Multi-dimensional anomaly early warning method and system for oil-immersed phase-separated power transformers
By constructing a six-layer operating condition matching system and multi-dimensional threshold calculation, the problem of low early warning accuracy of phase transformers in existing technologies has been solved. It has achieved dual early warning of faults in the main body and measurement circuit of 220kV~750kV phase transformers, thereby improving the safety and reliability of the power system.
Patent Information
- Application Number
- CN202611383242.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-08
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies fail to effectively distinguish the effects of temperature rise differences between active and reactive power, lack sufficient dimensions for matching operating conditions, fail to adapt cooling system warnings to reactive power conditions, and do not cover open-circuit hidden faults of dedicated CTs, resulting in low accuracy of phase transformer warnings and an inability to identify potential faults in real time.
By collecting and verifying the operating parameters and remote signaling signals of oil-immersed phase-separated power transformers, a six-layer operating condition matching system is constructed. Combined with active/reactive power correction coefficients, multi-dimensional thresholds are calculated to achieve dual early warning of faults in the main body and measurement circuit of 220kV~750kV phase-separated transformers, covering cooling system abnormalities and hidden faults such as open circuit of dedicated CT.
It improves the accuracy of early warning, reduces the false alarm rate, is compatible with phase transformers of different voltage levels, has wide applicability, is easy to operate and maintain, and enhances the power supply reliability and equipment safety of the power system.
Smart Images

Figure CN123121999A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power equipment condition monitoring technology, and in particular to a multi-dimensional anomaly early warning method and system for oil-immersed phase-separated power transformers. Background Technology
[0002] As a core piece of equipment in the transmission and transformation links of 220kV~750kV power systems, the operating status of phase transformers directly determines the reliability of power grid supply. Top oil temperature and winding temperature are core indicators reflecting the balance of internal heat generation and heat dissipation of the transformer. Abnormal temperatures usually indicate potential defects such as winding overheating and cooling system failure. If these defects are not detected and dealt with in a timely manner, they can easily lead to serious accidents such as equipment damage and power grid outages. At the same time, winding temperature measurement relies on dedicated current CTs to collect current signals. Open circuits in dedicated CTs can cause distortion in winding temperature calculations, making it impossible to reflect the true heating state of the windings. Such latent faults can only be detected through regular maintenance and cannot be detected in real time, posing a serious safety hazard.
[0003] Existing phase-splitting transformer temperature warning technologies generally have the following shortcomings, making it difficult to meet the actual engineering needs of 220kV~750kV phase-splitting transformers:
[0004] The current technology fails to differentiate the temperature rise effects of active and reactive power: It treats both active and reactive power as a single "load parameter," failing to clarify the fundamental difference in their impact on transformer temperature rise. Active power primarily generates heat through copper losses, while reactive power primarily generates heat through iron losses. Since the copper impedance of a 220kV~750kV phase transformer is much smaller than its iron impedance, under the same apparent power, a higher proportion of reactive power results in a higher actual transformer temperature rise. These factors prevent the current technology from accurately filtering similar historical data during operating condition matching, leading to distorted threshold calculations and consequently false alarms and missed alarms.
[0005] Insufficient dimensions for matching operating conditions: Existing technologies only match historical data using limited dimensions such as total load and season, without taking into account phase load, years of operation, and the differences in active and reactive power. This results in inconsistencies between the matched historical data and the heating characteristics of real-time operating conditions, making it impossible to accurately reflect the actual operating status of 220kV~750kV phase transformers.
[0006] The cooling system's early warning system is not adapted to the impact of reactive power: The existing cooler response threshold uses a fixed value and does not consider the cumulative effect of reactive power ratio on temperature rise. Under the high reactive power ratio operating conditions of 220kV~750kV, the actual heat dissipation demand of the cooler is higher than that under normal operating conditions. Using a fixed threshold will easily lead to a delay in early warning.
[0007] Hidden fault of open circuit in dedicated CT: Open circuit in dedicated CT for winding temperature measurement is a common measurement circuit fault in 220kV~750kV phase transformers. It will cause the winding temperature calculation to be distorted. However, the existing technology cannot identify this type of fault in real time. It can only be discovered through regular maintenance, which misses the opportunity to deal with the fault.
[0008] Therefore, it is necessary to develop a multi-dimensional anomaly early warning method and system for oil-immersed phase-separated power transformers that requires no additional hardware, is compatible with all voltage levels from 220kV to 750kV, and can cover both the transformer body and the measurement circuit for faults, in order to solve the above problems. Summary of the Invention
[0009] The purpose of this invention is to provide a multi-dimensional anomaly early warning method and system for oil-immersed phase-separated power transformers, in order to solve the problems of existing technologies such as failure to distinguish the differentiating effects of active and reactive power temperature rise, insufficient operating condition matching dimensions, cooler early warning not being adapted to reactive power conditions, failure to cover open-circuit hidden faults of dedicated CTs, and low early warning accuracy.
[0010] The technical solution adopted in this invention is as follows: the core parameters and thresholds are calibrated by referring to the operating data of 10 phase transformers with different voltage levels of 220kV~750kV, combined with the "Aging Guidelines for Oil-Immersed Power Transformers" DL / T 984-2018.
[0011] In a first aspect, the present invention provides a multi-dimensional anomaly early warning method for an oil-immersed phase-separated power transformer, comprising the following steps:
[0012] Step 1: Data Collection
[0013] Real-time acquisition of operating parameters and associated remote signaling signals of 220kV~750kV oil-immersed phase-separated power transformers. All parameters are acquired through the transformer's existing hardware sensors, requiring no additional acquisition equipment. Details are as follows:
[0014] 1. Operating parameters: Total active power P, Total reactive power Q, A / B / C three-phase active power P_φ, A / B / C three-phase reactive power Q_φ, A / B / C three-phase top oil temperature T_oil-φ, A / B / C three-phase winding temperature T_wind-φ, Transformer current operating years Y, Current month M.
[0015] 2. Associated remote signaling signals: remote signals for fan motor failure, cooler power circuit breaker status, and blower motor failure, used to jointly determine abnormal cooler power supply.
[0016] 3. Acquisition frequency: Default 5~10 seconds / time, which can be adjusted to 5~30 seconds / time according to the on-site operation and maintenance needs of 220kV~750kV. The remote signal acquisition frequency is consistent with the oil temperature acquisition frequency to balance data accuracy and hardware resource consumption. At the same time, it is adapted to the thermal inertia characteristics of 220kV~750kV phase transformers for 5~30 minutes to avoid abnormal missed detection due to too low acquisition frequency and waste of hardware resources due to too high acquisition frequency.
[0017] Step 2: Data Validation: The collected operating parameters and remote signaling signals undergo a four-layer validity filtering process to remove invalid data, ensuring the accuracy of subsequent calculations and adapting to the fault characteristics of 220kV~750kV phase transformers.
[0018] 1. Sensor fault verification: If any phase T_oil-φ < -10℃ or > 120℃, or any phase T_wind-φ < -10℃ or > 150℃, the corresponding sensor is determined to be faulty. The corresponding data is discarded, and only the valid data of the remaining two phases are used for subsequent calculations. If two or more phases are faulty, the current warning judgment is suspended and a sensor fault prompt is triggered to avoid invalid data interfering with the warning result.
[0019] 2. Load data verification: If the total P, total Q, or any phase P_φ or Q_φ is empty, or exceeds ±20% of the corresponding rated value, the load data is determined to be invalid, and the current warning judgment is suspended. The threshold is set based on the normal operating load fluctuation range of the 220kV~750kV phase transformer. Exceeding this range usually indicates an abnormal load or a data acquisition failure.
[0020] 3. Data jump verification: If the overall oil temperature difference between two adjacent acquisition times is >5℃, and the changes in total P and total Q are both <1%, it is determined to be a data jump (probably a communication disturbance or sensor false alarm), and all data at that adjacent acquisition time are removed.
[0021] 4. Remote signal jitter verification: If the remote signal switches ≥3 times within ±1 minute, it is determined to be remote signal jitter and will not be included in the cooler abnormal linkage judgment for the time being, so as to avoid false alarms caused by remote signal interference.
[0022] Step 3, Matching of six-layer operating conditions:
[0023] Based on valid data, and considering the differentiated impact of active and reactive power on the temperature rise of 220kV~750kV phase transformers (active power copper loss heating, reactive power iron loss heating), historical datasets consistent with the real-time operating condition heating characteristics are matched from the historical database to ensure the accuracy of the threshold calculation benchmark and to adapt to the thermal hysteresis characteristics of phase transformers. The specific matching conditions are as follows:
[0024] 1. Time matching: If the month of the real-time data is M, then only the data of months M-1, M, and M+1 in the historical database are searched to eliminate the influence of seasonal ambient temperature on oil temperature and ensure that the ambient temperature background of historical data and real-time data is consistent; among them, January matches December, January, and February, and December matches November, December, and January.
[0025] 2. Total active load matching: The historical total active power is within ±2% of the real-time total P, ensuring consistent copper loss heating level and conforming to the dominant influence of active power on temperature rise.
[0026] 3. Total reactive load matching: The historical total reactive power is within ±5% of the real-time total Q, ensuring consistent iron loss heating level and adapting to the operating characteristics of the reactive power ratio differences of 220kV~750kV phase transformers.
[0027] 4. Phase-by-phase active power load matching: The historical active power P_φ of each phase is within ±3% of the real-time corresponding phase P_φ, adapting to the phase-by-phase copper loss differences to ensure that the judgment benchmark for the uniformity of heating of each phase is consistent.
[0028] 5. Phase-by-phase reactive load matching: The historical reactive power Q_φ of each phase is within ±5% of the real-time corresponding phase Q_φ, which adapts to the phase-by-phase iron loss differences and conforms to the phase-by-phase operation characteristics of 220kV~750kV phase-by-phase transformers.
[0029] 6. Service life matching: The service life segment of the transformer corresponding to the historical data is consistent with the service life segment of the current service life Y, which is adapted to the heat dissipation efficiency differences of different aging stages of 220kV~750kV phase transformers.
[0030] 7. Thermal hysteresis adaptation: If the active power change rate is >10% / 10 minutes, extend the operating condition matching time window to 30 minutes to eliminate the operating condition matching distortion caused by the thermal inertia of the oil temperature of the 220kV~750kV phase transformer, and ensure that the matched historical data is completely consistent with the heating characteristics of the real-time operating condition.
[0031] Step 4: Data Preprocessing
[0032] The historical dataset matched in step three undergoes anomaly cleaning, and is stored in layers according to operating years to adapt to the aging characteristics of 220kV~750kV phase transformers, ensuring the accuracy of subsequent threshold calculations. The specific cleaning rules are customized adaptation rules tailored to the fault characteristics of 220kV~750kV phase transformers.
[0033] 1. Data with ΔT_max > 8℃ and P_φ change rate of each phase < 5% of the rated value is removed, as it is judged as transient disturbance or sensor failure data. Such data has no actual fault support and belongs to abnormal fluctuation. This rule is based on the statistical analysis of fault data from 10 220kV~750kV phase transformers, which can accurately filter out transient disturbances unique to phase transformers, rather than a general data cleaning rule. 2. Data with ΔT_oil-φ > 5℃ in any phase and ΔT_oil-φ < 2℃ in the other two phases is judged as abnormal data from the oil temperature sensor of that phase, avoiding omissions in cleaning due to sensor failure.
[0034] 3. Data with an overall oil temperature > 85℃ and total pressure (P) < 50% of the rated total pressure (P) are discarded as false overheating data without corresponding load support. Such data has no practical engineering significance and will interfere with threshold calculations. Specifically, ΔT_max = max(T_oil-A, T_oil-B, T_oil-C) - min(T_oil-A, T_oil-B, T_oil-C) is used to quantify the phase-separated oil temperature difference; ΔT_oil-φ = |T_oil-φ - T_avg| (φ = A, B, C), T_avg = (T_oil-A + T_oil-B + T_oil-C) / 3 is used to quantify the deviation of a single phase from the overall oil temperature.
[0035] 4. Layered Storage: The cleaned historical dataset is stored in layers according to the operating years. Different years correspond to different temperature rise margins ΔT_age. The specific division and margin values are in line with the aging characteristics of the entire life cycle of 220kV~750kV phase transformers: 0~1 years, ΔT_age=0℃; 1~5 years, ΔT_age=1℃; 5~10 years, ΔT_age=2~3℃; >10 years, ΔT_age=3~5℃. This margin has been verified by experiments on multiple 220kV~750kV phase transformers and can accurately compensate for the normal temperature rise caused by equipment aging, avoiding false alarms caused by aging.
[0036] Step 5: Threshold Calculation (including reactive power adaptation and CT open circuit verification)
[0037] Based on the historical dataset after stratified cleaning, various early warning thresholds are calculated, active / reactive power correction coefficients are introduced, and a new winding-oil surface temperature ratio threshold is added to cover both body faults and measurement circuit faults. All thresholds are adapted to the operating characteristics of 220kV~750kV phase transformers. The specific calculation method is as follows:
[0038] 1. Overall Oil Temperature Threshold: Calculate the 95th percentile T95_total of the historical overall oil temperature dataset for each layer, and superimpose the temperature rise margin ΔT_age for the corresponding operating years to obtain the overall oil temperature warning threshold T_total = T95_total + ΔT_age. The selection criteria for the 95th percentile are: to remove extreme outliers from the historical data, ensuring that the threshold can cover normal fluctuations and accurately identify anomalies. The calculation method uses the Excel function PERCENTILE.EXC(data range, 0.95) or the Python function numpy.percentile(data, 95). The 95th percentile calculation must be based on the historical data of the phase transformer layers, and extreme values caused by phase load imbalance must be removed. It is adapted to the phase operation characteristics of 220kV~750kV phase transformers to ensure that the calculation results are consistent and accurate, and compatible with commonly used data processing equipment in industrial sites.
[0039] 2. Phase oil temperature difference threshold: Calculate the 95th quantile T95_Δmax of the historical ΔT_max dataset for each phase and the 95th quantile T95_Δoil of the historical ΔT_oil-φ dataset. Add the temperature rise margin ΔT_age for the corresponding operating years to obtain the ΔT_max warning threshold T_Δmax = T95_Δmax + ΔT_age and the ΔT_oil-φ warning threshold T_Δoil = T95_Δoil + ΔT_age.
[0040] 3. Single-phase cooler failure threshold: After the cooler is started, the "difference in temperature drop" between any phase oil temperature and other phase oil temperatures is greater than 2℃, and the phase current deviation is ≤5%, with a duration of ≥10 minutes; the phase current deviation is calculated as (maximum value - minimum value) / average value × 100%. This threshold is set based on the oil temperature change pattern after the failure of the 220kV~750kV phase transformer cooler, which can accurately identify single-phase cooler failure faults.
[0041] 4. Cooler normal response threshold:
[0042] (1) After the first set of fans is started: the oil temperature drops by ≥4.0℃ in 10 minutes, ≥4.7℃ in 30 minutes, and the average rate of drop in the first 10 minutes is ≥0.40℃ / min.
[0043] (2) After the two sets of fans are started (starting conditions are set according to the on-site operation procedures): the oil temperature drop is ≥3.5℃ in 10 minutes and ≥5.5℃ in 30 minutes; the overall oil temperature is high when the two sets of fans are started, and the temperature difference between the oil and the environment is large, resulting in a relatively slow initial heat dissipation rate. Therefore, the 10-minute drop threshold is appropriately relaxed.
[0044] (3) Active / Reactive Power Correction Coefficient: When the active power increases by 5MW, the 10-minute temperature drop threshold of 1 group of wind turbines decreases by 0.8℃, and the 10-minute temperature drop threshold of 2 groups of wind turbines decreases by 1.2℃; when the reactive power increases by 10Mvar, the 10-minute temperature drop threshold of 1 group of wind turbines decreases by 0.5℃, and the 10-minute temperature drop threshold of 2 groups of wind turbines decreases by 0.8℃; this correction coefficient is adapted to the effect of the differential temperature rise of active / reactive power in 220kV~750kV phase transformers, and solves the problem of early warning lag under high reactive power ratio operating conditions.
[0045] 5. Winding-Oil Temperature Proportional Threshold (for CT Open Circuit Early Warning): Based on historical datasets, a correlation model between the winding temperature and oil surface temperature of the same phase under normal operating conditions is obtained through linear regression analysis. A proportionality coefficient k is set (range 0.01~0.03℃ / MVA). The normal threshold condition is T_wind-φ - T_oil-φ ≥ k×S_φ (S_φ=√(P_φ²+Q_φ²)), and T_wind-φ > T_oil-φ. The calibration process of k covers the historical operating data of 10 phase transformers of different voltage levels from 220kV to 750kV, covering the 20%~100% rated power range, to ensure the accuracy of the proportional relationship determination.
[0046] 6. Absolute value fallback threshold: To avoid missing extreme anomalies, a fixed fallback threshold is set to adapt to emergency fault scenarios of 220kV~750kV phase transformers: if the overall oil temperature is ≥85℃, or ΔT_max ≥5℃, or any phase ΔT_oil-φ ≥3℃, the corresponding level of warning will be triggered directly.
[0047] Step Six: Multi-dimensional Early Warning Judgment (Dual Fault Coverage)
[0048] Combining real-time operating parameters and various thresholds calculated in step five, along with trend verification and remote signaling linkage, a dual early warning judgment of "main transformer body fault + measurement circuit fault" is performed, adapting to fault types of 220kV~750kV phase transformers. The main transformer body temperature early warning adopts a dual threshold verification mechanism to avoid false judgments and missed judgments. The specific judgment logic is as follows:
[0049] 1. Cooler power supply abnormality warning: If the real-time response data is lower than the normal response threshold of the cooler after the fan is switched on or off, and the "cooler power supply failure / fan motor failure" remote signal is triggered within ±10 minutes, the cooler power supply abnormality warning will be triggered (blue audible and visual alarm + text prompt "suspected [abnormal phase] cooler power supply failure"), prompting maintenance personnel to check the corresponding phase power circuit breaker and fan motor status.
[0050] 2. Single-phase cooler failure warning: If the single-phase cooler failure threshold is met, a single-phase cooler failure warning will be triggered (blue audible and visual alarm + text prompt "Suspected [abnormal phase] cooler failure"), prompting you to check the corresponding phase fan motor and heat dissipation channel.
[0051] 3. Cooler operating condition abnormality warning: If the normal response threshold of the cooler is not met and there is no associated remote signaling trigger, a normal cooler operating condition abnormality warning (blue audible and visual alarm) will be triggered, prompting you to check the overall operating status of the cooling system.
[0052] 4. Level 1 Warning (Early Local Anomaly): Meets the following conditions (ΔT_max > T_Δmax or any phase ΔT_oil-φ > T_Δoil), and the corresponding phase oil temperature difference trend verification passes (ΔT_max continuously rises ≥1℃ within 5~8 minutes, or any phase ΔT_oil-φ continuously rises ≥0.8℃, and the corresponding phase P_φ change <3%, Q_φ change <5%), while the overall oil temperature ≤T_total; the corresponding anomaly type is single-phase cooling branch fault, phase load imbalance, and local winding aging, which is suitable for the local fault characteristics of 220kV~750kV phase transformers. The handling solution is to check the operating status of the phase cooling system and the phase load distribution.
[0053] 5. Level II Early Warning (Mid-term Overall Anomaly): The overall oil temperature is greater than T_total, and (ΔT_max > T_Δmax or the phase oil temperature difference trend verification is passed), while the overall oil temperature is less than 85℃; the corresponding anomaly type is overall temperature rise superimposed on local anomaly, and the handling plan is to reduce the load of the 220kV~750kV phase transformers and focus on investigating the abnormal phases.
[0054] 6. Level 3 Warning (Emergency Overheating Anomaly): Meets the following conditions: "Overall oil temperature ≥ 85℃, or (overall oil temperature ≥ 70℃ and ΔT_max ≥ 5℃), or (overall oil temperature ≥ 70℃ and any phase ΔT_oil-φ ≥ 3℃)"; The corresponding anomaly type is severe overheating and potential local short circuit. The handling plan is to shut down the machine for inspection to avoid equipment damage.
[0055] 7. Dedicated CT Open Circuit Warning (Measurement Circuit Fault): If for 5-10 consecutive minutes the condition “|T_wind-φ -T_oil-φ| < 2℃” is met, and the corresponding phase active power P_φ > 20% of the rated power (excluding normal scenarios with small temperature differences under low load), while not meeting the above-mentioned first-level, second-level, and third-level warning conditions for abnormal oil temperature, then the “Dedicated CT Open Circuit Warning for Winding Temperature Measurement” will be triggered (blue audible and visual alarm + text prompt “Suspected [abnormal phase] winding temperature gauge matching CT open circuit, check circuit continuity”), filling the gap in existing technology for warning of this type of hidden fault.
[0056] 8. Distinguishing between audible and visual warnings: To facilitate quick identification of warning types by maintenance personnel of 220kV~750kV substations, different audible and visual prompts are used: blue for cooler-related warnings and dedicated CT open circuit warnings, green for Level 1 warnings, yellow for Level 2 warnings, and red for Level 3 warnings. The audible and visual alarms continue until the fault is resolved, ensuring a rapid response from maintenance personnel.
[0057] Step 7: Early Warning Response and Historical Database Update
[0058] 1. Early Warning and Handling: Record detailed information for each early warning event, including early warning time, early warning level, real-time operating parameters, remote signaling status, and abnormal phase, forming an early warning ledger for easy follow-up and optimization; at the same time, output corresponding handling prompts for different early warning types, adapting to the operation and maintenance scenarios of 220kV~750kV phase transformers, ensuring that operation and maintenance personnel can handle the situation quickly.
[0059] 2. Historical Database Update: The real-time operating parameters that have passed this verification are added to the historical database according to the hierarchical rules in step four to achieve dynamic updates of historical data; at the same time, old data older than 5 years is automatically removed to avoid old data affecting the accuracy of operating condition matching and threshold calculation, thus adapting to the long-term operating requirements of 220kV~750kV phase transformers.
[0060] 3. Backup threshold scheme: When the number of historical dataset samples matched in the historical database is less than 20, it is determined to be statistically invalid. At this time, the backup threshold is activated: the overall oil temperature warning threshold is changed to the historical average oil temperature of the corresponding operating years + 8℃, and the phase oil temperature difference threshold is changed to the historical average ΔT_max + 0.5℃ and the historical average ΔT_oil-φ + 0.5℃ of the corresponding operating years. This ensures that accurate warnings can still be achieved when historical data is insufficient, thereby improving system reliability.
[0061] Secondly, the present invention provides a multi-dimensional anomaly early warning system for oil-immersed phase-separated power transformers, used to execute the multi-dimensional anomaly early warning method for oil-immersed phase-separated power transformers described above.
[0062] The beneficial effects of the present invention: The multi-dimensional anomaly early warning method and system for oil-immersed phase-splitting power transformers provided by the present invention have the following beneficial effects and meet the actual engineering needs of 220kV~750kV phase-splitting transformers.
[0063] Improved early warning accuracy: Clearly distinguish the differential temperature rise impact of active and reactive power, construct a six-layer precise operating condition matching system, introduce active / reactive power correction coefficients, significantly reduce false alarm and missed warning rates, and solve the pain point of existing technology in 220kV~750kV high reactive power ratio operating conditions with false alarms and missed warnings.
[0064] Comprehensive early warning coverage: It achieves dual early warning for "main transformer body faults + measurement circuit faults", covering not only body faults such as cooling system abnormalities and oil temperature abnormalities, but also adding a dedicated CT open circuit hidden fault early warning, which is compatible with the fault types of 220kV~750kV phase transformers.
[0065] Highly practical for engineering: No new hardware equipment is required; existing sensors, monitoring systems and remote signaling interfaces of 220kV~750kV phase transformers can be reused; it is compatible with the IEC 61850 protocol; it can be directly embedded into existing substation monitoring systems; low implementation cost; no need to modify existing equipment.
[0066] Wide adaptability: It can be adapted to phase transformers of all voltage levels from 220kV to 750kV. It automatically adjusts the matching deviation and correction coefficient for the differences in copper impedance and iron impedance of different voltage levels. It is also compatible with conventional field control logic such as air-cooled switching and reactive power compensation.
[0067] High ease of operation and maintenance: It outputs corresponding handling prompts, and the warning sound and light are clearly distinguishable, making it easy for 220kV~750kV substation operation and maintenance personnel to quickly identify the fault type and locate the fault. This invention breaks through the traditional understanding of "unified treatment of active and reactive power", clearly defines the different temperature rise effects of the two and constructs a six-layer operating condition matching system. At the same time, it adds a dedicated CT open circuit latent fault warning, forming a dual warning system of "body + measurement circuit".
[0068] High safety and reliability: Through strict data verification, abnormal data cleaning and dual threshold verification mechanisms, invalid data and interference signals are effectively filtered out, and the false alarm rate is significantly reduced. In test scenarios, it can issue an early warning more than 1 hour before a local anomaly develops into a serious fault. It can identify local anomalies and identify compound faults in advance, and can promptly detect potential faults and prompt handling, avoiding serious accidents such as equipment damage and power grid outages, and improving the power supply reliability and equipment operation safety of the 220kV~750kV power system.
[0069] Highly scalable: The system architecture is designed to fit the development trend of 220kV~750kV phase-splitting transformers. It can flexibly add early warning dimensions, adjust threshold parameters, and adapt to new phase-splitting transformer models according to subsequent operation and maintenance needs, without the need for large-scale modification of the existing system. It has good scalability and long-term applicability. Attached Figure Description
[0070] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.
[0071] Figure 1This is a schematic diagram of the module structure of the multi-dimensional anomaly early warning system for oil-immersed phase-separated power transformers of the present invention.
[0072] Diagram Explanation: 1-Data Acquisition Module, 2-Data Verification Module, 3-Operating Condition Matching Module, 4-Data Preprocessing Module, 5-Threshold Calculation Module, 6-Early Warning Judgment Module, 7-Handling Prompt Module, 8-Historical Database Update Module, 9-Historical Database. Detailed Implementation
[0073] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The technical solutions provided by various embodiments of this invention will be described in detail below with reference to the accompanying drawings.
[0074] This invention provides a multi-dimensional anomaly early warning method for oil-immersed phase-splitting power transformers. This method relies on existing monitoring sensors (oil temperature sensor, current sensor, voltage sensor, winding temperature sensor), monitoring devices, remote signaling interfaces, and historical databases for 220kV~750kV oil-immersed phase-splitting transformers. It cannot operate independently of the industrial hardware. Six-level operating condition matching is a necessary technical means to achieve early warning of single-phase partial faults in phase-splitting transformers; none of these can be omitted. The method includes the following steps:
[0075] Step 1: Data Acquisition: Real-time acquisition of operating parameters and associated remote signaling signals for 220kV~750kV oil-immersed phase-separated power transformers. Operating parameters include total active power P, total reactive power Q, A / B / C three-phase active power P_φ, A / B / C three-phase reactive power Q_φ, A / B / C three-phase top oil temperature T_oil-φ, A / B / C three-phase winding temperature T_wind-φ, transformer current operating years Y, and current month M. Associated remote signaling signals include fan motor fault, cooler power supply circuit breaker status, and fan motor fault remote signaling. The top oil temperature of the three phases A / B / C, T_oil-φ, is collected by the phase transformer oil temperature sensor. The overall oil temperature is the arithmetic average of the three-phase top oil temperatures, T_avg=(T_oil-A+T_oil-B+T_oil-C) / 3; T_oil-A, T_oil-B, and T_oil-C are the top oil temperatures of the three phases A / B / C, respectively. The default acquisition frequency is 5~10 seconds / time, or can be adjusted to 5~30 seconds / time according to the on-site operation and maintenance requirements of 220kV~750kV. The remote signal acquisition frequency is consistent with the oil temperature acquisition frequency.
[0076] Specifically, in step one, the remote signal acquisition is adapted to the IEC 61850 protocol for power field applications, supporting seamless integration with the monitoring system of 220kV~750kV substations; the phase active power P_φ and phase reactive power Q_φ are acquired through existing current and voltage sensors of the 220kV~750kV phase transformers, and the winding temperature T_wind-φ is acquired through existing winding temperature sensors; the acquisition frequency is adjusted according to the on-site operation and maintenance requirements, with an adjustment range of 5~30 seconds / time.
[0077] Step 2: Data Validity Verification: The collected operating parameters and remote signaling signals are filtered for validity, eliminating invalid data caused by sensor malfunctions, communication disturbances, and data jumps. Specific rules are as follows: If any phase T_oil-φ < -10℃ or > 120℃, or any phase T_wind-φ < -10℃ or > 150℃, it is determined that the corresponding sensor is faulty, and the corresponding data is discarded; only the remaining two phases' valid data are used for subsequent calculations. If two or more phases are faulty, the current warning judgment is suspended and a sensor fault prompt is triggered. If total P, total Q, or any phase P_φ or Q_φ is empty, or exceeds ±20% of the corresponding rated value, the load data is determined to be invalid, and the current warning judgment is suspended. If the overall oil temperature difference between two adjacent acquisition times is > 5℃, and the changes in total P and total Q are both < 1%, it is determined to be a data instantaneous jump, and all data from that adjacent acquisition time are discarded. If the remote signaling signal continuously jitters, switching ≥ 3 times within ± 1 minute, it is determined to be invalid remote signaling and is temporarily excluded from the linkage judgment.
[0078] Step 3: Operating Condition Matching: Based on the collected valid operating parameters and considering the differentiated impact of active and reactive power on the main transformer temperature rise, historical datasets meeting the following conditions are matched from the historical database to adapt to the thermal hysteresis characteristics of 220kV~750kV phase transformers:
[0079] Time matching: If the month of the real-time data is M, then only the data of months M-1, M, and M+1 in the historical database will be searched; where January matches December, January, and February, and December matches November, December, and January.
[0080] Total load matching: The historical total active power is within ±2% of the real-time total P, and the historical total reactive power is within ±5% of the real-time total Q.
[0081] Phase-by-phase load matching: The historical active power P_φ of each phase is within ±3% of the real-time corresponding phase P_φ, and the historical reactive power Q_φ of each phase is within ±5% of the real-time corresponding phase Q_φ.
[0082] Operating life matching: The transformer operating life segment corresponding to the historical data is consistent with the current operating life segment Y. If the active power change rate is >10% / 10 minutes, extend the operating condition matching time window to 30 minutes to eliminate the thermal inertia interference of oil temperature in the 220kV~750kV phase transformer.
[0083] Step 4: Data Preprocessing: Perform anomaly cleaning on the historical dataset matched in Step 3, and store it in layers according to the operating years. The operating years and corresponding temperature rise margin ΔT_age are preset as follows: 0~1 years, ΔT_age=0℃; 1~5 years, ΔT_age=1℃; 5~10 years, ΔT_age=2~3℃; >10 years, ΔT_age=3~5℃. This division of operating years is based on the aging characteristics of the entire life cycle of 220kV~750kV phase transformers.
[0084] The rules for abnormal data cleaning are customized rules adapted to the fault characteristics of 220kV~750kV phase transformers:
[0085] Data with ΔT_max > 8℃ and P_φ change rate of each phase < 5% of the rated value were removed and determined to be transient disturbance or sensor failure data.
[0086] Data that excludes any phase with ΔT_oil-φ > 5℃ and the other two phases with ΔT_oil-φ < 2℃ is considered abnormal data from the oil temperature sensor for that phase.
[0087] Data with an overall oil temperature > 85℃ and a total pressure (P) < 50% of the rated total pressure (P) are discarded as false overheating data without corresponding load support. Specifically, ΔT_max = max(T_oil-A, T_oil-B, T_oil-C) - min(T_oil-A, T_oil-B, T_oil-C) is used to quantify the phase-separated oil temperature difference; ΔT_oil-φ = |T_oil-φ - T_avg| (φ = A, B, C), T_avg = (T_oil-A + T_oil-B + T_oil-C) / 3 is used to quantify the deviation of a single phase from the overall oil temperature.
[0088] Specifically, the division of the operating lifespan in step four is based on the aging characteristics of the entire life cycle of 220kV~750kV phase-splitting transformers: 0~1 years is the new commissioning stage, with no aging of windings and oil, and optimal heat dissipation efficiency; 1~5 years is the stable operation stage, with slight aging of windings and oil, and copper loss increase ≤5%; 5~10 years is the moderate aging stage, with increased winding resistance, deterioration of oil, and a decrease in heat dissipation efficiency of about 10%; >10 years is the deep aging stage, with a decrease in cooling system efficiency of more than 15%, and a significant increase in copper and iron losses. The core basis for determining the temperature rise margin ΔT_age is: the rate of change of winding resistance, the rate of change of oil dielectric loss, and the rate of decay of cooling system efficiency of 220kV~750kV phase-splitting transformers in different operating lifespans.
[0089] Step 5: Threshold Calculation
[0090] Overall oil temperature threshold: Calculate the 95th percentile T95_total of the historical overall oil temperature dataset for each layer, and add the temperature rise margin ΔT_age for the corresponding operating years to obtain the overall oil temperature warning threshold T_total = T95_total + ΔT_age.
[0091] Phase-separated oil temperature difference thresholds: Calculate the 95th quantile T95_Δmax of the historical ΔT_max dataset for each layer and the 95th quantile T95_Δoil of the historical ΔT_oil-φ dataset, and superimpose the temperature rise margin ΔT_age for the corresponding operating years to obtain the ΔT_max warning threshold T_Δmax = T95_Δmax + ΔT_age and the ΔT_oil-φ warning threshold T_Δoil = T95_Δoil + ΔT_age.
[0092] Single-phase cooler failure threshold: After the cooler is started, the "difference in temperature drop" between any phase oil temperature and other phase oil temperatures is greater than 2℃, and the phase current deviation is ≤5%, with a duration of ≥10 minutes; the phase current deviation is calculated as (maximum value - minimum value) / average value × 100%.
[0093] Cooler normal response thresholds: After the first set of fans starts, the oil temperature drop should be ≥4.0℃ in 10 minutes, ≥4.7℃ in 30 minutes, and the average drop rate in the first 10 minutes should be ≥0.40℃ / min; After the second set of fans starts, the starting conditions should be set according to the on-site operating procedures, with an oil temperature drop of ≥3.5℃ in 10 minutes and ≥5.5℃ in 30 minutes; Active load correction factor: When the active power increases by 5MW, the 10-minute drop threshold for the first set of fans decreases by 0.8℃, and the 10-minute drop threshold for the second set of fans decreases by 1.2℃; Reactive load correction factor: When the reactive power increases by 10Mvar, the 10-minute drop threshold for the first set of fans decreases by 0.5℃, and the 10-minute drop threshold for the second set of fans decreases by 0.8℃, adapting to the operating characteristics of the reactive power ratio difference of the 220kV~750kV phase transformer.
[0094] Winding-oil surface temperature ratio threshold: Based on historical datasets, calculate the range of difference between the winding temperature and oil surface temperature of the same phase under normal operating conditions, and set the ratio coefficient k (range 0.01~0.03℃ / MVA). The normal threshold condition is T_wind-φ - T_oil-φ ≥ k×S_φ (S_φ=√(P_φ²+Q_φ²)), and T_wind-φ > T_oil-φ.
[0095] Preset absolute value fallback thresholds: The overall oil temperature absolute value warning threshold is 85℃, the ΔT_max fallback threshold is 5℃, and the fallback threshold for any phase ΔT_oil-φ is 3℃.
[0096] Specifically, the 95th percentile calculation in step five uses the Excel function PERCENTILE.EXC or the Python function numpy.percentile. The 95th percentile calculation is based on historical data of phase-separated transformers, and extreme values caused by phase load imbalance are eliminated to adapt to the phase-separated operation characteristics of 220kV~750kV phase-separated transformers. The calibration process for the proportional coefficient k in step five is as follows: based on historical operating data of 10 phase-separated transformers of different voltage levels from 220kV to 750kV, covering different load ranges and 20%~100% of rated power, a correlation model between T_wind-φ and T_oil-φ and S_φ is obtained through linear regression analysis. Finally, the value range of k is determined to be 0.01~0.03℃ / MVA to ensure the accuracy of the proportional relationship determination under normal operating conditions.
[0097] Step Six: Multi-dimensional Early Warning Judgment: Based on real-time operating parameters and various thresholds calculated in Step Five, combined with trend verification and remote signaling linkage, determine whether an early warning is triggered and the warning level. Adapt to the fault characteristics of 220kV~750kV phase transformers, the main transformer body temperature early warning adopts a dual threshold verification mechanism.
[0098] Cooler power supply abnormality warning: If the real-time response data is lower than the normal response threshold of the cooler after the fan is switched on or off, and a remote signal for cooler power supply failure / fan motor failure is triggered within ±10 minutes, a cooler power supply abnormality warning will be triggered, with a blue audible and visual alarm and a text prompt indicating a suspected abnormal phase cooler power supply failure.
[0099] Single-phase cooler failure warning: If the single-phase cooler failure threshold is met, a single-phase cooler failure warning will be triggered, with a blue audible and visual alarm and a text message indicating a suspected abnormal phase cooler failure.
[0100] Cooler operating condition abnormality warning: If the normal response threshold of the cooler is not met, and there is no associated remote signaling trigger, a normal cooler operating condition abnormality warning will be triggered, and a blue audible and visual alarm will be activated.
[0101] Level 1 warning, early local anomaly: meets the following conditions: ΔT_max > T_Δmax or any phase ΔT_oil-φ > T_Δoil, and the corresponding phase oil temperature difference trend verification passes. Within 5-8 minutes, ΔT_max continuously rises by ≥1℃, or any phase ΔT_oil-φ continuously rises by ≥0.8℃, and the corresponding phase P_φ change is <3% and Q_φ change is <5%, while the overall oil temperature is ≤T_total. The corresponding anomaly types are single-phase cooling branch fault, phase load imbalance, and local winding aging, and are applicable to the local fault characteristics of 220kV~750kV phase transformers.
[0102] Level 2 warning, mid-term overall anomaly: meets the following conditions: overall oil temperature > T_total, and ΔT_max > T_Δmax or phase oil temperature difference trend verification passes, while overall oil temperature < 85℃; the corresponding anomaly type is overall temperature rise superimposed on local anomaly, which is suitable for the overall operation anomaly scenario of 220kV~750kV phase transformers.
[0103] Level 3 warning, emergency overheating anomaly: meets the following conditions: overall oil temperature ≥85℃, or overall oil temperature ≥70℃ and ΔT_max ≥5℃, or overall oil temperature ≥70℃ and any phase ΔT_oil-φ ≥3℃; the corresponding anomaly type is severe overheating and local short circuit hazard, and is suitable for emergency fault scenarios of 220kV~750kV phase transformers.
[0104] Dedicated CT open circuit warning: If for 5~10 consecutive minutes the condition “|T_wind-φ - T_oil-φ| < 2℃” is met, and the corresponding phase active power P_φ > 20% of the rated power, but the warning conditions for Level 1, Level 2, and Level 3 warnings are not met, then the “Dedicated CT open circuit warning for winding temperature measurement will be triggered, with a blue audible and visual alarm and a text prompt indicating that the CT for the suspected abnormal phase winding temperature meter is open, and the circuit continuity status should be checked” will be activated.
[0105] Specifically, the abnormality types and handling solutions corresponding to each warning level in step six are as follows: Level 1 warning corresponds to single-phase cooling branch faults, phase load imbalance, and partial aging of windings. The handling solution is to check the operating status of the phase cooling system and the phase load distribution of the 220kV~750kV phase transformers; Level 2 warning corresponds to overall temperature rise superimposed on local abnormalities. The handling solution is to reduce the transformer load and focus on checking the abnormal phases; Level 3 warning corresponds to severe overheating and potential local short circuit hazards. The handling solution is to shut down the machine for inspection; the blue warnings related to the cooler and the open circuit warning of the dedicated CT both indicate that the hardware status of the corresponding circuit should be checked first.
[0106] Step 7, Early Warning Handling and Historical Database Update: Record early warning events, handling results, and remote sensing related data. Add the real-time operating parameters that have passed this verification to the historical database according to the hierarchical rules in Step 4, and continuously remove old data older than 5 years. When the sample size of the historical dataset matched in the historical database is <20, it is judged as statistically invalid. At this time, the overall oil temperature early warning threshold is changed to the historical average oil temperature of the corresponding operating years + 8℃, and the phase oil temperature difference threshold is changed to the historical average ΔT_max + 0.5℃ and the historical average ΔT_oil-φ + 0.5℃ of the corresponding operating years.
[0107] The above method is naturally compatible with the conventional control logic of 220kV~750kV phase transformers, including automatic switching of air-cooled systems, switching of on-load tap changers, switching of reactive power compensation, and sudden load increases and decreases. The above control logic will not cause the method to make misjudgments. The basis for the compatibility is that the above control logic makes the three-phase oil temperature change synchronously without changing the phase oil temperature difference. Moreover, trend verification and load stability constraints can filter the impact of instantaneous fluctuations. At the same time, because it clearly distinguishes the different impacts of active and reactive power on temperature rise, it can still maintain accurate early warning under high reactive power ratio conditions, avoiding false alarms and missed alarms caused by existing technologies ignoring this difference.
[0108] Please see Figure 1The present invention also provides a multi-dimensional anomaly early warning system for oil-immersed phase-separated power transformers, used to execute the multi-dimensional anomaly early warning method for oil-immersed phase-separated power transformers described above. The system includes the following modules bound to the physical hardware of the 220kV~750kV phase-separated transformer: data acquisition module 1, data verification module 2, operating condition matching module 3, data preprocessing module 4, threshold calculation module 5, early warning judgment module 6, handling prompt module 7, historical database update module 8, and historical database 9. The connections between the modules are as follows: Data acquisition module 1 is electrically connected to data verification module 2; data verification module 2 is electrically connected to operating condition matching module 3 and early warning judgment module 6 respectively; operating condition matching module 3 is electrically connected to data preprocessing module 4; data preprocessing module 4 is electrically connected to threshold calculation module 5; threshold calculation module 5 is electrically connected to early warning judgment module 6; early warning judgment module 6 is electrically connected to handling prompt module 7; data verification module 2 and early warning judgment module 6 are both electrically connected to historical database update module 8; historical database update module 8 is electrically connected to historical database 9; data acquisition module 1 is also directly electrically connected to historical database update module 8 to transmit verified real-time data to historical database update module 8, enabling dynamic updates of historical database 9. The connection relationships between the modules are based on the signal transmission path design of 220kV~750kV phase-splitting transformers. The direct connection between the data acquisition module and the historical database update module avoids phase-splitting data transmission delays and adapts to the thermal inertia response requirements of phase-splitting transformers.
[0109] Data acquisition module: Employs monitoring sensors suitable for 220kV~750kV phase-segment power transformers, including oil temperature sensors, current sensors, voltage sensors, winding temperature sensors, and remote signal acquisition interfaces, compatible with the IEC61850 protocol. It is used to acquire in real-time the total P, total Q, phase P_φ, phase Q_φ, phase T_oil-φ, phase T_wind-φ, operating years Y, month M, and associated remote signaling signals of the phase-segment transformer, transmitting them to the data verification module. The acquisition frequency is adjustable, defaulting to 5~10 seconds / time, and supports offline data caching, allowing for ≥72 hours of data caching in case of communication interruption.
[0110] Data verification module: Electrically connected to the data acquisition module, it is used to verify the validity of the acquired operating parameters and remote signaling signals, and to remove invalid data caused by sensor failure, communication disturbance, and data jump. If sensor failure, invalid data, or remote signaling jitter is detected, the module outputs the corresponding prompt information and transmits the valid data to the operating condition matching module.
[0111] Operating condition matching module: Electrically connected to the data verification module and the historical database respectively. The core is to adapt to the differential temperature rise effect of active power and reactive power. Based on valid data, it matches historical datasets from the historical database that meet the conditions of "same month ± 1 month + total active power + total reactive power + phase active power + phase reactive power + years of operation". It supports thermal hysteresis adaptation and automatically extends the matching time window when the active power fluctuation is large. The matched historical datasets are then transmitted to the data preprocessing module.
[0112] Data preprocessing module: Electrically connected to the operating condition matching module, used to clean abnormal data in the matched historical dataset, customize rules according to step four, and store the data in layers according to the operating years. It calculates the 95th percentile of the historical data in each layer and the temperature rise margin ΔT_age of the corresponding operating years, and transmits it to the threshold calculation module.
[0113] Threshold calculation module: Electrically connected to the data preprocessing module, it is used to calculate the overall oil temperature warning threshold, the phase oil temperature difference warning threshold, the single-phase cooler failure threshold, the cooler normal response threshold, the winding-oil surface temperature ratio threshold, and the absolute value fallback threshold, and transmit them to the warning judgment module.
[0114] Early warning judgment module: Electrically connected to the data verification module and threshold calculation module respectively, it is used to combine real-time valid data and various thresholds to perform trend verification, remote signaling linkage and early warning level judgment, output the main transformer body fault early warning and the measurement circuit fault early warning, and transmit the early warning signal to the power monitoring terminal; the early warning signal is distinguished by different audible and visual prompts: cooler related early warning and dedicated CT open circuit early warning is blue, level one early warning is green, level two early warning is yellow, and level three early warning is red; it has an IEC 61850 communication interface and can upload early warning signals to the 220kV~750kV power dispatching system.
[0115] Handling prompt module: Electrically connected to the early warning judgment module, it outputs corresponding handling prompt information for different early warning types. For example, if the cooler power supply is abnormal, it prompts to check the corresponding phase power supply circuit breaker / fan motor. If the single-phase cooler fails, it prompts to check the corresponding phase fan motor / heat dissipation channel. If the dedicated CT is open, it prompts to check the corresponding phase winding temperature, the dedicated CT circuit continuity and wiring status. For local / overall abnormalities, it prompts load adjustment and fault diagnosis according to the early warning level. It is suitable for the operation and maintenance scenarios of 220kV~750kV phase transformers.
[0116] Historical database update module: Electrically connected to the data verification module and the historical database respectively, it is used to add the real-time valid data that has passed the current verification to the historical database in layers according to the running years, and automatically remove old data older than 5 years. When the sample size of the historical dataset is <20, it outputs a threshold adaptation prompt and enables the backup threshold scheme.
[0117] Historical database: Used for hierarchical storage of historical operating parameters, remote signaling data and early warning event records after cleaning. It supports fast data query, hierarchical storage and rolling updates. It can be deployed on edge computing devices of 220kV~750kV substations or power monitoring center servers, and is suitable for different deployment scenarios on site. The historical operating parameters include active / reactive power differential matching identifiers and winding-oil surface temperature ratio data.
[0118] This system is compatible with 220kV, 330kV, 500kV, and 750kV oil-immersed phase-splitting transformers. It automatically adjusts the matching deviation range of total / phase active and reactive power, as well as the correction coefficient of the cooler response threshold, to address the differences in copper and iron impedance among transformers of different voltage levels. Simultaneously, it automatically calibrates the proportional coefficient k based on the winding structure and heat dissipation characteristics of different transformers, ensuring the accuracy of winding-oil surface temperature ratio verification. It maintains precise early warning under high reactive power ratios and different load conditions, adapting to the operational needs of phase-splitting transformers across all voltage levels from 220kV to 750kV.
[0119] To verify the effectiveness and accuracy of the method and system of this invention, a 220kV oil-immersed phase-separated transformer and a 500kV oil-immersed phase-separated transformer were selected as test objects. Based on the historical operating data of the two transformers for 5 years (including normal operating conditions, local anomalies, cooling system failures, open circuit failures of dedicated CTs, etc.), a field verification of no less than 75 days was carried out. The specific implementation process and results are as follows. All parameters are consistent with the actual engineering situation to ensure that the case can be reproduced.
[0120] 1. Test object parameters
[0121] (1) 220kV phase transformer: rated capacity 180MVA, rated voltage 220kV / 110kV, service life 6 years (belonging to the 5~10 year moderate aging stage, ΔT_age=2.5℃), with 3 sets of oil temperature sensors, current sensors, voltage sensors and winding temperature sensors (A / B / C phases), adapted to IEC 61850 protocol, and connected to the existing monitoring system of the substation.
[0122] (2) 500kV phase transformer: rated capacity 360MVA, rated voltage 500kV / 220kV, service life 3 years (belonging to the 1~5 year stable operation stage, ΔT_age=1℃), sensor configuration is the same as 220kV phase transformer, with complete remote signaling interface (fan motor, cooler power circuit breaker, fan motor fault remote signaling).
[0123] 2. Test conditions
[0124] (1) Data acquisition frequency: 8 seconds / time by default. The remote signal acquisition frequency is consistent with the oil temperature acquisition frequency.
[0125] (2) Operating condition matching parameters: The six-layer operating condition matching rules of this invention shall be followed. When the active power change rate is greater than 10% / 10 minutes, the matching time window shall be extended to 30 minutes.
[0126] (3) Threshold parameters: Calculated according to step five, where the proportional coefficient k is: k=0.02℃ / MVA for 220kV transformer and k=0.018℃ / MVA for 500kV transformer (calibrated by linear regression); the cooler response threshold contains active / reactive correction coefficients, which are executed according to the rules above.
[0127] (4) Test scenarios: covering five typical scenarios: normal operation, A-phase cooler failure, B-phase dedicated CT open circuit, local oil temperature abnormality under high reactive power ratio (60% reactive power ratio), and overall overheating abnormality. Each scenario will be tested for 15 days, and the warning response time and warning accuracy will be recorded.
[0128] 3. Implementation process and results
[0129] (1) Scenario 1: Normal operating condition (two transformers)
[0130] The system collects parameters such as total active power, total reactive power, phase oil temperature, and winding temperature in real time. After data verification, six-level operating condition matching, and threshold calculation, the early warning judgment module does not trigger any early warning. The historical database update module normally stores valid data into the historical database and continuously removes outdated data older than 5 years. The system operates stably and provides error-free early warnings.
[0131] (2) Scenario 2: Phase A cooler failure (220kV transformer)
[0132] A simulated A-phase cooler fan failure was initiated. After the cooler started, the temperature difference between phase A and phases B and C reached 2.3℃, and the phase current deviation was 3.2% (≤5%). After 10 minutes, the system triggered a single-phase cooler failure warning (blue audible and visual alarm), indicating "suspected A-phase cooler failure". The warning response time was 8 seconds, consistent with the actual fault. After handling, the warning was lifted, and the fault handling record was updated synchronously in the historical database.
[0133] (3) Scenario 3: Phase B dedicated CT open circuit (500kV transformer)
[0134] The dedicated CT circuit for measuring the temperature of phase B winding was manually disconnected. For 7 consecutive minutes, the condition "|T_wind-B-T_oil-B|=1.8℃<2℃" was met, and the active power of phase B was 35% (>20%) of the rated power. There was no abnormal oil temperature warning. The system triggered a dedicated CT open circuit warning (blue audible and visual alarm), indicating "Suspected open circuit of the CT associated with the phase B winding temperature gauge. Check the circuit continuity." The warning response time was 12 seconds. After investigation, the fault was confirmed, and it was completely consistent with the warning message.
[0135] (4) Scenario 4: Local oil temperature abnormality under high reactive power ratio (500kV transformer)
[0136] The system is set to have a total reactive power ratio of 60% and a total active power ratio of 70% of the rated power. Simulating localized aging of the B-phase winding, a temperature of ΔT_oil-B = 2.8℃ (exceeding the calculated T_Δoil = 2.2℃) is generated. Within 5 minutes, ΔT_oil-B continues to rise by 0.9℃. Changes in P_φ and Q_φ in B-phase are both <3% and 5% respectively. The overall oil temperature is ≤T_total. The system triggers a Level 1 warning (green audible and visual alarm) with a 3-minute response time, identifying localized anomalies 1.5 hours in advance. After handling the anomaly and resolving it, the warning is lifted.
[0137] (5) Scenario 5: Overall overheating anomaly (220kV transformer)
[0138] The simulation of a partial short circuit in the winding caused the overall oil temperature to rise to 86℃ (≥85℃). The system immediately triggered a level three warning (red audible and visual alarm), indicating "severe overheating, partial short circuit hazard, emergency shutdown for inspection". The warning response time was 5 seconds. After emergency shutdown, the short circuit hazard was confirmed and equipment damage was avoided, verifying the effectiveness of the fallback threshold.
[0139] 4. Verification Conclusion
[0140] The on-site verification, lasting no less than 75 days, demonstrates that the method and system of this invention can operate stably on 220kV and 500kV oil-immersed phase-splitting transformers. The early warning response time is ≤3 minutes under various scenarios. No false alarms or missed alarms occurred during the 75-day on-site verification period, meeting the actual engineering requirements. In particular, the dedicated CT open-circuit early warning fills a gap in existing technology, and there are no missed alarms or false alarms under high reactive power ratio conditions. The introduction of six-level operating condition matching and active / reactive power correction coefficients improves the accuracy of early warning. It can be directly applied to oil-immersed phase-splitting transformers of all voltage levels from 220kV to 750kV.
[0141] The embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention.
Claims
1. A multi-dimensional abnormality early warning method for oil-immersed split-phase power transformers, characterized in that, Includes the following steps: Step 1: Real-time acquisition of operating parameters and associated remote signaling signals of the oil-immersed phase-separated power transformer. Operating parameters include total active power, total reactive power, active power of each phase, reactive power of each phase, top oil temperature of each phase, winding temperature of each phase, current operating years of the transformer, and current month. Associated remote signaling signals include remote signaling for fan motor faults, cooler power supply circuit breaker status, and fan motor faults. The top oil temperature of each phase is acquired through phase-separated transformer oil temperature sensors, and the overall oil temperature is the arithmetic mean of the top oil temperatures of the three phases. The acquisition frequency is adjusted according to the on-site operation and maintenance needs, and the remote signal acquisition frequency is kept consistent with the oil temperature acquisition frequency. Step 2: Filter the collected operating parameters and remote signaling signals to remove invalid data caused by sensor malfunctions, communication disturbances, or data jumps. If the top oil temperature or winding temperature of any phase exceeds the set range, it is determined that the corresponding sensor is faulty and the data for that phase is removed. If two or more phases are faulty, the early warning judgment is suspended. If the load data is empty or exceeds the rated value range, the load data is determined to be invalid and the judgment is suspended. If the overall oil temperature difference between two adjacent acquisition times is too large and the changes in total active power and total reactive power are minimal, it is determined to be a momentary data jump and the data for that adjacent acquisition time is removed. If the remote signaling signal continues to jitter, it is determined to be invalid and is not included in the linkage judgment. Step 3: Based on the collected valid operating parameters, and combined with the differentiated impact of active power and reactive power on the temperature rise of the main transformer, match historical datasets from the historical database that meet the conditions of time matching, total load matching, phase load matching, and service life matching; if the active power change rate is >10% / 10 minutes, extend the operating condition matching time window to 30 minutes to eliminate the thermal inertia interference of oil temperature in 220kV~750kV phase transformers. Step 4: Perform anomaly cleaning on the matched historical dataset and store it in layers according to the number of years of operation; anomaly cleaning includes removing data of instantaneous disturbances or sensor failures, removing abnormal data of single mutual sensors, and removing false overheating data without load support. Step 5: Calculate the quantiles of the overall historical oil temperature dataset for each layer and superimpose the temperature rise margins for the corresponding operating years to obtain the overall oil temperature warning threshold. Calculate the quantiles of the historical phase oil temperature difference datasets for each layer and superimpose the temperature rise margin to obtain the phase oil temperature difference warning threshold; determine the single-phase cooler failure threshold, cooler normal response threshold, and winding to oil surface temperature ratio threshold; preset the overall oil temperature, phase oil temperature difference, and single-phase difference fallback thresholds. Step Six: Based on real-time operating parameters and various calculated thresholds, combined with trend verification and remote signaling linkage, the main transformer body temperature early warning adopts a dual threshold verification mechanism. Based on the dual verification of the dynamic threshold of historical data quantiles and the absolute value bottom-line threshold, it is determined whether an early warning is triggered and the warning level, including cooler power supply abnormality warning, single-phase cooler failure warning, cooler operating condition abnormality warning, early local abnormality warning, mid-term overall abnormality warning, emergency overheating abnormality warning and dedicated CT open circuit warning; Step 7: Record the early warning events, handling results, and remote signaling related data. Add the real-time operating parameters that have passed this verification to the historical database according to the hierarchical rules, and continuously remove outdated data. When the sample size of the historical dataset is insufficient, use the historical average oil temperature and phase oil temperature difference of the corresponding operating year period plus a fixed value as the early warning threshold.
2. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning method according to claim 1, characterized in that, include: Step 1: Data Acquisition: Real-time acquisition of operating parameters and associated remote signaling signals for 220kV~750kV oil-immersed phase-separated power transformers. Operating parameters include total active power P, total reactive power Q, A / B / C three-phase active power P_φ, A / B / C three-phase reactive power Q_φ, A / B / C three-phase top oil temperature T_oil-φ, A / B / C three-phase winding temperature T_wind-φ, transformer current operating years Y, and current month M. Associated remote signaling signals include fan motor fault, cooler power supply circuit breaker status, and fan motor fault remote signaling. The top oil temperature T_oil-φ of the three phases A / B / C is collected by the phase transformer oil temperature sensor. The overall oil temperature is the arithmetic mean of the three-phase top oil temperatures, T_avg=(T_oil-A+T_oil-B+T_oil-C) / 3; T_oil-A, T_oil-B, and T_oil-C are the top oil temperatures of the three phases A / B / C, respectively. The default acquisition frequency is 5~10 seconds / time, or can be adjusted to 5~30 seconds / time according to the on-site operation and maintenance requirements of 220kV~750kV. The remote signal acquisition frequency is consistent with the oil temperature acquisition frequency. Step 2: Data Validity Verification: The collected operating parameters and remote signaling signals are filtered for validity, and invalid data caused by sensor failures, communication disturbances, and data jumps are removed. The specific rules are as follows: If any phase T_oil-φ < -10℃ or > 120℃, or any phase T_wind-φ < -10℃ or > 150℃, it is determined that the corresponding sensor is faulty, and the corresponding data is removed. Only the valid data of the remaining two phases are used for subsequent calculations. If two or more phases are faulty, the current warning judgment is suspended and a sensor fault prompt is triggered. If the total P, total Q, or any phase P_φ or Q_φ is empty, or exceeds ±20% of the corresponding rated active power / rated capacity, the load data is deemed invalid, and the current warning judgment is suspended. If the overall oil temperature difference between two adjacent data collection times is >5℃, and the changes in total P and total Q are both <1%, the data is deemed to have a momentary jump, and all data at that time is discarded. If the remote signaling signal continues to jitter and switches ≥3 times within ±1 minute, the remote signaling is deemed invalid and is temporarily excluded from the linkage judgment. Step 3: Operating Condition Matching: Based on the collected valid operating parameters and considering the differentiated impact of active and reactive power on the main transformer temperature rise, historical datasets meeting the following conditions are matched from the historical database to adapt to the thermal hysteresis characteristics of 220kV~750kV phase transformers: Time matching: If the month of the real-time data is M, then only the data for months M-1, M, and M+1 in the historical database will be searched; where January matches December, January, and February, and December matches November, December, and January. Total load matching: Historical total active power is within ±2% of real-time total P, and historical total reactive power is within ±5% of real-time total Q. Phase-by-phase load matching: The historical active power P_φ of each phase is within ±3% of the real-time corresponding phase P_φ, and the historical reactive power Q_φ of each phase is within ±5% of the real-time corresponding phase Q_φ. Service life matching: The service life segment of the transformer corresponding to the historical data is consistent with the service life segment to which the current service life Y belongs; If the active power change rate is greater than 10% / 10 minutes, extend the operating condition matching time window to 30 minutes to eliminate the thermal inertia interference of oil temperature in 220kV~750kV phase transformers; Step 4: Data Preprocessing: The historical dataset matched in Step 3 is cleaned of outliers and stored hierarchically according to operating years. The operating years and corresponding temperature rise margin ΔT_age are preset as follows: 0~1 years, ΔT_age=0℃; 1~5 years, ΔT_age=1℃; 5~10 years, ΔT_age=2~3℃; >10 years, ΔT_age=3~5℃. This division of operating years is based on the aging characteristics of the entire life cycle of 220kV~750kV phase transformers. The rules for abnormal data cleaning are customized rules adapted to the fault characteristics of 220kV~750kV phase transformers: Data with ΔT_max > 8℃ and P_φ change rate of each phase < 5% of the rated value are excluded and determined to be transient disturbance or sensor failure data. Data that excludes any phase ΔT_oil-φ > 5℃ and the other two phases ΔT_oil-φ < 2℃ are determined to be abnormal data from the oil temperature sensor of that phase. Data with an overall oil temperature > 85℃ and total pressure (P) < 50% of the rated total pressure (P) are excluded as false overheating data without corresponding load support; where ΔT_max = max(T_oil-A,T_oil-B,T_oil-C)-min(T_oil-A,T_oil-B,T_oil-C), ΔT_oil-φ = |T_oil-φ - T_avg| (φ = A, B, C); Step 5: Threshold Calculation Overall oil temperature threshold: Calculate the 95th percentile T95_total of the historical overall oil temperature dataset for each layer, and add the temperature rise margin ΔT_age for the corresponding operating years to obtain the overall oil temperature warning threshold T_total = T95_total + ΔT_age; Phase-separated oil temperature difference thresholds: Calculate the 95th quantile T95_Δmax of the historical ΔT_max dataset for each layer and the 95th quantile T95_Δoil of the historical ΔT_oil-φ dataset, and superimpose the temperature rise margin ΔT_age for the corresponding operating years to obtain the ΔT_max warning threshold T_Δmax = T95_Δmax + ΔT_age and the ΔT_oil-φ warning threshold T_Δoil = T95_Δoil + ΔT_age; Single-phase cooler failure threshold: After the cooler starts, the "difference in temperature drop" between any phase oil temperature and other phase oil temperatures is >2℃, and the phase current deviation is ≤5%, with a duration ≥10 minutes; the phase current deviation is calculated as (maximum value - minimum value) / average value × 100%; Cooler normal response thresholds: After starting one set of fans, the oil temperature drop should be ≥4.0℃ in 10 minutes, ≥4.7℃ in 30 minutes, and the average drop rate in the first 10 minutes should be ≥0.40℃ / min; After starting two sets of fans, the starting conditions should be set according to the on-site operating procedures, with an oil temperature drop of ≥3.5℃ in 10 minutes and ≥5.5℃ in 30 minutes; Active load correction factor: When the active power increases by 5MW, the 10-minute drop threshold for one set of fans decreases by 0.8℃, and the 10-minute drop threshold for two sets of fans decreases by 1.2℃; Reactive load correction factor: When the reactive power increases by 10Mvar, the 10-minute drop threshold for one set of fans decreases by 0.5℃, and the 10-minute drop threshold for two sets of fans decreases by 0.8℃, adapting to the operating characteristics of the reactive power ratio difference of 220kV~750kV phase transformers; Winding-oil surface temperature ratio threshold: Based on historical datasets, calculate the range of difference between the winding temperature and oil surface temperature of the same phase under normal operating conditions, set the ratio coefficient k, and k ranges from 0.01℃ / MVA to 0.03℃ / MVA. The normal threshold condition is T_wind-φ - T_oil-φ ≥ k×S_φ (S_φ=√(P_φ²+Q_φ²)), and T_wind-φ > T_oil-φ; Preset absolute value fallback thresholds: The overall oil temperature absolute value warning threshold is 85℃, the ΔT_max fallback threshold is 5℃, and the ΔT_oil-φ fallback threshold for any phase is 3℃; Step Six: Multi-dimensional Early Warning Judgment: Based on real-time operating parameters and various thresholds calculated in Step Five, combined with trend verification and remote signaling linkage, determine whether an early warning is triggered and the warning level. Adapt to the fault characteristics of 220kV~750kV phase transformers. The main transformer body temperature early warning adopts a dual threshold verification mechanism. Cooler power supply abnormality warning: If the real-time response data is lower than the normal response threshold of the cooler after the fan is switched on or off, and the cooler power supply failure / fan motor failure remote signal is triggered within ±10 minutes, the cooler power supply abnormality warning will be triggered, and a blue audible and visual alarm will be set up with a text prompt indicating a suspected abnormal phase cooler power supply failure. Single-phase cooler failure warning: If the single-phase cooler failure threshold is met, a single-phase cooler failure warning will be triggered, with a blue audible and visual alarm and a text message indicating a suspected abnormal phase cooler failure. Cooler operating condition abnormality warning: If the normal response threshold of the cooler is not met and there is no associated remote signaling trigger, a normal cooler operating condition abnormality warning will be triggered, and a blue audible and visual alarm will be activated. Level 1 warning, early local anomaly: meets the following conditions: ΔT_max > T_Δmax or any phase ΔT_oil-φ > T_Δoil, and the corresponding phase oil temperature difference trend verification passes. Within 5-8 minutes, ΔT_max continuously rises by ≥1℃, or any phase ΔT_oil-φ continuously rises by ≥0.8℃, and the corresponding phase P_φ change is <3%, Q_φ change is <5%, while the overall oil temperature is ≤T_total. The corresponding anomaly types are single-phase cooling branch fault, phase load imbalance, and local winding aging, and are applicable to the local fault characteristics of 220kV~750kV phase transformers. Level 2 warning, mid-term overall anomaly: meets the following conditions: overall oil temperature > T_total, and ΔT_max > T_Δmax or phase oil temperature difference trend verification passes, while overall oil temperature < 85℃; the corresponding anomaly type is overall temperature rise superimposed on local anomaly, which is suitable for the overall operation anomaly scenario of 220kV~750kV phase transformers. Level 3 warning, emergency overheating anomaly: meets the following conditions: overall oil temperature ≥85℃, or overall oil temperature ≥70℃ and ΔT_max ≥5℃, or overall oil temperature ≥70℃ and any phase ΔT_oil-φ ≥3℃; the corresponding anomaly type is severe overheating and local short circuit hazard, and is suitable for emergency fault scenarios of 220kV~750kV phase transformers. Dedicated CT open circuit warning: If "|T_wind-φ -T_oil-φ| < 2℃" is met for 5~10 consecutive minutes, and the active power P_φ of the corresponding phase is greater than 20% of the rated power, but the above oil temperature abnormality warning conditions are not met, the dedicated CT open circuit warning for winding temperature measurement will be triggered. A blue audible and visual alarm will be activated, and a text prompt will be added indicating that the CT of the suspected abnormal phase winding temperature gauge is open. Check the circuit continuity. Step 7, Early Warning Handling and Historical Database Update: Record early warning events, handling results, and remote sensing related data. Add the real-time operating parameters that have passed this verification to the historical database according to the hierarchical rules in Step 4, and continuously remove old data older than 5 years. When the sample size of the historical dataset matched in the historical database is <20, it is judged as statistically invalid. At this time, the overall oil temperature early warning threshold is changed to the historical average oil temperature of the corresponding operating years + 8℃, and the phase oil temperature difference threshold is changed to the historical average ΔT_max + 0.5℃ and the historical average ΔT_oil-φ + 0.5℃ of the corresponding operating years.
3. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning method according to claim 1, characterized in that, In step one, the remote signal acquisition is adapted to the IEC 61850 protocol for power field applications, supporting seamless integration with the 220kV~750kV substation monitoring system; the phase active power P_φ and phase reactive power Q_φ are acquired through the existing current and voltage sensors of the 220kV~750kV phase transformer, and the winding temperature T_wind-φ is acquired through the existing winding temperature sensor; The sampling frequency is adjusted according to the on-site operation and maintenance needs, with an adjustment range of 5~30 seconds / time.
4. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning method according to claim 3, characterized in that, The division of the service life in step four is based on the aging characteristics of the entire life cycle of 220kV~750kV phase transformers: 0~1 years is the new commissioning stage, with no aging of windings and oil, and the best heat dissipation efficiency; 1~5 years is the stable operation stage, with slight aging of windings and oil, and copper loss increase ≤5%; 5~10 years is the moderate aging stage, with increased winding resistance, deterioration of oil, and a decrease in heat dissipation efficiency of about 10%; >10 years is the deep aging stage, with a decrease in cooling system efficiency of more than 15%, and a significant increase in copper loss and iron loss.
5. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning method according to claim 4, characterized in that, The core basis for determining the temperature rise margin ΔT_age is: the winding resistance change rate, oil dielectric loss change rate, and cooling system efficiency decay rate of 220kV~750kV phase transformers in different operating years.
6. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning method according to claim 5, characterized in that, Step 6 outlines the corresponding anomaly types and handling solutions for each warning level: Level 1 warning corresponds to single-phase cooling branch faults, phase load imbalance, and partial winding aging; the handling solution is to investigate the operating status of the phase cooling system and phase load distribution of the 220kV~750kV phase transformers. Level 2 warning corresponds to overall temperature rise combined with local anomalies; the handling solution is to reduce the transformer load and focus on investigating the abnormal phases. Level 3 warning corresponds to severe overheating and potential local short circuits; the handling solution is to shut down the transformer for inspection. Blue warnings related to the cooler and open-circuit warnings for dedicated CTs both indicate that the hardware status of the corresponding circuits should be investigated first.
7. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning method according to claim 6, characterized in that, Step five uses the Excel function PERCENTILE.EXC or the Python function numpy.percentile. The 95th percentile calculation is based on the historical data of the phase transformer and removes extreme values caused by phase load imbalance, adapting to the phase operation characteristics of 220kV~750kV phase transformers.
8. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning method according to claim 7, characterized in that, The calibration process for the proportional coefficient k in step five is as follows: Based on the historical operating data of 10 phase transformers of different voltage levels from 220kV to 750kV, covering different load ranges and 20% to 100% of rated power, the correlation model between T_wind-φ, T_oil-φ, and S_φ is obtained through linear regression analysis. Finally, the value range of k is determined to be 0.01~0.03℃ / MVA to ensure the accuracy of the proportional relationship determination under normal operating conditions.
9. An oil-immersed split-phase power transformer multi-dimensional anomaly early warning system, characterized in that, The system is used to implement the multi-dimensional anomaly early warning method for oil-immersed phase-splitting power transformers according to any one of claims 1-8. The system includes the following modules bound to the physical hardware of the 220kV~750kV phase-splitting transformer. The connection relationship between each module is designed based on the signal transmission path of the 220kV~750kV phase-splitting transformer. The data acquisition module is directly connected to the historical database update module: Data acquisition module: Employs monitoring sensors suitable for 220kV~750kV phase-separated power transformers, including oil temperature sensors, current sensors, voltage sensors, winding temperature sensors, and remote signal acquisition interfaces, compatible with the IEC 61850 protocol. It is used to acquire in real-time the total P, total Q, phase P_φ, phase Q_φ, phase T_oil-φ, phase T_wind-φ, operating years Y, month M, and associated remote signaling signals of the phase-separated transformer, and transmits them to the data verification module. The sampling frequency is adjustable, with a default of 5~10 seconds / time. It supports offline data caching, and can cache ≥72 hours of collected data when communication is interrupted. Data verification module: Electrically connected to the data acquisition module, it is used to verify the validity of the acquired operating parameters and remote signaling signals, and to remove invalid data caused by sensor failure, communication disturbance, and data jump. If sensor failure, invalid data or remote signaling jitter is detected, the corresponding prompt information is output and the valid data is transmitted to the operating condition matching module. Operating condition matching module: Electrically connected to the data verification module and the historical database respectively. The core is to adapt to the differential temperature rise effect of active power and reactive power. Based on valid data, it matches historical datasets from the historical database that meet the conditions of "same month ± 1 month + total active power + total reactive power + phase active power + phase reactive power + years of operation". It supports thermal hysteresis adaptation and automatically extends the matching time window when the active power fluctuation is large. The matched historical datasets are then transmitted to the data preprocessing module. Data preprocessing module: Electrically connected to the operating condition matching module, used to clean abnormal data in the matched historical dataset, customize rules according to step four, and store the data in layers according to the operating years. It calculates the 95th percentile of the historical data in each layer and the temperature rise margin ΔT_age of the corresponding operating years, and transmits it to the threshold calculation module. Threshold calculation module: Electrically connected to the data preprocessing module, used to calculate the overall oil temperature warning threshold, phase oil temperature difference warning threshold, single-phase cooler failure threshold, cooler normal response threshold, winding-oil surface temperature ratio threshold and absolute value fallback threshold, and transmit them to the warning judgment module; Early warning judgment module: It is electrically connected to the data verification module and the threshold calculation module respectively. It is used to combine real-time effective data and various thresholds to perform trend verification, remote signaling linkage and early warning level judgment, output the main transformer body fault early warning and the measurement circuit fault early warning, and transmit the early warning signal to the power monitoring terminal. The warning signals are distinguished by different audible and visual cues: blue for cooler-related warnings and dedicated CT open circuit warnings, green for Level 1 warnings, yellow for Level 2 warnings, and red for Level 3 warnings; it has an IEC 61850 communication interface and can upload warning signals to the 220kV~750kV power dispatching system; Handling prompt module: Electrically connected to the early warning judgment module, it outputs corresponding handling prompt information for different early warning types. For example, if the cooler power supply is abnormal, it prompts to check the corresponding phase power supply circuit breaker / fan motor. If the single-phase cooler fails, it prompts to check the corresponding phase fan motor / heat dissipation channel. If the dedicated CT is open, it prompts to check the corresponding phase winding temperature, the dedicated CT circuit continuity and wiring status. For local / overall abnormalities, it prompts load adjustment and fault diagnosis according to the early warning level. It is suitable for the operation and maintenance scenarios of 220kV~750kV phase transformers. Historical database update module: It is electrically connected to the data verification module and the historical database respectively. It is used to add the real-time valid data that has passed the verification to the historical database in layers according to the running years, and automatically remove old data that is more than 5 years old. When the sample size of the historical dataset is <20, it outputs a threshold adaptation prompt and enables the backup threshold scheme. Historical database: Used for hierarchical storage of historical operating parameters, remote signaling data and early warning event records after cleaning. It supports fast data query, hierarchical storage and rolling updates. It can be deployed on edge computing devices of 220kV~750kV substations or power monitoring center servers, and is suitable for different deployment scenarios on site. The historical operating parameters include active / reactive power differential matching identifiers and winding-oil surface temperature ratio data.
10. The oil-immersed split-phase power transformer multi-dimensional anomaly early warning system according to claim 9, characterized in that, The system is compatible with 220kV, 330kV, 500kV, and 750kV oil-immersed phase-splitting transformers. It automatically adjusts the matching deviation range of total / phase active and reactive power, as well as the correction coefficient of the cooler response threshold, to address the differences in copper and iron impedance among transformers of different voltage levels. Simultaneously, it automatically calibrates the value of the proportional coefficient k based on the winding structure and heat dissipation characteristics of different transformers, ensuring the accuracy of winding-oil surface temperature ratio verification. It maintains precise early warning under high reactive power ratios and different load conditions, adapting to the operational needs of phase-splitting transformers across all voltage levels from 220kV to 750kV.
