An artificial intelligence big data monitoring method and system for a power system

Abstract

This invention relates to the field of power system monitoring technology and discloses an artificial intelligence big data monitoring method and system for power systems, comprising: executing an upgrade period decision strategy, calculating the load activity index of the transformer based on the operating characteristics, constructing a transformer upgrade period evaluation function, and generating a target upgrade period; sorting multiple transformers to be upgraded based on the target upgrade period, and upgrading the transformers in staggered peak times according to the sorting results within each target upgrade period; executing an upgrade risk control strategy, calculating the transformer's regulation characteristics in real time, and performing graded control according to the upgrade risk value; executing a standby transformer scheduling strategy, calculating the standby compatibility score between the upgraded and verified transformer and the current transformer to be upgraded; and executing a switching collaborative maintenance strategy, adjusting the tap position through on-load tap changers and synchronously controlling the migration of metal particles in the insulation medium, achieving intelligent upgrades, uninterrupted power supply, and enhanced equipment safety.

Description

Technical Field

[0001] This invention relates to the field of power system monitoring technology, specifically to an artificial intelligence big data monitoring method and system for power systems. Background Technology

[0002] With the continuous improvement of the intelligence level of power systems, the demand for transformer software upgrades and operation and maintenance is becoming increasingly frequent. How to complete equipment updates and condition-based maintenance while ensuring power supply continuity has become an important issue in power system operation and management.

[0003] To avoid power outages during transformer upgrades, existing technologies typically assess load conditions before the upgrade and select a relatively low-risk period for execution. However, in scenarios such as data centers, industrial control systems, and research platforms, power outage-sensitive loads are present in the power supply system for extended periods, making it difficult to obtain conditions for complete load removal during upgrades. Furthermore, electrical loads are characterized by continuous operation and dynamic changes. While some research has explored load forecasting using historical load curves, existing forecasting methods are primarily geared towards long-term planning or short-term scheduling, lacking dedicated forecasting mechanisms for selecting upgrade windows. This makes it difficult to integrate terminal importance with real-time load fluctuations for comprehensive judgment, hindering maintenance personnel from accurately identifying long-term stable, low-risk upgrade windows and increasing the difficulty of upgrade implementation. In addition, existing technologies often rely on static load conditions before the upgrade for risk assessment, lacking the ability to continuously perceive dynamic changes in electrical terminals during the upgrade process. For example, sudden load increases, the startup of critical services, or the addition of new equipment can invalidate the original assessment results, introducing new power supply risks. Meanwhile, during long-term operation, the mechanical action of the on-load tap changer will generate metal particles that enter the insulating oil. When the transformer performs on-load tap change or tap switching, the electric field disturbance caused by the voltage change will exacerbate the migration and accumulation of particles, which may induce partial discharge or even insulation breakdown. Existing technologies usually require additional downtime for cleaning and maintenance, which not only increases the operation and maintenance costs, but also further reduces the equipment availability time.

[0004] This solution proposes an artificial intelligence big data monitoring method and system for power systems, addressing the problems raised in the background technology. Summary of the Invention

[0005] This invention provides an artificial intelligence big data monitoring method and system for power systems, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: an artificial intelligence big data monitoring method for power systems, comprising:

[0007] The transformer is upgraded based on the operating characteristics of the power consumption terminal, specifically as follows:

[0008] The transformer supplies power to multiple electrical terminals through power distribution lines, and the operating characteristics of each electrical terminal are extracted;

[0009] For a single transformer:

[0010] The decision-making strategy for the upgrade period is implemented, the load activity index of the transformer is calculated based on the operating characteristics, the time period evaluation function for transformer upgrade is constructed, and the target upgrade period is generated;

[0011] Multiple transformers to be upgraded are sorted according to the target upgrade period, and the transformers are upgraded in staggered time according to the sorting results within each target upgrade period.

[0012] When upgrading a transformer, an upgrade risk control strategy is implemented, the transformer's control characteristics are calculated in real time, the upgrade risk value is calculated based on the control characteristics, and graded control is implemented according to the upgrade risk value.

[0013] The hierarchical control includes a backup transformer scheduling strategy;

[0014] Implement the standby transformer scheduling strategy, and calculate the standby compatibility score between the upgraded and verified transformers obtained from the peak-shifting upgrade and the current transformer to be upgraded.

[0015] The upgraded and verified transformer with the highest standby adaptation score will be used as the standby transformer.

[0016] Control the backup transformer to take over the power consumption terminal of the transformer currently to be upgraded;

[0017] During the standby transformer takeover process:

[0018] The switching and collaborative maintenance strategy is implemented by adjusting the tap position through the on-load tap changer and simultaneously controlling the migration of metal particles in the insulating medium.

[0019] Optionally, the decision-making strategy for the upgrade period, based on the operating characteristics, calculates the transformer's load activity index, including:

[0020] The operating characteristics include the average power, peak power, and low-load operating periods of the power terminals;

[0021] A set of power-consuming terminals powered by a single transformer;

[0022] Calculate the electricity sensitivity index of each element in the electricity terminal set in sequence. ;

[0023] ,in, These represent the average power, peak power, and low-load operating period of the power consumption terminal, respectively. It is a very small positive number. Let be the aggregate average power of the set of electricity terminals, and These are the normalized values;

[0024] The power consumption sensitivity index is negatively correlated with low load operating time and aggregate average power, and positively correlated with average power and peak power.

[0025] The weight of any element in the set of electricity terminals is calculated as follows: , The total number of centralized elements in the power consumption terminal;

[0026] Based on the aforementioned weights, a weighted average of the centralized electricity consumption sensitivity indicators of the electricity terminals is calculated, and the result is recorded as the transformer load activity index. .

[0027] Optionally, the step of constructing a time-period evaluation function for transformer upgrades to generate target upgrade time periods includes:

[0028] Divide 24 hours into multiple time periods and iterate through any future time period in sequence;

[0029] Predicted power consumption of electrical terminals in future periods based on operational characteristics. ;

[0030] ,in, To predict the adjustment coefficient, , which is the power fluctuation term constructed based on peak power and ensemble average power;

[0031] The mean value of the centralized predicted power of the power terminals is obtained. ;

[0032] Constructing a time-period evaluation function for transformer upgrades ;

[0033] ,in, As a weighting coefficient, the lower the predicted power, the higher the load activity index, and the higher the time period evaluation function;

[0034] Set evaluation threshold ;

[0035] like The future time period currently being traversed is recorded as the target upgrade time period.

[0036] Optionally, the step of sorting multiple transformers to be upgraded based on target upgrade periods, and upgrading transformers in staggered shifts according to the sorting results within each target upgrade period, includes:

[0037] Set an upgrade evaluation period, and sort multiple transformers to be upgraded from high to low based on the transformer's time period evaluation function to determine the target transformer for upgrade;

[0038] The target upgrade period is included within the upgrade evaluation period;

[0039] Set the upgrade quantity c;

[0040] During the upgrade assessment period, the top-ranked transformer (ranked C) is selected as the target transformer for upgrade.

[0041] The remaining transformers will be re-submitted for re-evaluation during a subsequent upgrade assessment period.

[0042] Optionally, the implementation of the upgraded risk control strategy, which involves calculating the transformer's control characteristics in real time, calculating the upgraded risk value based on the control characteristics, and implementing tiered control according to the upgraded risk value, includes:

[0043] Obtain the control characteristics of the transformer;

[0044] The control features include load activity index and load change rate. and upgrade progress factor e;

[0045] Normalized load activity metrics, load change rate, and upgrade progress factor;

[0046] Calculate the load change rate ,in, Let be the power of the transformer at time t. For a moment The power of the transformer, For time intervals;

[0047] The upgrade progress factor is used to characterize the completion level of the current upgrade task, and is the ratio of the current completed upgrade workload to the total upgrade workload.

[0048] Calculate upgrade risk value , These are the weighting coefficients;

[0049] The upgrade risk value is positively correlated with the load activity index in a non-linear manner, and positively correlated with the load change rate and upgrade progress factor, so as to amplify the risk difference in high load scenarios and suppress risk fluctuations in low load scenarios.

[0050] Set a Level 1 risk threshold and secondary risk threshold ,and ;

[0051] like If so, the upgrade operation will continue;

[0052] like If so, the standby transformer scheduling strategy will be executed;

[0053] like If so, then an upgrade rollback operation will be performed.

[0054] Optionally, the implementation of the standby transformer scheduling strategy, based on the upgraded and verified transformers obtained from the peak-shifting upgrade, calculates the standby compatibility score between the upgraded and verified transformers and the current transformer to be upgraded, including:

[0055] Obtain any upgraded transformer and its remaining power supply capacity. Gear adjustment capability and operational reliability The remaining power supply capacity is then normalized.

[0056] Set performance evaluation values;

[0057] The operational evaluation value is the average of the remaining power supply capacity, the tap adjustment capability, and the operational reliability.

[0058] Based on the comparison between the operation evaluation value and the set evaluation threshold, when the operation evaluation value is greater than or equal to the evaluation threshold, the current transformer is recorded as an upgraded and verified transformer.

[0059] Calculate the standby adaptation score for the upgraded and verified transformer. :

[0060] ,in, These are the weighting coefficients;

[0061] The upgraded and verified transformer with the highest standby adaptation score will be used as the standby transformer.

[0062] The backup transformer is controlled to take over the power supply load of the transformer to be upgraded. Specifically, the backup transformer is switched and adjusted by means of an on-load tap changer according to the change in power supply load.

[0063] Optionally, the implementation of the switching collaborative maintenance strategy, which involves adjusting the tap position via an on-load tap changer and simultaneously controlling the migration of metal particles in the insulating medium, includes:

[0064] Obtain the output voltage of the on-load tap changer of the standby transformer before and after the tap position switching, respectively. and ;

[0065] Calculate the voltage change caused by gear shifting. ,in, It is used to characterize the degree of change in electric field distribution during gear switching. The greater the voltage change, the more obvious the electric field disturbance inside the insulating medium.

[0066] Set voltage change threshold ,like If so, dielectric electrophoresis control is initiated to migrate the metal particles in the insulating medium;

[0067] Obtain the number of metal particles before and after the treatment, respectively. and ;

[0068] Calculate the removal efficiency of metal particles When the removal efficiency is lower than the set removal efficiency threshold, the dielectrophoresis control intensity is increased.

[0069] An artificial intelligence big data monitoring system for power systems includes:

[0070] The feature extraction module is used to collect the operating data of each power terminal of the power distribution line, extract the average power, peak power and low load operating period, and normalize each parameter.

[0071] The upgrade period decision module is used to receive operational characteristic data, calculate the transformer's load activity index, construct a period evaluation function and compare it with a threshold to generate the target upgrade period;

[0072] The staggered scheduling module is used to sort multiple transformers to be upgraded from high to low based on the time period evaluation function, and select the transformers at the top of the sort to be upgraded within their respective target upgrade time periods.

[0073] Upgrade the risk control module to acquire control characteristics in real time during transformer upgrades, calculate upgrade risk values, and execute tiered control based on the upgrade risk values.

[0074] The switching collaborative maintenance module acquires the output voltage before and after the on-load tap changer switching and calculates the voltage change during the process of the standby transformer taking over the power terminal. It then drives the metal particles in the insulating medium to migrate and dynamically adjusts the dielectric electrophoresis control intensity based on the particle removal efficiency.

[0075] The present invention has the following beneficial effects:

[0076] 1. This artificial intelligence big data monitoring method for the power system calculates the power consumption sensitivity indicators of each terminal and performs a weighted average using weighting coefficients to obtain the transformer load activity index. The power consumption sensitivity index is negatively correlated with low-load operating periods and aggregated average power, and positively correlated with average power and peak power. This gives terminals with high peak power and short low-load windows higher sensitivity and assigns them higher weights, comprehensively reflecting the overall activity level of all terminals under the transformer. The predicted power for future periods is predicted through power fluctuation terms, and the average of the concentrated predicted power of power consumption terminals is calculated. A time period evaluation function is constructed for comprehensive evaluation, and the target upgrade time period is generated through threshold comparison. It aims to achieve both sufficiently low predicted power to ensure upgrade safety and sufficiently high load activity index to ensure upgrade necessity. This eliminates reliance on manual experience in the selection of upgrade time periods. Through mathematical models, multi-objective comprehensive optimization is achieved, effectively solving the technical challenges of long-term existence of power-sensitive loads in scenarios such as data centers and industrial control, and the difficulty in obtaining completely idle windows. The efficiency of time period decision-making is significantly improved.

[0077] 2. The artificial intelligence big data monitoring method for this power system sets an upgrade assessment period, sorts each transformer from highest to lowest based on its time-period assessment function value, and then sets the upgrade quantity, selecting only the top-ranked transformers to be upgraded within their respective target upgrade periods. Through the dual constraints of the sorting mechanism and quantity limit, the time-series distribution of multi-transformer upgrade tasks is achieved, ensuring that transformers with high urgency are prioritized for upgrades, effectively reducing the risk of grid structural vulnerability caused by simultaneous upgrades of multiple transformers. During the upgrade process, transformer control characteristics are acquired in real time, upgrade risk values ​​are calculated, and risk thresholds are set, realizing a three-level gradient control of continued upgrade—standby scheduling—upgrade rollback. Specifically, the nonlinear term significantly amplifies risk differences under high load scenarios and effectively suppresses risk fluctuations under low load scenarios, improving the discriminative power of risk scoring; the square root processing of the upgrade progress factor makes risk growth slow down as the upgrade approaches completion, reflecting the engineering intuition that the more work completed, the more inclined to continue. This upgrades risk control from static assessment to dynamic perception, from qualitative judgment to quantitative calculation, and from single decision-making to gradient response, improving the safety of the upgrade operation.

[0078] 3. The artificial intelligence big data monitoring method of this power system obtains the remaining power supply capacity, tap-level adjustment capability, and operational reliability of upgraded transformers. A first-level screening is performed using operational evaluation values; transformers with evaluation values ​​greater than or equal to the evaluation threshold are marked as verified upgraded transformers, while candidate transformers with insufficient reliability or capacity are eliminated. A second-level quantitative comparison is performed using standby adaptation scores, selecting the verified upgraded transformer with the highest adaptation score as the standby transformer to take over the power supply load of the transformer currently awaiting upgrade. This scheme constructs a dual-guarantee mechanism of "screening first, then selecting the best": the first-level screening ensures that all transformers entering the second round meet basic operating conditions; the second-level screening achieves a comprehensive evaluation of performance × reliability by weighted combination of remaining power supply capacity and tap-level adjustment capability multiplied by operational reliability, upgrading the selection of standby transformers from manual experience-based judgment to data-driven quantitative decision-making. The standby transformer is controlled to switch taps via on-load tap changers according to changes in power supply load, achieving smooth load transfer and effectively ensuring the power supply continuity of highly sensitive loads during the upgrade process, avoiding power outages caused by upgrade operations.

[0079] 4. The artificial intelligence big data monitoring method for this power system acquires the output voltage of the on-load tap changer of the standby transformer before and after the tap change during the switching process of the standby transformer to the power terminal, calculates the voltage change, and initiates dielectric electrophoresis control when the voltage exceeds a set threshold to migrate metal particles in the insulation medium. It identifies and utilizes the physical causal chain of tap change → voltage surge → electric field disturbance → particle migration, coupling the originally independent electrical voltage regulation operation with the insulation maintenance task through this bridge. Unlike traditional solutions that require scheduled shutdown windows for oil filtration maintenance, this invention utilizes the tap change action that inevitably occurs during the standby transformer takeover process, using the voltage change as a proxy indicator of electric field disturbance. Dielectric electrophoresis control is triggered only when the electric field disturbance is sufficiently significant, achieving precise condition-triggered maintenance. This avoids continuous operating losses of the dielectric electrophoresis device and ensures insulation protection during the most risky switching moments. It cross-domain integrates the originally independent technical branches of voltage regulation in electrical engineering and particle control in high-voltage insulation, achieving synergistic optimization of "electrical-insulation".

[0080] 5. The artificial intelligence big data monitoring method for this power system acquires the number of metal particles in the insulating medium before and after treatment, calculates the removal efficiency of metal particles, and automatically increases the dielectric electrophoresis control intensity when the removal efficiency is lower than the set removal efficiency threshold. This scheme constructs a complete monitoring-evaluation-adjustment closed-loop control link: real-time calculation of removal efficiency enables the quantification and evaluation of the current treatment effect; setting the removal efficiency threshold clarifies the control target; and automatically increasing the control intensity (such as increasing the amplitude of the dielectric electrophoresis electrode excitation voltage, reducing the excitation voltage frequency, or extending the action time) when the removal efficiency is not up to standard achieves dynamic adjustment of the control strategy. This ensures the controllability and consistency of the removal effect, avoiding the problem of inconsistent removal effects under different operating conditions caused by fixed parameter control. At the same time, by synchronously triggering dielectric electrophoresis control through tap switching, the insulation maintenance work that originally required shutdown is integrated into the switching process, eliminating the need for additional shutdown windows, effectively reducing operation and maintenance costs, reducing the average annual power outage time of transformers, and increasing equipment availability. Attached Figure Description

[0081] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0082] Figure 2 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0083] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0084] Example 1, refer to Figure 1 An artificial intelligence big data monitoring method for power systems, specifically:

[0085] The big data monitoring described in this invention refers to: based on the massive operational data of power terminals, through feature extraction, time period assessment, risk calculation and collaborative control, the data analysis results are directly used to guide transformer upgrade decisions, peak-shifting scheduling, risk classification and switching maintenance, so as to realize data-driven intelligent monitoring of the entire transformer upgrade process.

[0086] This example is an industrial park substation with a 10kV / 0.4kV oil-immersed transformer, numbered T-2026, which supplies power to the following four electrical terminals through four distribution lines:

[0087] Power terminal L1 is a data center cabinet with an average power of 280kW and a peak power of 350kW. Its low-load operating period is from 3:00 AM to 5:00 AM.

[0088] Power terminal L2 is an industrial robot production line with an average power of 420kW and a peak power of 500kW. Its low-load operation period is from 12:00 to 14:00 noon.

[0089] Power terminal L3 is a scientific research experimental platform with an average power of 150kW and a peak power of 220kW. Its low-load operation period is from 2:00 AM to 6:00 AM.

[0090] Power terminal L4 is the office building's air conditioning system, with an average power of 200kW and a peak power of 310kW. Its low-load operating period is from 22:00 at night to 6:00 the next day.

[0091] Normalization was performed using the transformer's rated capacity of 1250 kVA as a baseline, and further normalization was performed for low-load periods using a 24-hour baseline, resulting in the normalized parameters for each terminal: L1. The values ​​for L1, L2, and L3 are 0.224, 0.280, and 0.083 respectively; for L3, they are 0.336, 0.400, and 0.083 respectively; for L4, they are 0.120, 0.176, and 0.167 respectively; and for L5, they are 0.160, 0.248, and 0.333 respectively. The ensemble average power is calculated from these values. =(0.224+0.336+0.120+0.160) / 4=0.210.

[0092] Calculate the electricity sensitivity index of each element in the electricity terminal set in sequence. In this embodiment, It is 0.01;

[0093] The power sensitivity index of each terminal was calculated: L1 It is 0.156, L2 It is 0.335, L3 It is 0.026, L4 It is 0.025.

[0094] Among them, L2 has the highest power sensitivity index and L4 has the lowest. This indicates that the industrial robot production line corresponding to L2 has the highest urgency for upgrading due to its high peak power and short low load window, while the office building air conditioning system corresponding to L4 has the lowest urgency for upgrading due to its long low load period.

[0095] In real-world engineering scenarios, the continued operation of power-sensitive loads increases the risk of transformer upgrades, limits the upgrade window, and leads to delays or backlogs in upgrade tasks.

[0096] The weight of any element in the set of electricity terminals is calculated as follows: , The total number of centralized elements in the power consumption terminal;

[0097] The weighting coefficients for each terminal are calculated: L1 =0.224 / (4×0.210)=0.267; L2 =0.336 / (4×0.210)=0.400; L3 =0.120 / (4×0.210)=0.143; L4's =0.160 / (4×0.210)=0.190.

[0098] The load activity index is obtained by calculating the weighted average of the weighting coefficients and the electricity sensitivity index. =0.1842 / 1.000=0.184;

[0099] Construct a time-period evaluation function for transformer upgrades to generate target upgrade periods, specifically:

[0100] Predicted power consumption of electrical terminals in future periods based on operational characteristics. ;

[0101] ,in, To predict the adjustment coefficient, , which is the power fluctuation term constructed based on peak power and ensemble average power;

[0102] Calculate the terminals The values ​​are as follows: L1 is (0.280-0.210) / 0.210=0.333, L2 is (0.400-0.210) / 0.210=0.905, L3 is (0.176-0.210) / 0.210=-0.162, and L4 is (0.248-0.210) / 0.210=0.181;

[0103] Take the forecast adjustment coefficient =0.75, according to Calculate the predicted power of each terminal in the future time period. Taking 3:00 AM as an example, the predicted power of L1 to L4 are 0.262, 0.353, 0.184, and 0.239 respectively. Taking the average value, we get the representative value of the overall predicted power of the transformer in this time period: p5 = (0.262 + 0.353 + 0.184 + 0.239) / 4 = 0.260.

[0104] Similarly, the p5 at 12:00 noon is calculated to be 0.221, and the p5 at 14:00 in the afternoon is 0.265.

[0105] Constructing a time-period evaluation function for transformer upgrades ;

[0106] ,in, These are the weighting coefficients.

[0107] Substituting p5 for each time period into the time period evaluation function, we get:

[0108] 3:00 AM =0.6×(1-0.260 / 0.400)+0.4×0.290=0.6×0.350+0.116=0.326;

[0109] 12:00 noon =0.6×(1-0.221 / 0.400)+0.4×0.290=0.6×0.448+0.116=0.385; 2:00 PM =0.6×(1-0.265 / 0.400)+0.4×0.290=0.6×0.338+0.116=0.319;

[0110] Set evaluation threshold =0.35, only at 12:00 noon The value of 0.385 exceeds the threshold, therefore this period is recorded as the target upgrade period.

[0111] Multiple transformers to be upgraded are sorted according to the target upgrade period, and the transformers are upgraded in staggered time according to the sorting results within each target upgrade period.

[0112] The upgrade assessment period is set from 9:00 to 17:00 on the same day, and the number of upgrades is c=2.

[0113] There are a total of 5 transformers in the system that need to be upgraded, based on the time-period evaluation function. The transformers are sorted from highest to lowest quality, and the sorting result is: T-2027. Its value is 0.421, ranking 1st; T-2026's Its value is 0.385, ranking 2nd; T-2028's Its value is 0.362, ranking 3rd; T-2029's The value is 0.298, ranking 4th; T-2030's Its value is 0.275, ranking 5th.

[0114] Due to the top two transformers All values ​​exceeded the evaluation threshold of 0.35, and the target upgrade periods were all included within the upgrade evaluation period. Therefore, the top two ranked transformers, T-2027 and T-2026, were selected as target transformers for upgrades, with T-2026 being upgraded between 12:00 PM and 1:00 PM. The remaining transformers, T-2028, T-2029, and T-2030, were postponed to subsequent upgrade evaluation periods for re-ranking.

[0115] Regarding the T-2026 transformer:

[0116] During the upgrade of transformer T-2026 at 12:00 noon, the load change rate was calculated in real time. At time t=12:00, the transformer power p(t) was 420kW, with a standard value of 0.336; at time t-Δt=11:55, the transformer power p(t-Δt) was 380kW, with a standard value of 0.304; the time interval Δt was 5 minutes, or 0.083 hours.

[0117] According to the formula The load change rate was calculated to be Δp = (0.336 - 0.304) / 0.083 = 0.384, indicating that the load is increasing at a relatively rapid rate.

[0118] The current upgrade task involves a total of 10 steps, of which 6 steps have been completed, resulting in an upgrade progress factor e = 6 / 10 = 0.6. After obtaining the control characteristics, further load activity indicators will be obtained. =0.290, load change rate Δp=0.384, upgrade progress factor e=0.6, and these values ​​were normalized. Based on the upgrade risk value... , where κ1, κ2, and κ3 are 0.5, 0.3, and 0.2, respectively;

[0119] d=0.5×0.290²+0.3×0.384+0.2×√0.6=0.042+0.115+0.2×0.775=0.042+0.115+0.155=0.312.

[0120] The first-level risk threshold is set at 0.20, and the second-level risk threshold is set at 0.40. At this time, 0.20≤0.312<0.40, indicating that the current risk level is in the medium-risk range. Therefore, the standby transformer dispatch strategy is executed, and the standby transformer is called to take over the power terminal of the transformer to be upgraded.

[0121] In the standby transformer scheduling strategy, the transformers that have been upgraded in the system are first selected as candidate standby transformers, denoted as T-1024, T-1025 and T-1026 respectively.

[0122] The remaining power supply capacity, tap adjustment capability, and operational reliability of each transformer were obtained. The values ​​for T-1024 were 0.85, 0.90, and 0.98, respectively; for T-1025, they were 0.60, 0.75, and 0.92, respectively; and for T-1026, they were 0.95, 0.80, and 0.96, respectively.

[0123] Set the performance evaluation value R = (h1 + h2 + h3) / 3;

[0124] The calculated R-values ​​are (0.85+0.90+0.98) / 3=0.910 for T-1024, (0.60+0.75+0.92) / 3=0.757 for T-1025, and (0.95+0.80+0.96) / 3=0.903 for T-1026.

[0125] The evaluation threshold was set at 0.85. After comparison, the operation evaluation values ​​of T-1024 and T-1026 were both greater than or equal to the evaluation threshold, and were recorded as upgraded and verified transformers; the operation evaluation value of T-1025 was less than the evaluation threshold, and did not meet the standby conditions.

[0126] Subsequently, the standby adaptation score of the upgraded and verified transformer was calculated. ,in , .

[0127] T-1024:

[0128] T-1026: .

[0129] Comparing the two, T-1026 has a standby adaptation score of 0.854, which is higher than T-1024's 0.853. Therefore, T-1026 is selected as the standby transformer.

[0130] The backup transformer T-1026 is controlled to take over the power terminal of the current transformer T-2026 to be upgraded. Specifically, according to the changes in the power supply load of T-2026, the tap position of the on-load tap changer of T-1026 is switched and adjusted so that the output voltage of T-1026 matches the load demand, thus completing the load transfer.

[0131] During the process of connecting the standby transformer T-1026 to the power terminal of T-2026, the output voltage of the on-load tap changer of T-1026 before and after the tap changer switch is executed is obtained, including the voltage before the switch. The voltage after switching is 398V. The voltage is 402V. The voltage change is calculated as |398-402|=4V. The voltage change threshold is set to 3V. Since 4V≥=3V, it indicates that the electric field disturbance caused by the gear switching is significant. Dielectric-phoretic control is activated to migrate the metal particles in the insulating medium.

[0132] Obtain the number of metal particles in the insulating oil before dielectric electrophoresis treatment The number of metal particles after treatment was 580 / mL. The value was 175 cells / mL.

[0133] The metal particle removal efficiency is calculated as (580-175) / 580 = 405 / 580 = 0.698, which is 69.8%. A removal efficiency threshold of 0.75 is set. Since 0.698 < 0.75, the current removal efficiency has not reached the expected target. Therefore, the dielectrophoresis control intensity is increased, specifically by increasing the excitation voltage of the dielectrophoresis electrode from 500V to 650V until the removal efficiency reaches or exceeds the set threshold.

[0134] The improvement of dielectrophoresis control intensity includes at least one of the following methods: increasing the excitation voltage amplitude of the dielectrophoresis electrode, decreasing the excitation voltage frequency, and extending the dielectrophoresis action time;

[0135] Voltage surges during shift changes can drastically alter the internal electric field distribution of a transformer, exacerbating the risk of migration and accumulation of metal particles in the insulating oil. This invention identifies this hidden physical connection and simultaneously initiates dielectric electrophoresis control during the standby transformer takeover process. It actively applies a directional electric field to guide particles to the collection area and remove them, thus timely interrupting the physical chain of "voltage surge → particle accumulation → insulation degradation" at the moment of most intense electric field disturbance. This upgrades the switching process from a simple electrical operation to a composite operation of electrical operation and insulation synergistic maintenance, preventing particle accumulation from evolving into partial discharge or even insulation breakdown accidents. Dielectric electrophoresis control is a prior art technology.

[0136] There is a brief physical gap in time during gear switching, namely a millisecond-level transition period from the current gear being disconnected to the target gear being connected. During this gap, the transformer is in a state of no excitation or transient excitation, and the internal electric field undergoes severe disturbances. This is precisely the period when metal particles are most actively driven by dielectric electrophoresis. This invention utilizes this inherent switching gap to simultaneously apply dielectric electrophoresis control, transforming particle migration from "disordered aggregation" to "ordered removal." This embeds insulation maintenance operations without increasing additional time costs, realizing the transformation from a "switching gap" to a "maintenance window," which is the core advantage of this solution.

[0137] Since the switching gap itself has an effective duration of several seconds to tens of seconds, and each switching can produce a maintenance effect, it has a significant effect on suppressing the continuous rise of the concentration of metal particles in the insulating oil over a long period of time, avoiding the problem of passively shutting down the machine only after the particle concentration gradually rises to a dangerous level.

[0138] Thus, transformer T-2026 completed its upgrade within the target upgrade period of 12:00-13:00 noon, and the standby transformer T-1026 successfully took over its power terminal. During the takeover process, it completed the tap switching and coordinated maintenance of metal particles in the insulation medium. The entire process formed a complete closed loop of "time period decision-making → peak shifting → risk control → standby scheduling → switching and coordinated maintenance", and the verification was successful.

[0139] To verify the effectiveness of this solution, a 10kV / 0.4kV oil-immersed transformer (rated capacity 1250kVA) with serial number T-2026 in an industrial park substation was used as the experimental object. Comparative experiments were conducted using both the traditional manual upgrading method and the method of this invention. The experimental data are as follows:

[0140] I. Comparison of Upgrade Window Identification Efficiency: Before adopting this solution, maintenance personnel needed to check all four power terminals one by one, manually analyze historical load data, and predict future trends. The entire process took approximately 4.5 hours. Furthermore, due to the subjectivity of manual judgment, the low-risk window of 12:00-13:00 was not identified, and 3:00-5:00 AM was ultimately chosen as the upgrade period. After adopting this solution, the system automatically completes the extraction of operational features, normalization processing, calculation of power sensitivity indicators, and multi-period traversal evaluation. The entire process takes approximately 6 minutes, accurately identifying 12:00-13:00 as the target upgrade period, improving the efficiency of time-period decision-making by approximately 97.8%.

[0141] II. Comparison of Risk Control During the Upgrade Process: From 3:00 AM to 5:00 AM, while the L1 data center cabinet and L3 research and experimental platform were under low load, the L2 industrial robot production line suddenly encountered an emergency production task at 4:17 AM, causing the load to surge from 60kW (standard 0.048) to 420kW (standard 0.336). The traditional solution, lacking dynamic risk perception capabilities, resulted in transformer overload tripping, causing a power outage on the L3 research and experimental platform and directly resulting in the loss of approximately 2.5 hours of experimental data. After adopting this solution, the upgrade was performed from 12:00 PM to 1:00 PM. During the process, the load change rate Δp = 0.384 was monitored in real time, and the upgrade risk value d = 0.287 was calculated. This successfully triggered the medium-risk standby scheduling strategy, and the T-1026 standby transformer completed takeover at 12:03 PM, achieving zero-interruption load switching and improving power supply reliability from 89.3% to 99.97%.

[0142] III. Comparison of Backup Transformer Selection Accuracy: The traditional approach relies on maintenance personnel's subjective judgment based on experience to select a backup transformer from three candidate transformers. Historical data shows that the accuracy rate of manually selected backup transformers matching the current transformer to be upgraded is approximately 66.7% (selecting 2 out of 3). With this approach, qualified backup transformers (T-1024 and T-1026) are screened using the operational evaluation value R, and then precisely compared using the backup compatibility score g. The g values ​​are 0.853 and 0.854 respectively, a difference of only 0.001, achieving quantitative optimization and a matching accuracy rate of 100%.

[0143] IV. Comparison of Insulation Maintenance Efficiency: The traditional solution requires one shutdown window per year for insulating oil filtration maintenance, with each shutdown lasting approximately 4 hours. Based on the average annual power outage time of the transformer, this equates to approximately 6.5 hours per year. With this solution, the dielectric electrophoresis control is synchronously triggered during the tap change process of the standby transformer. When the voltage change ΔU=4V, particle migration treatment is automatically initiated, achieving a removal efficiency of 69.8%. Insulating oil maintenance can be completed without additional shutdown. It is projected that the average annual power outage time of the transformer will be reduced from 6.5 hours / year to 2.2 hours / year, increasing equipment uptime by approximately 66.2%.

[0144] Overall experimental results: After adopting this solution, the efficiency of window recognition is improved by approximately 97.8%, the power supply reliability is improved from 89.3% to 99.97%, the accuracy of backup transformer matching is improved from 66.7% to 100%, and the average annual power outage time is expected to be reduced by approximately 66.2%.

[0145] Example 2, refer to Figure 2 An artificial intelligence big data monitoring system for power systems, comprising:

[0146] The feature extraction module is used to collect the operating data of each power terminal of the power distribution line, extract the average power, peak power and low load operating period, and normalize each parameter.

[0147] The upgrade period decision module is used to receive operational characteristic data, calculate the transformer's load activity index, construct a period evaluation function and compare it with a threshold to generate the target upgrade period;

[0148] The staggered scheduling module is used to sort multiple transformers to be upgraded from high to low based on the time period evaluation function, and select the transformers at the top of the sort to be upgraded within their respective target upgrade time periods.

[0149] Upgrade the risk control module to acquire control characteristics in real time during transformer upgrades, calculate upgrade risk values, and execute tiered control based on the upgrade risk values.

[0150] The switching collaborative maintenance module acquires the output voltage before and after the on-load tap changer switching and calculates the voltage change during the process of the standby transformer taking over the power terminal. It then drives the metal particles in the insulating medium to migrate and dynamically adjusts the dielectric electrophoresis control intensity based on the particle removal efficiency.

[0151] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0152] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An artificial intelligence big data monitoring method for power systems, characterized in that: include: The transformer is upgraded based on the operating characteristics of the power consumption terminal, specifically as follows: The transformer supplies power to multiple electrical terminals through power distribution lines, and the operating characteristics of each electrical terminal are extracted; For a single transformer: The decision-making strategy for the upgrade period is implemented, the load activity index of the transformer is calculated based on the operating characteristics, the time period evaluation function for transformer upgrade is constructed, and the target upgrade period is generated; Multiple transformers to be upgraded are sorted according to the target upgrade period, and the transformers are upgraded in staggered time according to the sorting results within each target upgrade period. When upgrading a transformer, an upgrade risk control strategy is implemented, the transformer's control characteristics are calculated in real time, the upgrade risk value is calculated based on the control characteristics, and graded control is implemented according to the upgrade risk value. The hierarchical control includes a backup transformer scheduling strategy; Implement the standby transformer scheduling strategy, and calculate the standby compatibility score between the upgraded and verified transformers obtained from the peak-shifting upgrade and the current transformer to be upgraded. The upgraded and verified transformer with the highest standby adaptation score will be used as the standby transformer. Control the backup transformer to take over the power consumption terminal of the transformer currently to be upgraded; During the standby transformer takeover process: The switching and collaborative maintenance strategy is implemented by adjusting the tap position through the on-load tap changer and simultaneously controlling the migration of metal particles in the insulating medium.

2. The artificial intelligence big data monitoring method for power systems according to claim 1, characterized in that: The decision-making strategy for the upgrade period, based on the operating characteristics, calculates the transformer's load activity index, including: The operating characteristics include the average power, peak power, and low-load operating periods of the power terminals; A set of power-consuming terminals powered by a single transformer; Calculate the electricity sensitivity index of each element in the electricity terminal set in sequence. ; ,in, These represent the average power, peak power, and low-load operating period of the power consumption terminal, respectively. It is a very small positive number. Let be the aggregate average power of the set of power terminals, and These are the normalized values; The power consumption sensitivity index is negatively correlated with low load operating time and aggregate average power, and positively correlated with average power and peak power. The weight of any element in the set of electricity terminals is calculated as follows: , The total number of centralized elements in the power consumption terminal; Based on the aforementioned weights, a weighted average of the centralized electricity consumption sensitivity indicators of the electricity terminals is calculated, and the result is recorded as the transformer load activity index. .

3. The artificial intelligence big data monitoring method for power systems according to claim 2, characterized in that: The aforementioned time-period evaluation function for transformer upgrades generates the target upgrade time period, including: Divide 24 hours into multiple time periods and iterate through any future time period in sequence; Predicted power consumption of electrical terminals in future periods based on operational characteristics. ; ,in, To predict the adjustment coefficient, , which is the power fluctuation term constructed based on peak power and ensemble average power; Constructing a time-period evaluation function for transformer upgrades ; The mean value of the centralized predicted power of the power terminals is obtained. ; ,in, As a weighting coefficient, the lower the predicted power, the higher the load activity index, and the higher the time period evaluation function; Set evaluation threshold ; like The future time period currently being traversed is recorded as the target upgrade time period.

4. The artificial intelligence big data monitoring method for power systems according to claim 3, characterized in that: The process of prioritizing multiple transformers to be upgraded based on target upgrade periods, and then upgrading the transformers in staggered shifts according to the prioritization results within each target upgrade period, includes: Set an upgrade evaluation period, and sort multiple transformers to be upgraded from high to low based on the transformer's time period evaluation function to determine the target transformer for upgrade; The target upgrade period is included within the upgrade evaluation period; Set the upgrade quantity c; During the upgrade assessment period, the top-ranked transformer (ranked C) is selected as the target transformer for upgrade. The remaining transformers will be re-submitted for re-evaluation during a subsequent upgrade assessment period.

5. The artificial intelligence big data monitoring method for power systems according to claim 2, characterized in that: The implementation of the upgraded risk control strategy involves calculating the transformer's control characteristics in real time, calculating the upgraded risk value based on the control characteristics, and implementing tiered control according to the upgraded risk value, including: Obtain the control characteristics of the transformer; The control features include load activity index and load change rate. and upgrade progress factor e; Normalized load activity metrics, load change rate, and upgrade progress factor; Calculate the load change rate ,in, Let be the power of the transformer at time t. For a moment Transformer power, For time intervals; The upgrade progress factor is used to characterize the completion level of the current upgrade task, and is the ratio of the current completed upgrade workload to the total upgrade workload. Calculate upgrade risk value , These are the weighting coefficients; The upgrade risk value is positively correlated with the load activity index in a non-linear manner, and positively correlated with the load change rate and upgrade progress factor, so as to amplify the risk difference in high load scenarios and suppress risk fluctuations in low load scenarios. Set a Level 1 risk threshold and secondary risk threshold ,and ; like If so, the upgrade operation will continue; like If so, the standby transformer scheduling strategy will be executed; like If so, then an upgrade rollback operation will be performed.

6. The artificial intelligence big data monitoring method for power systems according to claim 5, characterized in that: The implementation of the standby transformer scheduling strategy, based on the upgraded and verified transformers obtained from the peak-shifting upgrade, calculates the standby adaptability score between the upgraded and verified transformers and the current transformer to be upgraded, including: Obtain any upgraded transformer and its remaining power supply capacity. Gear adjustment capability and operational reliability The remaining power supply capacity is then normalized. Set performance evaluation values; The operational evaluation value is the average of the remaining power supply capacity, the tap adjustment capability, and the operational reliability. Based on the comparison between the operation evaluation value and the set evaluation threshold, when the operation evaluation value is greater than or equal to the evaluation threshold, the current transformer is recorded as an upgraded and verified transformer. Calculate the standby adaptation score for the upgraded and verified transformer. : ,in, These are the weighting coefficients; The upgraded and verified transformer with the highest standby adaptation score will be used as the standby transformer. The backup transformer is controlled to take over the power supply load of the transformer to be upgraded. Specifically, the backup transformer is switched and adjusted by means of an on-load tap changer according to the change in power supply load.

7. The artificial intelligence big data monitoring method for power systems according to claim 6, characterized in that: The implementation of the switching and coordinated maintenance strategy, which involves adjusting the tap position via an on-load tap changer and simultaneously controlling the migration of metal particles in the insulating medium, includes: Obtain the output voltage of the on-load tap changer of the standby transformer before and after the tap position switching, respectively. and ; Calculate the voltage change caused by gear shifting. ,in, It is used to characterize the degree of change in electric field distribution during gear switching. The greater the voltage change, the more obvious the electric field disturbance inside the insulating medium. Set voltage change threshold ,like If so, dielectric electrophoresis control is initiated to migrate the metal particles in the insulating medium; Obtain the number of metal particles before and after the treatment, respectively. and ; Calculate the removal efficiency of metal particles When the removal efficiency is lower than the set removal efficiency threshold, the dielectrophoresis control intensity is increased.

8. An artificial intelligence big data monitoring system for a power system, applied to the artificial intelligence big data monitoring method for a power system as described in claims 1-7, characterized in that: include: The feature extraction module is used to collect the operating data of each power terminal of the power distribution line, extract the average power, peak power and low load operating period, and normalize each parameter. The upgrade period decision module is used to receive operational characteristic data, calculate the transformer's load activity index, construct a period evaluation function and compare it with a threshold to generate the target upgrade period; The staggered scheduling module is used to sort multiple transformers to be upgraded from high to low based on the time period evaluation function, and select the transformers at the top of the sort to be upgraded within their respective target upgrade time periods. Upgrade the risk control module to acquire control characteristics in real time during transformer upgrades, calculate upgrade risk values, and execute tiered control based on the upgrade risk values. The switching collaborative maintenance module acquires the output voltage before and after the on-load tap changer switching and calculates the voltage change during the process of the standby transformer taking over the power terminal. It then drives the metal particles in the insulating medium to migrate and dynamically adjusts the dielectric electrophoresis control intensity based on the particle removal efficiency.

Images