An oil-immersed transformer energy efficiency monitoring system based on multi-parameter fusion

Abstract

The present application relates to transformer operation monitoring technical field, disclose a kind of oil-immersed transformer energy efficiency monitoring system based on multi-parameter fusion, comprising: data acquisition and window module, collect high-low voltage side electrical quantity, oil temperature, ambient temperature and cooling state, obtain active power and temperature rise and divide identification, correction and verification data;Transient parameter identification module, according to power increment and temperature rise dynamic determination candidate thermal parameter;Steady-state double-path correction module, according to steady-state temperature rise inversion loss and determine high-low voltage side candidate electrical correction parameter;Cross-validation and update module, according to oil temperature residual and loss residual select target path and update parameter;Energy efficiency and abnormality determination module, calculate transformer energy efficiency and determine measurement drift, heat dissipation drop or real loss increase.Transient identification, double-path correction and cross-validation, distinguish high-low voltage side measurement deviation and update thermal parameter and electrical parameter jointly, improve energy efficiency monitoring credibility.

Description

Technical Field

[0001] This invention relates to the field of transformer operation monitoring technology, specifically to an oil-immersed transformer energy efficiency monitoring system based on multi-parameter fusion. Background Technology

[0002] Oil-immersed transformers are electrical devices that use insulating oil as both the insulating and heat dissipation medium. During operation, they experience no-load losses, load losses, and additional losses. Energy efficiency monitoring is used to collect voltage, current, and temperature data from the high and low voltage sides of the transformer to determine the input active power, output active power, and operating energy efficiency. To reflect the energy conversion status of the transformer under conditions of load changes, ambient temperature changes, and cooling device switching, energy efficiency monitoring typically requires analysis in conjunction with electrical operating data, oil temperature, ambient temperature, and cooling status.

[0003] Existing methods for monitoring the energy efficiency of oil-immersed transformers typically calculate losses based on the difference in active power between the high-voltage and low-voltage sides, or estimate losses based on fixed thermal parameters and steady-state temperature rise, and then correct the measurement results using pre-calibrated electrical correction parameters. However, during long-term operation, both electrical measurement link parameters and transformer heat dissipation parameters may change. Existing methods struggle to distinguish the impact of high-voltage side measurement deviations, low-voltage side measurement deviations, and changes in thermal parameters on the monitoring results. This can easily lead to mutual compensation between electrical correction parameters and thermal parameters, resulting in insufficient reliability of the parameters used for energy efficiency monitoring. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion, which solves the problem that existing technologies have difficulty distinguishing between electrical measurement deviations and changes in thermal parameters, resulting in insufficient reliability of energy efficiency monitoring parameters.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion, comprising: The data acquisition and windowing module collects high and low voltage, current, oil temperature, ambient temperature and cooling status to obtain input and output active power and temperature rise, and divides the synchronous data into transient identification, steady-state correction, transient verification and steady-state verification data. The transient parameter identification module fixes the previous effective electrical correction parameter and dynamically determines candidate thermal parameters, including equivalent thermal conductivity and equivalent heat capacity parameters, based on power increment and temperature rise. The previous effective electrical correction parameter is initially adopted as the calibration parameter. The steady-state dual-path correction module fixes candidate thermal parameters, determines heat inversion loss based on steady-state temperature rise, establishes high- and low-voltage side correction paths, fixes the previous effective electrical correction parameters of the non-correction side for each path, and determines candidate proportional correction parameters and candidate bias correction parameters of the correction side based on at least two load levels. The cross-validation and update module processes transient and steady-state validation data using candidate parameters for each path, determines oil temperature residual and loss residual respectively, and enters the candidate path if it meets the threshold of the two types of residuals. The target path is determined based on the comprehensive residual and the corresponding candidate parameters are jointly updated. Otherwise, the effective parameters are not updated. The energy efficiency calculation module corrects the input and output active power based on the current effective electrical correction parameters to determine the transformer's energy efficiency; The anomaly detection module determines the main anomaly type among electrical measurement link drift, heat dissipation performance degradation, or actual loss increase based on electrical correction parameter offset, equivalent thermal conductivity decrease, and the joint increase of electrical loss and heat inversion loss.

[0006] Preferably, the data acquisition and windowing module includes: The electrical acquisition unit collects the three-phase voltage and three-phase current on the high-voltage side and the low-voltage side according to a unified clock, and determines the input active power and output active power respectively based on the three-phase voltage and three-phase current within the same power frequency cycle. The thermal state acquisition unit collects the top oil temperature, bottom oil temperature, ambient temperature, and cooling status. It determines the oil temperature based on the top oil temperature and bottom oil temperature, and determines the temperature rise based on the difference between the oil temperature and the ambient temperature. The synchronization processing unit combines the input active power, output active power, oil temperature, ambient temperature, temperature rise, and cooling status according to the sampling time and a preset synchronization cycle to form synchronization data.

[0007] Preferably, the synchronized data for data acquisition and windowing module includes: The load change amplitude, load change rate, temperature rise response amplitude, temperature rise change rate, and cooling state switching time are determined based on the synchronous data. Data that reach the corresponding thresholds for load change amplitude and temperature rise response amplitude are identified as transient identification data, and data that are lower than the corresponding thresholds for load change rate and temperature rise change rate are identified as steady-state correction data. When switching cooling states, the transition data is removed and the data is segmented according to the cooling state. From the data that has never participated in the candidate parameter calculation and covers the cooling state corresponding to the parameter to be verified, transient and steady-state data with different load change amplitudes or initial temperature rises are selected and determined as transient verification data and steady-state verification data, respectively.

[0008] Preferably, the transient parameter identification module includes: The incremental generation unit uses the data before the load change in the transient identification data as a reference to determine the input active power increment, output active power increment, and temperature rise increment. The incremental correction unit uses the previous effective proportional correction parameters of the high-voltage side and the low-voltage side to correct the incremental input active power and the incremental output active power respectively, and determines the incremental loss based on the difference between the two after correction. The thermal parameter identification unit determines candidate equivalent thermal conductivity parameters and candidate equivalent heat capacity parameters based on incremental loss, temperature rise increment, and temperature rise rate.

[0009] Preferably, the thermal parameter identification unit determines candidate equivalent thermal parameters by: Based on at least two transient identification data with different load change amplitudes or load change rates, establish a thermal balance relationship between incremental loss, temperature rise increment, and temperature rise rate. Set the corresponding equivalent thermal conductivity parameters according to the cooling state, and jointly determine the equivalent heat capacity parameters based on each transient identification data; Solve for the equivalent thermal conductivity and equivalent heat capacity values ​​that make the residual of the thermal balance relationship satisfy the preset conditions, and determine them as candidate equivalent thermal parameters when they are within the corresponding parameter range.

[0010] Preferably, the steady-state dual-path correction module includes: The heat inversion unit determines the heat inversion loss based on the temperature rise of the steady-state correction data and the candidate equivalent thermal conductivity parameters of the corresponding cooling state, under the condition of fixed candidate equivalent thermal parameters. The correction path generation unit establishes a high-voltage side correction path and a low-voltage side correction path. The high-voltage side correction path fixes the previous effective electrical correction parameter of the low-voltage side, and the low-voltage side correction path fixes the previous effective electrical correction parameter of the high-voltage side. The candidate parameter generation unit determines the candidate proportional correction parameter and the candidate bias correction parameter corresponding to the two correction paths based on steady-state correction data of at least two different load levels.

[0011] Preferably, the candidate parameter generation unit specifically includes: In the high-voltage side correction path, the sum of the output active power corrected according to the previous effective electrical correction parameters and the heat inversion loss is determined as the high-voltage side target power. In the low-voltage side correction path, the difference between the input active power corrected according to the previous effective electrical correction parameters and the heat inversion loss is determined as the low-voltage side target power. Based on the relationship between the measured active power and the corresponding target power under at least two different load levels, candidate proportional correction parameters and candidate bias correction parameters for the high-voltage side correction path and the low-voltage side correction path are determined respectively.

[0012] Preferably, the cross-validation and update module includes: The transient verification unit processes the transient verification data using the candidate electrical correction parameters corresponding to the high-voltage side correction path and the low-voltage side correction path respectively. Based on the corrected loss, candidate equivalent thermal conductivity parameters and candidate equivalent heat capacity parameters, the predicted oil temperature is determined, and the oil temperature residual between the predicted oil temperature and the measured oil temperature is obtained. The steady-state verification unit uses candidate electrical correction parameters corresponding to the high-voltage side correction path and the low-voltage side correction path to determine the electrical loss in the steady-state verification data, determines the heat inversion loss based on the candidate equivalent thermal conductivity parameters and temperature rise, and obtains the loss residual between the electrical loss and the heat inversion loss. The path determination unit determines the target correction path based on the oil temperature residual and loss residual corresponding to the high-pressure side correction path and the low-pressure side correction path.

[0013] Preferably, the path determination unit determines the target correction path and updates the parameters according to the following process: The transient and steady-state verification data were processed using the previous effective parameters to obtain the reference oil temperature residual and the reference loss residual, respectively. The correction path that is lower than the corresponding residual threshold and the corresponding benchmark residual is selected as the candidate correction path. The comprehensive residual of each candidate correction path is determined based on the ratio of oil temperature residual to oil temperature residual threshold and the ratio of loss residual to loss residual threshold. When there is only one candidate correction path, the candidate correction path is determined as the target correction path; when there are two candidate correction paths, when the difference between the comprehensive residuals of the two candidate correction paths reaches the path discrimination threshold, the candidate correction path with the smaller comprehensive residual is determined as the target correction path. Update the candidate electrical correction parameters and candidate thermal parameters corresponding to the target correction path to the current valid parameters; if the target correction path is not determined, retain the previous valid parameters.

[0014] Preferably, the energy efficiency calculation module and the anomaly detection module include: The energy efficiency determination unit corrects the input active power and output active power according to the current effective electrical correction parameters on the high-voltage side and the low-voltage side, respectively, and determines the transformer energy efficiency based on the ratio of the corrected output active power to the input active power. The abnormal index generation unit determines the electrical parameter offset index based on the offset of the current effective electrical correction parameter relative to the historical benchmark, determines the thermal conductivity decrease index based on the decrease of the current effective equivalent thermal conductivity parameter relative to the corresponding cooling state historical benchmark, and determines the loss growth index based on the common increase of electrical loss and heat inversion loss relative to the historical benchmark. The anomaly classification unit determines the main anomaly type of the indicator that reaches the corresponding threshold and has the largest normalized exceedance among the electrical parameter deviation index, thermal conductivity decrease index, and loss increase index when the transformer energy efficiency is lower than the energy efficiency benchmark. When two or three indicators reach the corresponding threshold, a composite anomaly label composed of the corresponding anomaly types is generated. The electrical parameter deviation index, thermal conductivity decrease index, and loss increase index correspond to electrical measurement link drift, heat dissipation performance degradation, and actual loss increase, respectively.

[0015] This invention provides an energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion. It has the following beneficial effects: 1. This invention combines transient thermal parameter identification, steady-state dual-path electrical correction, and independent cross-validation to distinguish measurement deviations on the high and low voltage sides under the condition of fixed non-corrected side parameters, and jointly updates candidate thermal parameters and electrical correction parameters to improve the reliability of energy efficiency monitoring parameters.

[0016] 2. This invention divides transient identification, steady-state correction, transient verification, and steady-state verification data according to different load change characteristics and cooling states, reducing the impact of data mixing and cooling state switching on parameter identification and correction results.

[0017] 3. This invention performs dual verification of two correction paths using oil temperature residual and loss residual, and maintains the previous valid parameter when the verification result does not meet the threshold or the path distinction is insufficient, thereby reducing the impact of erroneous parameter updates on energy efficiency calculation results.

[0018] 4. This invention integrates electrical correction parameters, equivalent thermal conductivity, electrical loss, and heat inversion loss, and can distinguish between electrical measurement link drift, decreased heat dissipation performance, and increased actual loss, providing a basis for transformer operating status analysis. Attached Figure Description

[0019] Figure 1 This is an architecture diagram of an oil-immersed transformer energy efficiency monitoring system based on multi-parameter fusion according to the present invention. Figure 2 This is an architecture diagram of the cross-validation and update module of the present invention; Figure 3 This is an architectural diagram of an embodiment of the present invention. Detailed Implementation

[0020] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see the appendix Figure 1 and attached Figure 2 This invention provides an energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion, comprising: The data acquisition and windowing module collects high and low voltage, current, oil temperature, ambient temperature and cooling status to obtain input and output active power and temperature rise, and divides the synchronous data into transient identification, steady-state correction, transient verification and steady-state verification data. Furthermore, the data acquisition and windowing module includes: The electrical acquisition unit collects the three-phase voltage and three-phase current on the high-voltage side and the low-voltage side according to a unified clock, and determines the input active power and output active power respectively based on the three-phase voltage and three-phase current within the same power frequency cycle. The thermal state acquisition unit collects the top oil temperature, bottom oil temperature, ambient temperature, and cooling status. It determines the oil temperature based on the top oil temperature and bottom oil temperature, and determines the temperature rise based on the difference between the oil temperature and the ambient temperature. The synchronization processing unit combines the input active power, output active power, oil temperature, ambient temperature, temperature rise, and cooling status according to the sampling time and a preset synchronization cycle to form synchronization data.

[0022] Specifically, during the electrical data acquisition process, three-phase voltage acquisition channels and three-phase current acquisition channels are set up on the high-voltage side and the low-voltage side, respectively. Each channel is triggered for sampling by the same clock source, ensuring that the electrical quantities on the high-voltage side and the low-voltage side have a consistent time reference. The voltage signal is connected to the acquisition terminal via a voltage transformer, and the current signal is connected to the acquisition terminal via a current transformer. The sampling frequency is set according to the requirements of power frequency signal analysis. The processor uses a complete power frequency cycle as the calculation interval, calculates the active power of each phase voltage and corresponding phase current, and then sums the three-phase calculation results to obtain the input active power on the high-voltage side and the output active power on the low-voltage side. If there are missing sampling points, abnormal timestamps, or asynchronous voltage and current channels within the power frequency cycle, the data of that cycle will not participate in subsequent processing, thereby avoiding incomparability between input and output power due to sampling misalignment.

[0023] The top oil temperature sensor is located at the upper part of the oil tank, and the bottom oil temperature sensor is located at the lower part of the oil tank. The ambient temperature sensor is located away from direct sunlight and local heat sources of the transformer. The cooling status is determined by the cooling device control signal, and the start-up status and switching time of the fan or oil pump are recorded. The arithmetic mean of the top and bottom oil temperatures is taken as the oil temperature. The temperature rise is obtained by subtracting the ambient temperature at the same sampling time from the oil temperature. Using both the top and bottom oil temperatures to determine the oil temperature can reduce the fluctuation caused by local oil flow and installation position of a single measuring point, and provide continuous thermal state data for subsequent thermal parameter identification.

[0024] During synchronization, a timestamp formed by a unified clock is used as the basis for data alignment, and data records are established according to a preset synchronization cycle. When the electrical sampling frequency is higher than the sampling frequency of temperature and cooling status, the input active power and output active power are first counted in each synchronization cycle, and then the oil temperature, ambient temperature, temperature rise and cooling status corresponding to the synchronization cycle are written into the same data record. When the temperature data update time is later than the synchronization time, the valid temperature data that is adjacent in time and does not exceed the allowable retention time is used. When the allowable retention time is exceeded, the synchronization cycle is marked as invalid. When the cooling status changes, the switching time is accurately recorded so that the transitional data caused by the cooling status switching can be removed later.

[0025] Taking a single data acquisition process with a synchronization period of 1 second as an example, at a certain synchronization moment, the input active power is 630.4 kW, the output active power is 618.7 kW, the top oil temperature is 64.2 degrees Celsius, the bottom oil temperature is 56.8 degrees Celsius, the ambient temperature is 31.5 degrees Celsius, the cooling fan is in operation, and the oil temperature determined by the top and bottom oil temperatures is 60.5 degrees Celsius, corresponding to a temperature rise of 29.0 degrees Celsius. The above data, along with the sampling time and cooling status, constitute a single synchronization data point. Continuously generated synchronization data retains the original sampling time, valid data markers, and cooling status switching information, which can be used in subsequent data partitioning processes for transient identification, steady-state correction, transient verification, and steady-state verification.

[0026] Furthermore, the data acquisition and windowing module divides and synchronizes the following data: The load change amplitude, load change rate, temperature rise response amplitude, temperature rise change rate, and cooling state switching time are determined based on the synchronous data. Data that reach the corresponding thresholds for load change amplitude and temperature rise response amplitude are identified as transient identification data, and data that are lower than the corresponding thresholds for load change rate and temperature rise change rate are identified as steady-state correction data. When switching cooling states, the transition data is removed and the data is segmented according to the cooling state. From the data that has never participated in the candidate parameter calculation and covers the cooling state corresponding to the parameter to be verified, transient and steady-state data with different load change amplitudes or initial temperature rises are selected and determined as transient verification data and steady-state verification data, respectively.

[0027] Specifically, the data acquisition and windowing module establishes continuous analysis windows according to the time sequence of synchronized data. The input active power and temperature rise at the start of the window are used as comparison benchmarks. The load change amplitude is determined by the change in input active power within the window relative to the initial input active power. The load change rate is determined by the change in input active power between adjacent synchronization cycles and the synchronization cycle. The temperature rise response amplitude is determined by the change in temperature rise within the window relative to the initial temperature rise. The temperature rise change rate is determined by the change in temperature rise between adjacent synchronization cycles and the synchronization cycle. The cooling state switching time is determined based on the timestamp of the fan control signal or oil pump control signal state change and is aligned with the timestamp of the synchronized data. For example, when the synchronization cycle is 1 second, if the input active power increases from 420 kW to 610 kW, the load change amplitude is 190 kW; if the temperature rise increases from 25.6 degrees Celsius to 30.4 degrees Celsius, the temperature rise response amplitude is 4.8 degrees Celsius. This processing allows for comparison of electrical load changes, thermal state response, and cooling state changes on the same time benchmark.

[0028] The transient identification data is filtered by sequentially judging whether the load change amplitude and temperature rise response amplitude reach the corresponding thresholds. After a load change occurs, if the load change amplitude reaches 15% of the rated active power and the temperature rise response amplitude reaches 3 degrees Celsius within the preset observation period, then the baseline data before the load change and the temperature rise response data after the load change are combined to form a transient identification window. Steady-state correction data is selected from continuous data where the load change rate is lower than the preset load change rate threshold, the temperature rise change rate is lower than the preset temperature rise change rate threshold, and the duration reaches the steady-state duration. For example, if the fluctuation of the input active power does not exceed 1% of the rated active power and the temperature rise change does not exceed 0.3 degrees Celsius within 30 consecutive minutes, the continuous data can be determined as steady-state correction data. When the cooling fan or oil pump is put into or taken out of service, the data before and after the state switch is divided by the state switch time, and the data within the preset transition time before and after the switch is discarded, so that the retained data segments correspond to the determined cooling state, reducing the impact of changes in heat dissipation conditions on parameter calculation.

[0029] Transient and steady-state verification data were selected from synchronous data that were not involved in candidate parameter acquisition and were matched according to the cooling state corresponding to the thermal parameter to be verified. Transient verification data selected were transient processes where the load change amplitude, load change rate, or initial temperature rise differed from the transient identification data. For example, if the transient identification data corresponded to an input active power increase from 420 kW to 610 kW with an initial temperature rise of 25.6 degrees Celsius, data with an input active power increase from 455 kW to 585 kW, an initial temperature rise of 28.1 degrees Celsius, and the same cooling state could be selected as transient verification data. Steady-state verification data selected were load levels that were not involved in steady-state correction. For example, if the steady-state correction data covered 40% and 60% of the rated load, stable operating data corresponding to 50% of the rated load could be selected as steady-state verification data. The same synchronous data was used only for one process in candidate parameter acquisition or parameter verification to maintain the independence of the verification results.

[0030] The transient parameter identification module fixes the previous effective electrical correction parameter and dynamically determines candidate thermal parameters, including equivalent thermal conductivity and equivalent heat capacity parameters, based on power increment and temperature rise. The previous effective electrical correction parameter is initially the calibration parameter. Furthermore, the transient parameter identification module includes: The incremental generation unit uses the data before the load change in the transient identification data as a reference to determine the input active power increment, output active power increment, and temperature rise increment. The incremental correction unit uses the previous effective proportional correction parameters of the high-voltage side and the low-voltage side to correct the incremental input active power and the incremental output active power respectively, and determines the incremental loss based on the difference between the two after correction. The thermal parameter identification unit determines candidate equivalent thermal conductivity parameters and candidate equivalent heat capacity parameters based on incremental loss, temperature rise increment, and temperature rise rate.

[0031] Specifically, after transient parameter identification begins, the previous effective electrical correction parameters for both the high-voltage and low-voltage sides remain unchanged within the same identification cycle to avoid simultaneous changes in electrical and thermal parameters in the solution. During the initial identification, the previous effective electrical correction parameters are the calibration parameters of the high-voltage and low-voltage side measurement links. After parameter verification and updating, subsequent identifications use the most recently verified electrical correction parameters. The incremental generation unit selects continuous stable data from the transient identification data before load changes, determines the input active power reference value, output active power reference value, and temperature rise reference value, respectively, and then subtracts each synchronous data point after load changes from the corresponding reference values ​​to form the input active power increment, output active power increment, and temperature rise increment. Taking an input active power of 420 kW, an output active power of 410 kW, and a temperature rise of 25.6 degrees Celsius before the load change as an example, the corresponding data at a certain synchronous moment after the load change are 610 kW, 594 kW, and 30.4 degrees Celsius. Then, the increment of input active power at that moment is 190 kW, the increment of output active power is 184 kW, and the increment of temperature rise is 4.8 degrees Celsius.

[0032] The incremental correction unit only uses the previous effective proportional correction parameter to process the power increment, and determines the incremental loss according to the following formula: ; In the formula, k is the synchronous sampling sequence number in the transient identification data. The incremental loss at the k-th synchronization time is expressed in kilowatts. and These are the previous effective proportional correction parameters for the high-pressure side and the low-pressure side, respectively. and Let be the input active power and the output active power at the k-th synchronization time, respectively. and These are the baseline values ​​for input and output active power before the load change, respectively. The power increment is formed by subtracting the data before and after the load change on the same side. The fixed bias is eliminated during the subtraction process, so the previous effective bias correction parameter is not invoked. When the aforementioned proportional correction parameters are 1.002 and 0.998, respectively, the corrected input active power increment is 190.38 kW, the corrected output active power increment is 183.632 kW, and the corresponding incremental loss is 6.748 kW.

[0033] The thermal parameter identification unit reads incremental losses and temperature rise increments sequentially according to the synchronization time, and determines the rate of temperature rise change based on the temperature rise increment changes at adjacent synchronization moments. The rate of temperature rise change reflects the rate of heat accumulation inside the oil-immersed transformer, the temperature rise increment reflects the current heat dissipation state, and the incremental loss serves as the heat input causing the temperature rise change. After establishing a correspondence among these three types of data under a unified time reference, they are used to determine candidate equivalent heat capacity parameters and candidate equivalent thermal conductivity parameters. The cooling state remains consistent during identification; when the cooling state changes, the corresponding data segment formed after removing the transitional data is used for identification. The candidate thermal parameters obtained in this way do not replace the current effective thermal parameters, but are passed to the subsequent steady-state correction and cross-validation process, and are jointly updated only after the residual conditions are met.

[0034] Furthermore, the thermal parameter identification unit determines the candidate equivalent thermal parameters, including: Based on at least two transient identification data with different load change amplitudes or load change rates, establish a thermal balance relationship between incremental loss, temperature rise increment, and temperature rise rate. Set the corresponding equivalent thermal conductivity parameters according to the cooling state, and jointly determine the equivalent heat capacity parameters based on each transient identification data; Solve for the equivalent thermal conductivity and equivalent heat capacity values ​​that satisfy the preset conditions for the residuals of the thermal equilibrium relationship, and determine them as candidate equivalent thermal parameters when they are within the corresponding parameter range.

[0035] Specifically, the thermal parameter identification unit reads the incremental loss, temperature rise increment, temperature rise rate of change, synchronization period, and cooling state from at least two transient identification windows. The selected transient identification windows differ in terms of load change amplitude or load change rate, so that the temperature rise increment and temperature rise rate of change form different combinations of changes. This avoids the inability of a single transient process to distinguish between equivalent thermal conductivity parameters and equivalent heat capacity parameters. Each transient identification window establishes an incremental thermal balance relationship according to the synchronization time sequence. The incremental loss is used to characterize the new heat input, the temperature rise rate of change is used to characterize the heat accumulation, and the temperature rise increment is used to characterize the new heat dissipation relative to the state before the load change. For example, a first transient identification window with an increase of 190 kW in input active power and a second transient identification window with an increase of 130 kW in input active power can be selected. The two have different temperature rise response processes and neither contains cooling state switching transition data.

[0036] The candidate thermal parameters are solved jointly according to the following formula: ; In the formula, The number of transient identification windows participating in joint identification. For transient identification window sequence number, For the first The number of valid synchronized data within each transient identification window. For synchronizing data sequence numbers, For the first The first transient identification window The incremental loss at each synchronization moment, in kilowatts. and This represents the temperature increase at adjacent synchronization points, in degrees Celsius. The synchronization period is in seconds. This is the equivalent heat capacity, expressed in kilojoules per degree Celsius. For the first Each transient identification window corresponds to the cooling status. The equivalent thermal conductivity is given below, in kilowatts per degree Celsius. and The equivalent heat capacity and cooling state obtained from the solution are respectively. The corresponding equivalent thermal conductivity value is shared by all transient identification windows. The equivalent thermal capacity parameter is set separately for different cooling states. The transient identification windows under the same cooling state take the corresponding equivalent thermal conductivity parameter.

[0037] Before solving, the amount and change characteristics of data in each transient identification window are checked. When the temperature rise increment and the temperature rise change rate cannot form a distinguishable relationship, the current identification is stopped and no candidate thermal parameters are generated. After solving, the thermal balance residual at each synchronous moment is calculated, and it is determined whether the residual meets the preset conditions. At the same time, it is determined whether the equivalent thermal conductivity and equivalent heat capacity values ​​are within the corresponding parameter range. The parameter range can be predetermined based on the factory temperature rise test data and historical normal operation data. For example, if the preset equivalent heat capacity range is 1500 kJ / °C to 5000 kJ / °C, and the equivalent thermal conductivity range under wind turbine operation is 0.2 kW / °C to 1.2 kW / °C, and the combined solution yields an equivalent heat capacity of 2800 kJ / °C and an equivalent thermal conductivity of 0.48 kW / °C, and the heat balance residuals meet the preset conditions, then both are determined as candidate equivalent thermal parameters. If the residuals do not meet the preset conditions or any parameter exceeds the corresponding range, the results of this round will not be included in the subsequent parameter update process.

[0038] The steady-state dual-path correction module fixes candidate thermal parameters, determines heat inversion loss based on steady-state temperature rise, establishes high- and low-voltage side correction paths, fixes the previous effective electrical correction parameters of the non-correction side for each path, and determines candidate proportional correction parameters and candidate bias correction parameters of the correction side based on at least two load levels. Furthermore, the steady-state dual-path correction module includes: The heat inversion unit determines the heat inversion loss based on the temperature rise of the steady-state correction data and the candidate equivalent thermal conductivity parameters of the corresponding cooling state, under the condition of fixed candidate equivalent thermal parameters. The correction path generation unit establishes a high-voltage side correction path and a low-voltage side correction path. The high-voltage side correction path fixes the previous effective electrical correction parameter of the low-voltage side, and the low-voltage side correction path fixes the previous effective electrical correction parameter of the high-voltage side. The candidate parameter generation unit determines the candidate proportional correction parameter and the candidate bias correction parameter corresponding to the two correction paths based on steady-state correction data of at least two different load levels.

[0039] Specifically, after receiving candidate equivalent thermal conductivity parameters and candidate equivalent heat capacity parameters, the heat inversion unit keeps the candidate thermal parameters unchanged during the current steady-state correction period and calls up the corresponding candidate equivalent thermal conductivity parameters according to the cooling state recorded in the steady-state correction data. If the rate of temperature rise change in the steady-state correction data is lower than a preset threshold, and the amount of heat storage change in the oil within that data segment is lower than a preset range, the heat inversion unit multiplies the steady-state temperature rise by the candidate equivalent thermal conductivity parameter for the corresponding cooling state to obtain the heat inversion loss for that steady-state data segment. For example, if the candidate equivalent thermal conductivity parameter corresponding to the fan operation state is 0.48 kW / °C, and the average temperature rise of a certain steady-state data segment is 29.0°C, then the heat inversion loss for that steady-state data segment is 13.92 kW. Each steady-state data segment selects the corresponding candidate equivalent thermal conductivity parameter according to its own cooling state, so that the heat inversion loss corresponds to the current heat dissipation conditions.

[0040] The calibration path generation unit establishes separate calibration paths for the high-voltage side and the low-voltage side for the same set of steady-state calibration data. The high-voltage side calibration path keeps the previous effective proportional correction parameter and the previous effective bias correction parameter of the low-voltage side unchanged, only using the high-voltage side proportional correction parameter and the high-voltage side bias correction parameter as parameters to be determined. The low-voltage side calibration path keeps the previous effective electrical correction parameter of the high-voltage side unchanged, only determining the corresponding parameters on the low-voltage side. During the initial calibration, the previous effective electrical correction parameter uses the calibration parameters of the measurement link; subsequent calibrations use the most recently verified effective parameter. The two paths correspond to the calibration assumptions of high-voltage side measurement link offset and low-voltage side measurement link offset, respectively, ensuring that only one side's electrical parameters change within the same path, reducing the uncertainty caused by simultaneous changes in high-voltage side parameters, low-voltage side parameters, and thermal parameters.

[0041] The candidate parameter generation unit reads the steady-state input active power, steady-state output active power, and corresponding heat inversion loss at at least two different load levels. The high-voltage side correction path uses the output active power corrected with the previous effective parameter and the heat inversion loss to form a high-voltage side power reference. Under fixed low-voltage side parameters, candidate proportional correction parameters and candidate bias correction parameters are determined based on the linear correspondence between the measured input active power and the high-voltage side power reference. The low-voltage side correction path uses the opposite data transmission direction to determine the low-voltage side candidate parameters. For example, selecting two steady-state load levels with input active power of approximately 420 kW and 610 kW, the power span formed by the load level change can be used to distinguish between proportional deviation and fixed deviation. The obtained candidate electrical correction parameters are only used as verification results for the two correction paths and do not replace the previous effective parameter at this stage. They are transmitted together with the candidate thermal parameters to the subsequent cross-validation process.

[0042] Furthermore, the candidate parameter generation unit specifically includes: In the high-voltage side correction path, the sum of the output active power corrected according to the previous effective electrical correction parameters and the heat inversion loss is determined as the high-voltage side target power. In the low-voltage side correction path, the difference between the input active power corrected according to the previous effective electrical correction parameters and the heat inversion loss is determined as the low-voltage side target power. Based on the relationship between the measured active power and the corresponding target power under at least two different load levels, candidate proportional correction parameters and candidate bias correction parameters for the high-voltage side correction path and the low-voltage side correction path are determined respectively.

[0043] Specifically, the candidate parameter generation unit keeps the previous effective proportional correction parameter and the previous effective bias correction parameter unchanged in the high-pressure side correction path. It first uses the parameters to correct the output active power in the steady-state correction data, and then adds the correction result to the heat inversion loss corresponding to the same steady-state data to obtain the target power on the high-pressure side. The relationship is as follows: ; In the formula, This is the sequence number of the steady-state correction data. For the first The target power on the high-voltage side corresponding to the set of data. This represents the measured output active power on the low-voltage side. and These are the previous effective proportional correction parameter and the previous effective bias correction parameter on the low-pressure side, respectively. To address heat loss inversion, this process uses the low-voltage side, where parameters remain stable, as the power reference, ensuring that the high-voltage side correction path is determined solely for the electrical parameters of the high-voltage side.

[0044] The candidate parameter generation unit keeps the previous effective proportional correction parameter and the previous effective bias correction parameter unchanged on the high-pressure side in the low-pressure side correction path. It subtracts the corresponding heat inversion loss from the corrected input active power to obtain the target power on the low-pressure side, with the following relationship: ; In the formula, For the first The target power on the low-voltage side corresponding to the set of data. This represents the measured input active power on the high-voltage side. and These are the previous effective proportional correction parameter and the previous effective bias correction parameter on the high-voltage side, respectively. The target power on both the high-voltage and low-voltage sides uses the same power unit as the measured active power. The two correction paths each fix the previous effective parameter on the non-corrected side, preventing the high-voltage and low-voltage side parameters from changing simultaneously within the same path.

[0045] After obtaining the target power corresponding to each steady-state data, the candidate parameter generation unit establishes a linear relationship between the measured input active power on the high-voltage side and the target power on the high-voltage side, as well as a linear relationship between the measured output active power on the low-voltage side and the target power on the low-voltage side. Each correction path selects at least two different load levels to form a power span between different power points that is sufficient to distinguish between proportional deviation and bias deviation. Based on this, the candidate proportional correction parameters and candidate bias correction parameters on the correction side are obtained. For example, the high-voltage side correction path uses two sets of steady-state data with measured input active power of 405 kW and 587 kW, corresponding to target power on the high-voltage side of 406.81 kW and 589.45 kW. The candidate proportional correction parameters and candidate bias correction parameters on the high-voltage side can be determined based on the two sets of correspondences. The candidate parameters obtained from the two paths are saved separately and are not replaced in the previous effective electrical correction parameters at this stage, but are transferred to the subsequent verification process together with the candidate thermal parameters.

[0046] Please see the appendix Figure 2 The cross-validation and update module processes transient and steady-state validation data using candidate parameters for each path, determines oil temperature residual and loss residual respectively, and enters the candidate path if it meets the threshold of the two types of residuals. The target path is determined based on the comprehensive residual and the corresponding candidate parameters are updated together; otherwise, the effective parameters are not updated. Furthermore, the cross-validation and update module includes: The transient verification unit processes the transient verification data using the candidate electrical correction parameters corresponding to the high-voltage side correction path and the low-voltage side correction path respectively. Based on the corrected loss, candidate equivalent thermal conductivity parameters and candidate equivalent heat capacity parameters, the predicted oil temperature is determined, and the oil temperature residual between the predicted oil temperature and the measured oil temperature is obtained. The steady-state verification unit uses candidate electrical correction parameters corresponding to the high-voltage side correction path and the low-voltage side correction path to determine the electrical loss in the steady-state verification data, determines the heat inversion loss based on the candidate equivalent thermal conductivity parameters and temperature rise, and obtains the loss residual between the electrical loss and the heat inversion loss. The path determination unit determines the target correction path based on the oil temperature residual and loss residual corresponding to the high-pressure side correction path and the low-pressure side correction path.

[0047] Specifically, the transient verification unit processes the transient verification data that were not involved in the candidate parameter calculation using both high-voltage and low-voltage correction paths. The high-voltage correction path uses high-voltage candidate electrical correction parameters and the previous valid electrical correction parameters on the low-voltage side, while the low-voltage correction path uses low-voltage candidate electrical correction parameters and the previous valid electrical correction parameters on the high-voltage side. Each path subtracts the corrected output active power from the corrected input active power to obtain the electrical losses during the transient verification process, and uses the measured oil temperature or temperature rise at the start of the transient verification window as the basis for calculation. As an initial state, the predicted oil temperature is determined moment by moment according to the aforementioned thermal balance relationship, taking into account the ambient temperature, synchronization cycle, candidate equivalent heat capacity parameters, and candidate equivalent thermal conductivity parameters corresponding to the current cooling state. The oil temperature residual is determined by the root mean square value of the difference between the predicted oil temperature and the measured oil temperature. For example, in a transient verification window containing 600 synchronization data, the oil temperature residual corresponding to the high-pressure side correction path is 0.72 degrees Celsius, and the oil temperature residual corresponding to the low-pressure side correction path is 1.35 degrees Celsius. This respectively characterizes the degree of fitting of the two paths to the dynamic process of temperature rise.

[0048] The steady-state verification unit uses the same path parameter combination as the transient verification to process the steady-state verification data, ensuring that the data used is not involved in the calculation of candidate electrical correction parameters. Each path first corrects the input and output active power, using the difference to determine the electrical loss; then, it determines the heat inversion loss based on the temperature rise in the steady-state verification data and the candidate equivalent thermal conductivity parameters under the corresponding cooling state. The loss residual is determined using the root mean square value of the difference between the electrical loss and the heat inversion loss in each steady-state verification data point. Transient verification focuses on verifying the applicability of candidate parameters to the oil temperature change process, while steady-state verification focuses on verifying the consistency between the electrical loss and the thermal state inversion results, requiring the same path to be verified simultaneously by both dynamic and steady-state processes.

[0049] The path determination unit receives the oil temperature residual and loss residual corresponding to the high-pressure side correction path and the low-pressure side correction path, and compares them with preset oil temperature residual thresholds and loss residual thresholds, respectively. Only when both types of residuals for the same path meet their corresponding thresholds is the path retained as a candidate correction path, avoiding parameter update judgments based solely on oil temperature residuals or loss residuals. For retained candidate correction paths, the path determination unit forms a comprehensive residual based on the normalization results of the two types of residuals relative to their respective thresholds, and transmits the candidate path and its comprehensive residual to the parameter update process. If no path meeting both types of residual thresholds is formed, the current effective electrical correction parameters and current effective thermal parameters remain unchanged.

[0050] Furthermore, the path determination unit determines the target correction path and updates the parameters according to the following process: The transient and steady-state verification data were processed using the previous effective parameters to obtain the reference oil temperature residual and the reference loss residual, respectively. The correction path that is lower than the corresponding residual threshold and the corresponding benchmark residual is selected as the candidate correction path. The comprehensive residual of each candidate correction path is determined based on the ratio of oil temperature residual to oil temperature residual threshold and the ratio of loss residual to loss residual threshold. When there is only one candidate correction path, the candidate correction path is determined as the target correction path; when there are two candidate correction paths, when the difference between the comprehensive residuals of the two candidate correction paths reaches the path discrimination threshold, the candidate correction path with the smaller comprehensive residual is determined as the target correction path. Update the candidate electrical correction parameters and candidate thermal parameters corresponding to the target correction path to the current valid parameters; if the target correction path is not determined, retain the previous valid parameters.

[0051] Specifically, the path determination unit first reprocesses the transient and steady-state verification data using the previous effective electrical correction parameters and the previous effective thermal parameters. During transient verification, the measured oil temperature or temperature rise at the start of the verification window is used as the initial state. A baseline predicted oil temperature is obtained based on the previous effective parameters, and the baseline oil temperature residual is determined by the difference between the baseline predicted oil temperature and the measured oil temperature. During steady-state verification, a baseline electrical loss is obtained based on the previous effective electrical correction parameters, and a baseline heat inversion loss is obtained based on the previous effective equivalent thermal conductivity parameters. The baseline loss residual is determined by the difference between the two. During the initial run, the previous effective electrical correction parameters are taken from the calibration parameters, and the previous effective thermal parameters are taken from the initial thermal parameters determined by the factory temperature rise test data or normal operation calibration data, ensuring that the candidate parameters have comparable baseline states.

[0052] For both high-pressure and low-pressure correction paths, their oil temperature residuals, loss residuals, corresponding residual thresholds, and baseline residuals are compared. A correction path is selected as a candidate correction path only if its oil temperature residual is lower than both the oil temperature residual threshold and the baseline oil temperature residual, and simultaneously, its loss residual is lower than both the loss residual threshold and the baseline loss residual. Therefore, candidate parameters must not only meet preset absolute error requirements but also show consistent improvement compared to the previous effective parameter, avoiding parameter updates triggered when the residual, although decreasing, remains within an unacceptable range.

[0053] The path determination unit determines the overall residual of the candidate correction path according to the following formula: ; in, To identify the calibration path, either the high-voltage side calibration path or the low-voltage side calibration path should be used. For path The overall residual, For path oil temperature residual, For path Loss residual, This is the oil temperature residual threshold. The threshold for loss residuals, and The weights are oil temperature residual weight and loss residual weight, respectively, and the sum of the two is 1. The two types of residuals are first divided by their respective thresholds to convert the temperature dimension and power dimension into comparable dimensionless quantities, and then fused according to the preset weights.

[0054] If only one correction path meets the selection criteria, the path determination unit identifies that path as the target correction path. If both correction paths meet the selection criteria, the combined residual difference between the two paths is determined according to the following formula: ; In the formula, This represents the combined residual difference between the two candidate correction paths. and These are the combined residuals of the high-voltage side correction path and the low-voltage side correction path, respectively. When the path discrimination threshold is reached, the path with the smaller overall residual is identified as the target correction path. When the difference does not reach the path discrimination threshold, it indicates that the verification results of the two paths are insufficient to form a reliable distinction, and the target correction path is not determined in this round. For example, if the overall residuals of the high-voltage side path and the low-voltage side path are 0.58 and 0.76 respectively, and the path discrimination threshold is 0.10, the difference between the two is 0.18, so the high-voltage side correction path can be identified as the target correction path.

[0055] After determining the target correction path, the candidate proportional correction parameters, candidate bias correction parameters, candidate equivalent thermal conductivity parameters, and candidate equivalent heat capacity parameters corresponding to that path are simultaneously updated to the current valid parameters, ensuring that the electrical and thermal parameters remain in the same verification state for this round. Non-target sides continue to retain their previous valid electrical correction parameters. If no candidate correction path is formed, or if two candidate correction paths fail to meet the path discrimination conditions, the current valid electrical correction parameters and current valid thermal parameters are not changed, and the previous valid parameters are retained for use in the next identification cycle.

[0056] The energy efficiency and anomaly determination module corrects the input active power and output active power based on the current effective electrical correction parameters, determines the transformer energy efficiency, and determines the main anomaly type among electrical measurement link drift, heat dissipation performance degradation, or actual loss increase based on the offset of electrical correction parameters, the decrease in equivalent thermal conductivity, and the joint increase of electrical loss and heat inversion loss.

[0057] Furthermore, the energy efficiency and anomaly detection module includes: The energy efficiency determination unit corrects the input active power and output active power according to the current effective electrical correction parameters on the high-voltage side and the low-voltage side, respectively, and determines the transformer energy efficiency based on the ratio of the corrected output active power to the input active power. The abnormal index generation unit determines the electrical parameter offset index based on the offset of the current effective electrical correction parameter relative to the historical benchmark, determines the thermal conductivity decrease index based on the decrease of the current effective equivalent thermal conductivity parameter relative to the corresponding cooling state historical benchmark, and determines the loss growth index based on the common increase of electrical loss and heat inversion loss relative to the historical benchmark. The anomaly classification unit determines the main anomaly type when the transformer energy efficiency is lower than the energy efficiency benchmark. Among the electrical parameter deviation index, thermal conductivity decrease index, and loss increase index, the index that reaches the corresponding threshold and has the largest normalized exceedance is identified. When two or three indexes reach the corresponding threshold, a composite anomaly label composed of the corresponding anomaly types is generated. The electrical parameter deviation index, thermal conductivity decrease index, and loss increase index correspond to electrical measurement link drift, heat dissipation performance degradation, and actual loss increase, respectively.

[0058] Specifically, the energy efficiency determination unit calls upon the currently valid high-voltage side proportional correction parameters and bias correction parameters to correct the input active power, and calls upon the currently valid low-voltage side proportional correction parameters and bias correction parameters to correct the output active power. The corrected input active power and output active power use data from the same synchronization period, and the difference between the two determines the current electrical loss. The ratio of the corrected output active power to the input active power determines the transformer energy efficiency. During historical normal operation, an energy efficiency benchmark is formed under the same load range and cooling conditions to reduce the impact of differences in load level and heat dissipation method on the energy efficiency comparison results. For example, if the corrected input active power is 610.5 kW, the corrected output active power is 596.8 kWh, the current electrical loss is 13.7 kW, and the transformer energy efficiency is 97.76%.

[0059] The anomaly indicator generation unit generates electrical parameter deviation indicators, thermal conductivity decrease indicators, and loss increase indicators, respectively. The electrical parameter deviation indicator is determined according to the following formula: ; The thermal conductivity decrease index and the loss increase index are determined according to the following formulas: ; ; In the formula, , and In order, they are electrical parameter deviation index, thermal conductivity decrease index, and loss increase index. , , , These are the currently valid electrical correction parameters. Parameters with a subscript of 0 correspond to historical normal baselines. , , and Allowed offset for the corresponding parameter. and Cooling state The current effective equivalent thermal conductivity parameters and historical thermal conductivity benchmarks are as follows. The threshold for thermal conductivity decrease, and These are the current electrical losses and the heat inversion losses, respectively. and This serves as a historical loss benchmark under the same load range and cooling conditions. and This corresponds to the loss growth threshold. The loss growth index takes the smaller of the two normalized increments, so that the index only reaches the judgment condition when electrical loss and heat inversion loss increase together.

[0060] When the current energy efficiency is lower than the corresponding energy efficiency benchmark, the anomaly classification unit compares three anomaly indicators. An indicator reaching 1 indicates that the corresponding threshold has been reached. The anomaly type corresponding to the indicator that reaches the threshold and has a larger value is determined as the primary anomaly type. Specifically, the electrical parameter deviation indicator corresponds to electrical measurement link drift, the thermal conductivity decrease indicator corresponds to decreased heat dissipation performance, and the loss increase indicator corresponds to increased actual loss. When two or three indicators reach the threshold, a composite anomaly label composed of the corresponding anomaly types is generated simultaneously. For example, when the three indicators are 1.35, 0.82, and 1.18 respectively, electrical measurement link drift is determined as the primary anomaly type, and a composite anomaly label of electrical measurement link drift and increased actual loss is generated. When two indicators have the same value, the primary anomaly type is determined according to a preset judgment order. When the energy efficiency is lower than the energy efficiency benchmark but none of the three indicators have reached the threshold, it is recorded as pending review.

[0061] Example Please see the appendix Figure 3 The monitoring object is an oil-immersed transformer with a rated capacity of 1000 kVA and a voltage level of 10 kV / 0.4 kV. The data acquisition and windowing module collects three-phase voltage, three-phase current, top oil temperature, bottom oil temperature, ambient temperature, and cooling status on both the high-voltage and low-voltage sides according to a unified clock, forming synchronized data with a 1-second synchronization cycle. Data with a load change amplitude reaching 15% of the rated active power and a temperature rise response amplitude of 3 degrees Celsius are used as transient identification data. Data with load change rate and temperature rise change rate below the corresponding threshold for 30 minutes are used as steady-state correction data. Data not involved in parameter calculation are used as transient verification data and steady-state verification data.

[0062] During a transient process, the input active power, output active power, and temperature rise before the load change were 420 kW, 410 kW, and 25.6 °C, respectively. After the load change, the corresponding data were 610 kW, 594 kW, and 30.4 °C. After processing the power increment using the previous effective proportional correction parameter, the incremental loss was found to be 6.558 kW. Combining this with another set of transient identification data with different load change amplitudes, the candidate equivalent heat capacity parameter was determined to be 2800 kJ / °C, and the candidate equivalent thermal conductivity parameter under the fan-operated state was determined to be 0.48 kW / °C.

[0063] During steady-state correction, the average temperature rise of a certain steady-state data segment was 25 degrees Celsius, and the heat inversion loss was determined to be 12 kW. The high-voltage side correction path fixed the previous effective electrical correction parameter on the low-voltage side. Based on the data of the two load levels, the candidate proportional correction parameter 1.0035 and the candidate bias correction parameter 0.386 kW were obtained on the high-voltage side. The corresponding candidate parameters were formed in the same way on the low-voltage side correction path.

[0064] Verification results show that the oil temperature residual and loss residual of the high-voltage side correction path are 0.72 degrees Celsius and 0.65 kW, respectively, both lower than the corresponding thresholds and baseline residuals. The two types of residuals in the low-voltage side correction path do not meet the threshold conditions. Therefore, the candidate electrical correction parameters and candidate thermal parameters of the high-voltage side are jointly updated to the current valid parameters. After the update, the corrected input active power is 610.5 kW, the corrected output active power is 596.8 kW, and the transformer efficiency is 97.76%. In subsequent operation, if the equivalent thermal conductivity parameter continues to decrease relative to the historical baseline under the same cooling condition, and the electrical parameter deviation index and loss growth index do not reach the corresponding thresholds, then the decline in heat dissipation performance will be identified as the main anomaly type.

[0065] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-parameter fusion-based energy efficiency monitoring system for oil-immersed transformers, characterized in that, include: The data acquisition and windowing module collects high and low voltage, current, oil temperature, ambient temperature and cooling status to obtain input and output active power and temperature rise, and divides the synchronous data into transient identification, steady-state correction, transient verification and steady-state verification data. The transient parameter identification module fixes the previous effective electrical correction parameter and dynamically determines candidate thermal parameters, including equivalent thermal conductivity and equivalent heat capacity parameters, based on power increment and temperature rise. The previous effective electrical correction parameter is initially adopted as the calibration parameter. The steady-state dual-path correction module fixes candidate thermal parameters, determines heat inversion loss based on steady-state temperature rise, establishes high- and low-voltage side correction paths, fixes the previous effective electrical correction parameters of the non-correction side for each path, and determines candidate proportional correction parameters and candidate bias correction parameters of the correction side based on at least two load levels. The cross-validation and update module processes transient and steady-state validation data using candidate parameters for each path, determines oil temperature residual and loss residual respectively, and enters the candidate path if it meets the threshold of the two types of residuals. The target path is determined based on the comprehensive residual and the corresponding candidate parameters are jointly updated. Otherwise, the effective parameters are not updated. The energy efficiency and anomaly determination module corrects the input active power and output active power based on the current effective electrical correction parameters, determines the transformer energy efficiency, and determines the main anomaly type among electrical measurement link drift, heat dissipation performance degradation, or actual loss increase based on the offset of electrical correction parameters, the decrease in equivalent thermal conductivity, and the joint increase of electrical loss and heat inversion loss.

2. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 1, characterized in that, The data acquisition and windowing module includes: The electrical acquisition unit collects the three-phase voltage and three-phase current on the high-voltage side and the low-voltage side according to a unified clock, and determines the input active power and output active power respectively based on the three-phase voltage and three-phase current within the same power frequency cycle. The thermal state acquisition unit collects the top oil temperature, bottom oil temperature, ambient temperature, and cooling status. It determines the oil temperature based on the top oil temperature and bottom oil temperature, and determines the temperature rise based on the difference between the oil temperature and the ambient temperature. The synchronization processing unit combines the input active power, output active power, oil temperature, ambient temperature, temperature rise, and cooling status according to the sampling time and a preset synchronization cycle to form synchronization data.

3. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 2, characterized in that, The data acquisition and windowing module divides and synchronizes the following data: The load change amplitude, load change rate, temperature rise response amplitude, temperature rise change rate, and cooling state switching time are determined based on the synchronous data. Data that reach the corresponding thresholds for load change amplitude and temperature rise response amplitude are identified as transient identification data, and data that are lower than the corresponding thresholds for load change rate and temperature rise change rate are identified as steady-state correction data. When switching cooling states, the transition data is removed and the data is segmented according to the cooling state. From the data that has never participated in the candidate parameter calculation and covers the cooling state corresponding to the parameter to be verified, transient and steady-state data with different load change amplitudes or initial temperature rises are selected and determined as transient verification data and steady-state verification data, respectively.

4. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 3, characterized in that, The transient parameter identification module includes: The incremental generation unit uses the data before the load change in the transient identification data as a reference to determine the input active power increment, output active power increment, and temperature rise increment. The incremental correction unit uses the previous effective proportional correction parameters of the high-voltage side and the low-voltage side to correct the incremental input active power and the incremental output active power respectively, and determines the incremental loss based on the difference between the two after correction. The thermal parameter identification unit determines candidate equivalent thermal conductivity parameters and candidate equivalent heat capacity parameters based on incremental loss, temperature rise increment, and temperature rise rate.

5. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 4, characterized in that, The thermal parameter identification unit determines candidate equivalent thermal parameters including: Based on at least two transient identification data with different load change amplitudes or load change rates, establish a thermal balance relationship between incremental loss, temperature rise increment, and temperature rise rate. Set the corresponding equivalent thermal conductivity parameters according to the cooling state, and jointly determine the equivalent heat capacity parameters based on each transient identification data; Solve for the equivalent thermal conductivity and equivalent heat capacity values ​​that make the residual of the thermal balance relationship satisfy the preset conditions, and determine them as candidate equivalent thermal parameters when they are within the corresponding parameter range.

6. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 5, characterized in that, The steady-state dual-path correction module includes: The heat inversion unit determines the heat inversion loss based on the temperature rise of the steady-state correction data and the candidate equivalent thermal conductivity parameters of the corresponding cooling state, under the condition of fixed candidate equivalent thermal parameters. The correction path generation unit establishes a high-voltage side correction path and a low-voltage side correction path. The high-voltage side correction path fixes the previous effective electrical correction parameter of the low-voltage side, and the low-voltage side correction path fixes the previous effective electrical correction parameter of the high-voltage side. The candidate parameter generation unit determines the candidate proportional correction parameter and the candidate bias correction parameter corresponding to the two correction paths based on steady-state correction data of at least two different load levels.

7. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 6, characterized in that, The candidate parameter generation unit specifically includes: In the high-voltage side correction path, the sum of the output active power corrected according to the previous effective electrical correction parameters and the heat inversion loss is determined as the high-voltage side target power. In the low-voltage side correction path, the difference between the input active power corrected according to the previous effective electrical correction parameters and the heat inversion loss is determined as the low-voltage side target power. Based on the relationship between the measured active power and the corresponding target power under at least two different load levels, candidate proportional correction parameters and candidate bias correction parameters for the high-voltage side correction path and the low-voltage side correction path are determined respectively.

8. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 1, characterized in that, The cross-validation and update module includes: The transient verification unit processes the transient verification data using the candidate electrical correction parameters corresponding to the high-voltage side correction path and the low-voltage side correction path respectively. Based on the corrected loss, candidate equivalent thermal conductivity parameters and candidate equivalent heat capacity parameters, the predicted oil temperature is determined, and the oil temperature residual between the predicted oil temperature and the measured oil temperature is obtained. The steady-state verification unit uses candidate electrical correction parameters corresponding to the high-voltage side correction path and the low-voltage side correction path to determine the electrical loss in the steady-state verification data, determines the heat inversion loss based on the candidate equivalent thermal conductivity parameters and temperature rise, and obtains the loss residual between the electrical loss and the heat inversion loss. The path determination unit determines the target correction path based on the oil temperature residual and loss residual corresponding to the high-pressure side correction path and the low-pressure side correction path.

9. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 8, characterized in that, The path determination unit determines the target correction path and updates the parameters according to the following process: The transient and steady-state verification data were processed using the previous effective parameters to obtain the reference oil temperature residual and the reference loss residual, respectively. The correction path that is lower than the corresponding residual threshold and the corresponding benchmark residual is selected as the candidate correction path. The comprehensive residual of each candidate correction path is determined based on the ratio of oil temperature residual to oil temperature residual threshold and the ratio of loss residual to loss residual threshold. When there is only one candidate correction path, the candidate correction path is determined as the target correction path; when there are two candidate correction paths, when the difference between the comprehensive residuals of the two candidate correction paths reaches the path discrimination threshold, the candidate correction path with the smaller comprehensive residual is determined as the target correction path. Update the candidate electrical correction parameters and candidate thermal parameters corresponding to the target correction path to the current valid parameters; if the target correction path is not determined, retain the previous valid parameters.

10. The energy efficiency monitoring system for oil-immersed transformers based on multi-parameter fusion according to claim 9, characterized in that, The energy efficiency and anomaly detection module includes: The energy efficiency determination unit corrects the input active power and output active power according to the current effective electrical correction parameters on the high-voltage side and the low-voltage side, respectively, and determines the transformer energy efficiency based on the ratio of the corrected output active power to the input active power. The abnormal index generation unit determines the electrical parameter offset index based on the offset of the current effective electrical correction parameter relative to the historical benchmark, determines the thermal conductivity decrease index based on the decrease of the current effective equivalent thermal conductivity parameter relative to the corresponding cooling state historical benchmark, and determines the loss growth index based on the common increase of electrical loss and heat inversion loss relative to the historical benchmark. The anomaly classification unit determines the main anomaly type of the indicator that reaches the corresponding threshold and has the largest normalized exceedance among the electrical parameter deviation index, thermal conductivity decrease index, and loss increase index when the transformer energy efficiency is lower than the energy efficiency benchmark. When two or three indicators reach the corresponding threshold, a composite anomaly label composed of the corresponding anomaly types is generated. The electrical parameter deviation index, thermal conductivity decrease index, and loss increase index correspond to electrical measurement link drift, heat dissipation performance degradation, and actual loss increase, respectively.

Images