the smallest dilution ratios are compared, and the scheme with the largest smallest dilution ratio is the optimal scheme.

6. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 5, characterized in that, The two evaluation metrics, threshold trigger probability and minimum dilution ratio, are constructed respectively, including: Set the preset alarm threshold as For the first For various working conditions, when the monitoring point scheme The concentration of the fault characteristic gas at at least one monitoring point is greater than or equal to the preset alarm threshold. At that time, determine the scheme to which the monitoring point belongs. In the The alarm was successfully triggered under the following conditions; Statistical monitoring point scheme In all Number of working conditions that successfully triggered alarms in various working conditions Construct the threshold trigger probability: in, The range of values ​​is The larger the value, the stronger the ability of the monitoring point scheme to successfully detect fault characteristic gases and trigger alarms under multiple operating conditions; For the Fault origin under various working conditions and plan The first in monitoring points Construct a single-point dilution ratio for: in: For monitoring points In the Characteristic gas concentration under certain operating conditions The source of the fault Characteristic gas concentration in situ; The scheme is determined based on the single-point dilution ratio. The minimum dilution ratio for this scheme is: in, Monitoring point scheme The monitoring point index in the middle, ; For operating condition index, ; Indicates the first Solution under various working conditions The Middle The single-point dilution ratio corresponding to each monitoring point.

7. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 1, characterized in that, The optimal monitoring point layout scheme is constructed using the particle swarm optimization algorithm, and the obtained optimal monitoring point layout scheme is verified. The three-dimensional coordinates of each monitoring point in the optimal monitoring point layout scheme are output, including: Each particle represents a scheme containing a preset number of monitoring points. By iteratively updating the position and velocity of the particles and applying monitoring point spacing constraints and no-displacement zone constraints, the scheme is optimized generation by generation until convergence, thus obtaining the globally optimal monitoring point scheme. The obtained global optimal monitoring point scheme is compared with the preset comparison point layout scheme, the performance evaluation index is calculated, and the three-dimensional coordinates of each monitoring point in the global optimal monitoring point scheme are output.

8. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 7, characterized in that, The scheme, where each particle represents a set of monitoring points with a preset number of monitoring points, iteratively updates the particle's position and velocity, applies monitoring point spacing constraints and no-go zone constraints, and optimizes generation by generation until convergence, obtaining the globally optimal monitoring point scheme, including: Set the size as , A positive integer, the position vector of each particle. Indicates a set containing The plan includes monitoring points, totaling [number] points. Dimension, that is: In the candidate point set Random selection Each particle has an initial position consisting of _ points, and the distance between any two points is not less than _ points. Initialize the velocity vector of each particle. It is a zero vector; For each particle, its position vector is... Decoded as The coordinates of each monitoring point are calculated, and the threshold trigger probability of the monitoring point scheme corresponding to the particle is determined. and minimum dilution ratio of the scheme To determine the ranking of each particle in the population; Update the individual historical best position of each particle. and the global optimal position of the population ; According to the above and Update the velocity and position of each particle; After each position update, the coordinates of each particle are checked. If the coordinates of any monitoring point fall into the forbidden region or are not in the candidate point set, the check is performed. If the Euclidean distance between any two monitoring points is less than 1, then map it to the nearest valid candidate point; if the Euclidean distance between any two monitoring points is less than 1, then map it to the nearest valid candidate point. Then, the coordinates are adjusted by random perturbation until the constraint conditions are met; When the maximum number of iterations is reached or the global optimal position of the population The iteration terminates when there is no improvement within a preset consecutive algebra, and the current iteration is output. Corresponding monitoring point scheme .

9. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 8, characterized in that, According to the and The velocity and position of each particle are updated, including: In the formula: and Particles In the The position vector and velocity vector of the step; As an inertial weight, it is preset to decrease linearly from a larger value to a smaller value; and As a learning factor, and for Independent random numbers within the interval.

10. The intelligent optimization layout method for oil chromatography monitoring points of oil-immersed transformers based on gas distribution simulation according to claim 7, characterized in that, The process of comparing the obtained globally optimal monitoring point scheme with a preset comparison point layout scheme, calculating performance evaluation indicators, and outputting the three-dimensional coordinates of each monitoring point in the globally optimal monitoring point scheme includes: Selecting the globally optimal monitoring point scheme Pre-set comparison point layout scheme with the same number of monitoring points ; The and stated Substitute into the above The dataset shows the three-dimensional spatial distribution of various working conditions, and the calculations are performed for each solution. and plan ; Calculate the The increase in trigger probability and the increase in minimum dilution ratio of the solution: Output the Description of the three-dimensional coordinates and location of each monitoring point.

Images