A monitoring and maintenance system and method for a substation

Abstract

The present application relates to the technical field of power system automation, in particular to a kind of monitoring and maintenance system and method of transformer substation.It includes quantification classification module, difference analysis module and hierarchical adjustment module.The present application classifies the equipment in transformer substation by quantification classification module, difference analysis module introduces correction parameters such as aging and load loss for single equipment monitoring, combined with multi-level threshold, reduces the misjudgment and missed judgment caused by equipment state change, correlates the equipment with GNN model modeling, aggregates neighbor state, both outputs equipment health degree, and quantifies fault propagation probability, early warning cascading failure risk, hierarchical adjustment module takes risk response measures at the same time, linkage adjustment single equipment threshold, makes single equipment monitoring and correlation equipment risk state adaptive matching, realizes full-coverage accurate monitoring, and gives consideration to local and global risk.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system automation, in particular to a monitoring and maintenance system and method for a substation. BACKGROUND

[0002] As the core hub of the power generation, power transmission, power transformation and power distribution link of the power system, the equipment operation state of the substation directly determines the power supply reliability and operation and maintenance economy of the power grid. The traditional substation monitoring and maintenance system generally takes equipment state acquisition, abnormal alarm and manual maintenance as the core logic, acquires the electrical parameters (voltage, current, partial discharge) and physical state (temperature, vibration) of the transformer, switch cabinet and busbar through sensors, transmits the data to the station control layer or cloud platform, and after the operation and maintenance personnel on the cloud platform receive the alarm, they check the fault equipment on site, formulate and execute the maintenance scheme, and manually record the maintenance results after completion.

[0003] However, with the expansion of the substation scale, the increase of the equipment types and the improvement of the user's requirement for power supply reliability, the equipment in the substation is correlated when monitoring and maintaining the equipment in the substation, and different equipment will have different effects when being maintained. Some effects are large, and some effects are small. The traditional substation monitoring and maintenance system only monitors the equipment individually, ignores the correlation between the equipment, and may cause misjudgment when monitoring and maintaining the equipment, take wrong maintenance measures, and cause the equipment that needs to be maintained not to be maintained in time, resulting in irreparable loss to the substation. Therefore, we propose a monitoring and maintenance system and method for a substation. SUMMARY

[0004] The present application aims to provide a monitoring and maintenance system and method for a substation, which can solve any technical problem proposed in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides a monitoring and maintenance system and method for a substation, comprising:

[0006] The quantitative classification module acquires the electrical connection data and parameter correlation data, calculates the electrical connection degree and data correlation degree of the equipment, and quantifies the fault influence range of the equipment to obtain a comprehensive score. The comprehensive score is compared with a preset threshold value to divide the equipment type into a single equipment type and an associated equipment type.

[0007] The difference analysis module uses a threshold comparison method to monitor the state of the single equipment, acquires the functional parameters of the single equipment in real time, compares them with the rated parameters of the single equipment, determines the abnormal condition of the single equipment, uses a GNN global propagation model to monitor the state of the associated equipment, models the associated equipment and its connection relationship as a graph structure, takes the node feature matrix and the adjacency matrix as the input, uses the graph convolution layer to aggregate the information, and outputs the node health degree and the fault propagation probability.

[0008] The hierarchical adjustment module associates the device with high-risk anomalies, medium-risk anomalies and low-risk anomalies according to the node health degree and the fault propagation probability, and takes adjustment measures aiming at suppressing faults and reducing the impact on the normal operation of the power grid according to the anomaly level.

[0009] As a further improvement of the technical solution, the quantitative classification module comprises:

[0010] The acquisition calculation unit extracts the static electrical connection relationship and dynamic switch state between devices from the substation design drawings, calculates the electrical connection degree by counting the number of devices that can be directly and indirectly affected by the electrical connection and the connection strength, collects device operating parameters from sensors and intelligent instruments, and selects strong correlation parameter pairs, and calculates the data correlation degree by counting the number of strong correlation parameter pairs;

[0011] The quantitative comparison unit quantifies the fault influence range of the device by weighted summation, calculates the comprehensive score of the device, compares the comprehensive score of the device with the preset threshold value, and if the comprehensive score is lower than the preset threshold value, it is a single device type, and if the comprehensive score exceeds the preset threshold value, it is an associated device type.

[0012] As a further improvement of the technical solution, the quantitative comparison unit collects historical operation data of the substation devices, calculates the fault influence range of each device, and draws a distribution histogram of device fault influence using a data statistical analysis method, and takes the inflection point of the distribution curve as the initial preset threshold value.

[0013] As a further improvement of the technical solution, the quantitative comparison unit real-time collects the actual operation load of the devices in the substation, calculates the load rate of the device, classifies the operating conditions of the device, determines the adjustment range of the initial threshold value according to the amplitude of the device fault influence of the operating condition level, and establishes a rule mapping table of operating conditions and threshold values.

[0014] As a further improvement of the technical solution, the difference analysis module comprises:

[0015] The single device monitoring unit determines the correction parameter according to the cumulative running time of the single device, corrects the rated parameter, compares the real-time functional parameter of the single device with the corrected rated parameter, outputs the abnormal condition of the single device, and sends a warning signal;

[0016] The associated device monitoring unit takes each associated device as a node, takes the electrical connection relationship between devices as an edge and assigns a weight, and then converts the node and edge into an adjacency matrix, inputs the adjacency matrix and node feature matrix into a GNN global propagation model, and through a graph convolution layer, each node fuses its own and neighbor device state information, and outputs the node health degree and fault propagation probability of the associated device.

[0017] As a further improvement of the technical solution, the single-device monitoring unit obtains the cumulative running time of the single device, calculates the load loss coefficient and the aging correction coefficient according to the cumulative full-load time proportion, and determines the final correction parameter.

[0018] As a further improvement of the technical solution, the associated device monitoring unit divides the connection into hard connection, semi-hard connection and soft connection according to the functional attribute of the electrical connection, determines the basic weight of the edge between devices according to the connection type, and corrects the basic weight through the key parameters of different connections.

[0019] As a further improvement of the technical solution, the hierarchical adjustment module aims to minimize the outage range for high-risk abnormalities, to prevent fault escalation for medium-risk abnormalities, to fine-tune stability for low-risk abnormalities, and to minimize intervention for normal operation.

[0020] As a further improvement of the technical solution, the hierarchical adjustment module determines the adjustment direction of the threshold value in the single device according to the division of the associated device into different abnormal levels after taking the adjustment measures, determines the threshold adjustment amplitude according to the node health degree and the fault propagation probability, and adjusts the threshold value of the single device.

[0021] Compared with the prior art, the present application has the following advantages:

[0022] 1. The substation monitoring and maintenance system and method, the quantitative classification module quantifies and classifies the devices in the substation based on electrical connection and operating parameter, avoids subjective bias, sets the initial threshold value according to historical fault data, and dynamically adjusts the threshold value according to the device load rate to adapt to different operating scenarios of the substation, ensuring that the classification is consistent with the actual fault influence of the device.

[0023] 2. The difference analysis module introduces correction parameters such as aging and load loss for single device monitoring, combines multi-level threshold values, reduces misjudgment and missed judgment caused by changes in device state, uses GNN model for associated devices, aggregates neighbor states, outputs device health degree and quantifies fault propagation probability, and early warns of the risk of chain failure, realizing full-coverage and accurate monitoring, and taking into account local and global risks.

[0024] 3. The hierarchical adjustment module quickly isolates faults to protect core power supply for high-risk, fine-tunes to prevent escalation for medium-risk, reduces intervention for low-risk, avoids excessive operation to disturb the power grid, and adjusts the threshold value of the single device while taking risk response measures, so that the single device monitoring and the risk state of the associated device are adaptively matched.

[0025] In addition to the purposes, features and advantages described above, the present application has other purposes, features and advantages. The present application will be further described below with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a schematic diagram of the overall flow of the present application;

[0027] Figure 2 is a schematic diagram of the overall detailed flow of the present application;

[0028] Figure 3 is a schematic diagram of the method flow of the present application.

[0029] The meanings of the various reference numerals in the figures are as follows:

[0030] 100, quantification classification module; 110, acquisition calculation unit; 120, quantification comparison unit; 200, difference analysis module; 210, single device monitoring unit; 220, associated device monitoring unit; 300, hierarchical adjustment module. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0032] At present, with the expansion of the scale of the substation, the increase of the types of equipment and the improvement of the requirement of the user for the reliability of power supply, when the equipment in the substation is monitored and maintained, the equipment in the substation is associated with each other, and when the maintenance is performed, different influences will be produced on different equipment, some influences are larger, and some influences are smaller. The traditional substation monitoring and maintenance system only monitors the equipment individually, ignores the association between the equipment, and will lead to misjudgment when the equipment is monitored and maintained, takes wrong maintenance measures, and leads to that the equipment needing maintenance is not maintained in time, and causes irreparable loss to the substation.

[0033] Therefore, the present application proposes that the quantification classification module divides the equipment into single device type and associated device type by calculating the electrical connection degree, the data association degree and the comprehensive score, directly provides the classification basis of the monitoring object for the difference analysis module, the difference analysis module outputs the core monitoring result for the two types of equipment, directly provides the risk judgment and the measure making input for the hierarchical adjustment module, and the hierarchical adjustment module reversely adjusts the threshold value of the single device local threshold diagnosis model in the difference analysis module according to the risk level of the associated equipment.

[0034] Specifically as follows:

[0035] Please refer to Figure 1As shown, the present application provides a monitoring and maintenance system and method for a substation, including a quantitative classification module 100, collecting electrical connection data and parameter correlation data, calculating the electrical connection degree and data correlation degree of the equipment, and quantifying the fault influence range of the equipment to obtain a comprehensive score, comparing the comprehensive score with a preset threshold, and dividing the equipment type into single equipment type and associated equipment type.

[0036] Among them, the electrical connection data refers to the data reflecting the physical and functional connection relationship and real-time on-off state between the substation equipment, and the core role is to define which devices are associated with each other, providing a basis for subsequent determination of whether the fault will spread between the devices, and the parameter correlation data refers to the data reflecting the running state of the substation equipment itself and the strong correlation between different devices and parameters, and the core role is to determine whether the equipment is running normally and which parameter changes will affect each other, providing a basis for equipment abnormal diagnosis and risk warning.

[0037] As shown, Figure 2 The quantitative classification module 100 includes:

[0038] The acquisition and calculation unit 110 extracts the static electrical connection relationship and dynamic switch state between the devices from the substation design drawings, calculates the electrical connection degree by counting the number of devices that can be directly and indirectly affected by the electrical connection and the connection strength, and the formula is as follows:

[0039] Electrical connection degree = (directly connected device number × a + indirectly connected device number × b) ÷ maximum possible connection number;

[0040] Among them, a is the weight of the number of directly connected devices, and the directly connected device is the core path of fault propagation. Fault current, voltage anomaly, etc. can be transmitted immediately and without attenuation through direct connection (such as bus short circuit which will directly cause overload of connected main transformer), and the influence on the target device has directness and certainty, which is the main carrier of fault chain propagation, and needs to be given the highest weight to reflect its decisive role, generally set to 1;

[0041] b is the weight of the number of indirectly connected devices, and the indirectly connected device needs to be transferred through the directly connected device. The fault propagation has delay, attenuation and blockability (such as indirect device fault which can be isolated by intermediate switch tripping without affecting the target device), and the influence on the target device has indirectness and contingency, which is determined according to the number of devices between the device and the target device (the more the number of intermediate devices, the lower the weight of the number of indirectly connected devices)

[0042] Statistical current in the on state of the connected device, wherein, the directly effective connected device refers to the device which has physical direct connection with the target device (such as wire, bus, copper directly connected), and the switch device (circuit breaker, disconnecting switch) on the connection path is in the closed on state, the list of all physical connection devices of the target device is obtained from the CAD primary wiring diagram of the substation, the opening and closing state of the switch device on the connection path is obtained from the SCADA system in real time, the connection disconnected by opening is excluded, and the number of devices directly connected with the target device and the switch closed in the connection list is counted, that is, the number of directly effective connected devices,

[0043] The indirectly effective connected device refers to the device which is not directly physically connected with the target device, but is indirectly connected through the first-level directly effective connected device, and all the switches in the middle are in the closed on state, the number of devices directly connected with the target device and the switch closed in the connection list is counted, that is, the number of indirectly effective connected devices;

[0044] The maximum possible connection number refers to the maximum number of devices that can be directly connected in theory based on the physical wiring topology and functional positioning of the target device in the design stage of the substation, which is directly obtained from the design drawings of the substation and the technical manual of the device;

[0045] The electrical connection degree is calculated by the above formula, which not only considers whether there is physical connection, but also combines the real-time state (closing / opening) of the switch to filter the effective connection, avoiding counting the invalid connection disconnected by opening into the actual operation scene;

[0046] The running parameters of the sensor and intelligent instrument are collected, and the strong correlation parameter pairs are selected, and the data correlation degree is calculated by counting the number of strong correlation parameter pairs, and the formula is as follows:

[0047] Data correlation degree = (core strong correlation pair number × c + secondary strong correlation pair number × d) ÷ target device key parameter total number;

[0048] Wherein, c is the weight of the number of core strong correlation pairs, the core parameter is the direct cause of device failure, such as overrated current directly leading to overload failure, power anomaly directly reflecting the overloading of load, its correlation to the judgment of device abnormality has decisive role, and the highest weight should be given to ensure the accuracy of fault diagnosis, generally set to 1;

[0049] D is the weight of the number of secondary strong correlation pairs, the secondary parameter is the indirect signal of device failure, such as temperature rise, which is the result (not the reason) of overload, and vibration anomaly needs to be combined with current parameter to judge whether it is a fault (pure vibration high may be installation problem), its correlation only plays an auxiliary role, which is determined according to the functional priority of the parameter (the greater the influence on the core parameter, the higher the weight d of the number of secondary strong correlation pairs);

[0050] Among them, the core strong correlation pair refers to the vital parameter pair of equipment operation. The linear and nonlinear correlation between the parameters is very strong, and directly reflects the safe operation boundary of the equipment. The abnormality will directly cause equipment failure or trigger protection action. The real-time running parameters (such as current, voltage, power, temperature, vibration, and partial discharge amount) of the target equipment are obtained from the intelligent instrument. The time granularity is unified to 15 minutes. The correlation of the parameter pair is calculated by statistical method. The number of parameter pairs directly related to the equipment is recorded, that is, the number of core strong correlation pairs;

[0051] The secondary strong correlation pair refers to the auxiliary parameter pair of equipment operation. The correlation between the parameters is relatively strong, and only indirectly reflects the equipment state. The abnormality is a signal of potential problem of the equipment. The number of parameters that only assist in reflecting the equipment state is counted, that is, the number of secondary strong correlation pairs.

[0052] The total number of target equipment key parameters refers to the total number of all core and secondary operating parameters that need to be monitored for a certain target equipment (such as a main transformer, a circuit breaker, and a busbar) from the perspective of ensuring its safe and stable operation. The monitoring parameters are obtained from the factory technical manual of the equipment.

[0053] The above formula is used to calculate the data correlation degree of the equipment, distinguish the core and secondary strong correlation pairs, avoid regarding the auxiliary key parameters as the same, and ensure that the data correlation degree meets the priority requirement of fault diagnosis.

[0054] Among them, the quantitative comparison unit 120 quantifies the fault influence range of the equipment by weighted summation, calculates the comprehensive score of the equipment, compares the comprehensive score of the equipment with the preset threshold value, and if the comprehensive score is lower than the preset threshold value, it is a single equipment type, and if the comprehensive score exceeds the preset threshold value, it is an associated equipment type.

[0055] According to the above calculated electrical connection degree and data correlation degree, in the power system, the electrical connection degree has a stronger decisive influence on the fault influence range, and the parameter correlation degree only plays an auxiliary reflection role, which is the core basis of weight allocation. According to the past 3-5 years of substation fault cases, the number of cases of associated equipment failure caused only by electrical connection and the number of cases of associated equipment failure indirectly caused only by parameter correlation are counted respectively. According to the formula:

[0056]

[0057] Among them, is the weight of the electrical connection degree, is the weight of the data correlation degree, is the number of cases of associated equipment failure caused only by electrical connection, is the number of cases of associated equipment failure indirectly caused only by parameter correlation.

[0058] The failure influence range of the equipment is weighted and summed according to the formula, and the formula of the comprehensive score is as follows:

[0059]

[0060] Wherein, is the comprehensive score, is the electrical connection degree, is the data correlation degree, is the electrical connection degree weight, is the data correlation degree weight.

[0061] Wherein, the quantitative comparison unit 120 collects the historical operation data of the substation equipment, calculates the failure influence range of each equipment, and draws a distribution histogram of the equipment failure influence by using a data statistical analysis method, and takes the inflection point of the distribution curve as the initial preset threshold;

[0062] According to the formula of the comprehensive score, the failure influence range of each equipment is calculated, the interval is divided according to the comprehensive score range (0 to 1), the group interval is set to 0.1 (taking into account the resolution and simplicity), there are 10 groups, the score interval is taken as the X axis, the equipment frequency is taken as the Y axis, each interval corresponds to a rectangular column, the column height is the frequency of the interval, and the column width is the group interval (0.1). Add chart title (such as substation equipment failure influence comprehensive score distribution histogram), axis label (X axis: failure influence comprehensive score, Y axis: equipment quantity);

[0063] On the basis of the histogram, the frequency distribution curve is added, the shape of the distribution curve is observed, the point where the curve changes from steep rise (representing rapid increase of equipment quantity) to gentle rise and starts to decline is found, the X axis (comprehensive score) corresponding to the point is the inflection point, and the comprehensive score corresponding to the located inflection point is taken as the initial preset threshold;

[0064] For example, in a 110kV substation with 30 equipment, the main transformer and bus are classified as associated equipment according to traditional experience, and the rest are classified as single equipment. In practice, 1 circuit breaker failure affects the area (should be associated) but is not monitored; 1 voltage transformer failure only affects metering (should be single equipment) but is globally monitored; the threshold is set to 0.5, which cannot explain the rationality, and the adjustment still misjudges;

[0065] The weights (electrical connection degree 0.6, data correlation degree 0.4) are determined by data statistics, the comprehensive score is calculated according to the comprehensive score formula F=electrical connection degree×0.6+data correlation degree×0.4: main transformer 0.76, fault circuit breaker 0.56, surge arrester 0.16, voltage transformer 0.26, the score is divided into 10 groups (0-0.1 to 0.9-1.0, group interval 0.1), the number of equipment in each interval is counted, the histogram is drawn and the distribution curve is added;

[0066] The observation curve shape shows that the curve gently rises in the 0-0.5 interval (the number of devices slowly increases, mostly weakly affected devices such as lightning arresters and voltage transformers); the curve steeply rises in the 0.5-0.6 interval (the number of devices rapidly increases to 5, including moderately affected devices such as faulty circuit breakers); the curve gently rises and then falls above 0.6 (the number of devices increases at a slower rate, mostly strongly affected devices such as main transformers and busbars), and the critical point where the curve changes from gentle rise to steep rise corresponds to a score of 0.5 on the X-axis. This point is the dividing point, so 0.5 is set as the initial preset threshold value.

[0067] By combining the score and the threshold value, the faulty circuit breaker (0.56>0.5) is accurately classified as an associated device to avoid underestimating the impact range, and the voltage transformer (0.26<0.5) is classified as a single device to reduce invalid monitoring.

[0068] In order to adjust the initial threshold value according to the actual operating conditions of the device, the quantitative comparison unit 120 real-time collects the actual operating load of the device in the substation, calculates the load rate of the device, classifies the operating conditions of the device, determines the adjustment range of the initial threshold value according to the magnitude of the device fault impact of the operating condition level, and establishes a rule mapping table of the operating condition and the threshold value.

[0069] The essence of the device load rate is the percentage of the actual load and the rated load, and the formula is:

[0070]

[0071] Among them, is the device load rate, is the actual operating load of the device in a certain time period, is the rated load of the device as specified by the manufacturer.

[0072] Under different operating condition levels, the device load, aging speed, and fault conduction efficiency are different, which directly leads to differences in fault impact magnitude. According to the device load rate, the device operating conditions are divided into light load conditions, normal conditions, and heavy load conditions, and the fault impact magnitude difference is quantified. The light load condition is reduced by 20%-30%, the normal condition has no significant change, and the heavy load condition is expanded by 15%-40%.

[0073] The adjustment range needs to be based on the percentage change of the fault impact magnitude, and the percentage adjustment method is used. The specific adjustment ratio is as follows:

[0074]

[0075] According to the four core dimensions of device type, working condition level (light load, normal, heavy load), corresponding load rate range, and initial threshold value, the adjusted threshold value is calculated by the logic of increasing by 5-10% for light load, decreasing by 10-20% for heavy load, and remaining unchanged for normal, and the application rules for determining the device type are supplemented by whether the comprehensive score exceeds the threshold value, forming a structured table;

[0076] The table is then digitized and entered into the substation automation system, indexed by device ID and working condition level, facilitating real-time acquisition of load rate and automatic matching of threshold value. Finally, the table is updated quarterly in combination with device parameter changes and fault data calibration to ensure adaptation to actual operation and maintenance needs.

[0077] The comprehensive score obtained by the above calculation is compared with the adjusted preset threshold value. If the comprehensive score is lower than the preset threshold value, it is a single device type, and if the comprehensive score exceeds the preset threshold value, it is a related device type.

[0078] In addition, the difference analysis module 200 uses the threshold comparison method for single device state monitoring. By real-time acquisition of single device functional parameters and comparison with single device rated parameters, the abnormal condition of single device is determined.

[0079] The core of the threshold comparison method is to take the device state as the core, compare the real-time parameters of single device with the corrected rated parameters, and not rely on neighboring device information. It is suitable for single device type (such as lightning arrestor, voltage transformer) whose fault impact is limited to itself. According to the correction of rated parameters, the real-time acquisition of single device functional parameters is compared with the rated parameters of single device to determine the abnormal condition of single device. If the real-time acquisition of functional parameters is lower than the rated parameters, the single device is normal, and if the real-time acquisition of functional parameters is higher than the rated parameters, the single device is abnormal, and an alarm signal is issued.

[0080] For example, in a certain 110kV substation, 3 sets of 10kV lightning arrestors (whose fault impact is limited to itself, meeting the definition of single device type) need to be monitored by the threshold comparison method to avoid the problem of missed judgment of real-time parameters not exceeding the original rated value but close to the actual safety critical value due to lightning arrestor insulation aging and load loss, or false alarm of normal fluctuation triggering a fixed threshold.

[0081] Real-time collection of the operating voltage and leakage current (two key functional parameters) of each surge arrester through sensors, filtering and denoising preprocessing (excluding parameter jumps caused by transient interference of the power grid), for example, the operating voltage of the surge arrester is 10.2 kV and the leakage current is 0.35 mA at a certain time period, and the original rated parameters are extracted from the surge arrester equipment nameplate, wherein the rated operating voltage is 10 kV (the upper limit of the voltage for long-term safe operation) and the rated leakage current is 0.5 mA (the maximum allowable leakage current under normal insulation state);

[0082] Extracting the surge arrester basic data from the substation PMS system, the cumulative total operating time (excluding maintenance time) is 50,000 hours, the cumulative full load time is 30,000 hours, and the load interval time is 0-5 kV (T1) 5,000 hours, 5-9.5 kV (T2) 15,000 hours, and ≥9.5 kV (T3) 30,000 hours, and the load loss coefficient is calculated according to the formula;

[0083] Substituting the corrected parameters into the threshold comparison method, the original rated parameters are corrected, and the preprocessed real-time functional parameters are compared with the corrected rated parameters. The operating voltage comparison is that the real-time 10.2 kV is greater than the corrected rated 9.4 kV, and the leakage current comparison is that the real-time 0.35 mA is less than the corrected rated 0.47 mA, and it is determined that the parameters are normal.

[0084] There is one real-time parameter higher than the corrected rated parameter, so the surge arrester is determined to be abnormal as a whole, and the system immediately sends out an audible and visual alarm signal, and a pop-up window is prompted on the background operation and maintenance interface that the operating voltage of the surge arrester is out of limit, and the insulation state needs to be checked.

[0085] Considering that the traditional single-device monitoring only compares the real-time parameters with the rated parameters, without considering the decline in adaptability of the rated parameters caused by device load loss and aging, false positives may occur, wherein the difference analysis module 200 comprises:

[0086] The single-device monitoring unit 210 determines the correction parameters according to the cumulative operating time of the single device, corrects the rated parameters, compares the real-time functional parameters of the single device with the corrected rated parameters, outputs the abnormal conditions of the single device, and sends out a warning signal;

[0087] Real-time collection of device functional parameters (such as current, temperature, pressure, etc.) and device rated parameters (such as rated current, maximum allowable temperature, from the device nameplate) through sensors or intelligent instruments, and filtering and denoising preprocessing are performed;

[0088] In order to better determine the correction parameters, the single-device monitoring unit 210 acquires the cumulative operating time of the single device, calculates the load loss coefficient according to the cumulative full load time proportion, and determines the final correction parameters;

[0089] Extract basic data from substation PMS or SCADA system: obtain the total running time of equipment (Ttotal, deducting downtime for maintenance), the total time of full load (Tfull, the running time when the load rate is greater than 90%), and the running time of equipment in different load intervals (such as 0-50%, 50%-80%, and greater than 90%) (T1, T2, T3);

[0090] Calculate the load loss coefficient (Kloss): first calculate the proportion of each load interval (such as T1 / Ttotal, T2 / Ttotal), then take the average load of each interval (such as 25% for 0-50%, 65% for 50%-80%), and according to the formula:

[0091]

[0092] Calculate the degree of additional loss caused by load fluctuation. The larger the Kloss, the more serious the loss. Determine the final correction parameter. If the correction current, power, and other load-related thresholds are corrected, the device aging effect (a fixed aging coefficient such as 0.95 can be simplified) needs to be combined. The final correction parameter = Kloss x aging coefficient; If only the load loss related threshold is corrected, Kloss is directly used as the final correction parameter, and the rated parameter is adjusted;

[0093] For example, a substation transformer (in operation for 8 years) has a rated current of 1000A. In traditional monitoring, the real-time current is 950A (lower than the rated value) and is determined to be normal, but in fact, due to the long-term full load operation (the cumulative full load time accounts for 60%), the winding loss causes the actual safe current to drop to 920A, and the continuous operation at 950A has an overload risk and no warning. A newly commissioned arrester has a rated temperature of 80°C. In traditional monitoring, the real-time temperature is 78°C (close to the rated value) and is misjudged as abnormal. In fact, there is no load loss, and the safe temperature still meets the requirements, resulting in ineffective operation and maintenance.

[0094] Extract main transformer data from PMS / SCADA: Ttotal is 65000 hours, Tfull (load rate > 90%) is 39000 hours, and the time of each load interval T1 (0-50%) is 13000 hours, T2 (50%-80%) is 13000 hours, and T3 (> 90%) is 39000 hours;

[0095] Calculate the interval proportion (T1 / Ttotal is 20%, T2 / Ttotal is 20%, and T3 / Ttotal is 60%), take the interval average (25%, 65%, and 95%), and calculate the load loss coefficient (Kloss) according to the above formula (quantify the additional loss, Kloss is 1.08 here). The main transformer needs to correct the current threshold, combined with the aging coefficient 0.95, and the final correction parameter = 1.08 x 0.95 ≈ 1.03;

[0096] The main transformer corrected rated current = 1000A ÷ 1.03 ≈ 970A, the real-time current 950A is less than 970A, so it is determined to be normal (if not corrected, the original rated value 1000A will miss the risk of 950A close to the actual safe value); the new arrester has no obvious loss, the K loss is 1, the correction parameter is 1, and the rated temperature is still 80℃, and the real-time temperature is 78℃, so it is determined to be normal, avoiding false positives;

[0097] Through the above steps, the problem of decline in rated parameter adaptability caused by load loss and aging is solved, the rated parameter is fitted to the actual state of the equipment, and the monitoring accuracy is improved.

[0098] The GNN global propagation model is used for state monitoring of the associated equipment, the associated equipment and its connection relationship are modeled as a graph structure, the node feature matrix and the adjacency matrix are used as input, the graph convolution layer is used for information aggregation, and the node health degree (which is an index for quantifying the health degree of the current running state of the associated equipment, the closer the value is to 1, the more stable the equipment running state is) and the fault propagation probability (which is an index for quantifying the possibility of the fault of a certain associated equipment being transmitted to other associated equipment (target node), the value range is also 0-1) are output. The steps are as follows:

[0099] Each associated equipment is taken as a node, the electrical connection relationship between the equipment is taken as an edge and is assigned a weight, and then the node and the edge are converted into an adjacency matrix. The adjacency matrix and the node feature matrix are input into the GNN global propagation model, each node fuses the state information of itself and the neighbor equipment through the graph convolution layer, the fused node features output by the graph convolution layer are input into the fully connected layer, the feature dimension is mapped to 1 (single value output) through linear transformation, and then the output value is processed through the Sigmoid activation function, and the output value is compressed to the interval of 0-1. The value is the health degree of the corresponding node (equipment), that is, the node health degree.

[0100] For each source node (assuming it is the fault source equipment), the fused features of the source node and the features of all target nodes (other associated equipment) are input into another fully connected layer, and the probability mapping relationship from the source node to all target nodes is constructed through linear transformation. Then, the Softmax activation function is processed, so that the probability sum of the source node to each target node is 1. The output value corresponding to each target node is the probability of the source node transmitting to the target node after the fault, and the probability is the fault propagation probability.

[0101] All devices defined as associated equipment (such as main transformers, busbars, and circuit breakers) are defined as node objects, each node is assigned a unique ID (such as main transformer T1 and busbar B2), and the device type is recorded. Determine the node attribute. If there is an effective electrical connection (hard connection, semi-hard connection, and soft connection) between two associated equipment, an edge is established between the two nodes, and the edge weight is assigned a value.

[0102] Hard connection refers to a physical connection that can directly and continuously transmit electrical energy (or strong electrical signals) without any controllable switching devices (or switching devices are always closed and do not need to be disconnected). Faults can be immediately transmitted through the connection.

[0103] Semi-hard connection refers to an electrical energy transmission connection achieved through a controllable switching device (such as a circuit breaker or disconnector). When closed, it is equivalent to a hard connection (can transmit electrical energy and faults can be immediately transmitted). When open, the connection is disconnected (no electrical energy transmission and faults cannot be transmitted). The effectiveness of the connection depends on the state of the switch.

[0104] Soft connection refers to a connection used only for transmitting control signals and monitoring data (non-electrical energy). It has no electrical energy transmission capability, and faults need to be indirectly triggered through signals (with a delay and not necessarily transmitted). The effectiveness of the connection depends on the quality of the signal.

[0105] Convert the graph structure into an adjacency matrix, which is a numerical representation of the graph structure used to describe the connection relationship and weight between nodes. Let the matrix be A, and A[i][j] represents the edge weight from node i (device i) to node j (device j). Assuming there are 3 nodes, 1 is the main transformer T1, 2 is the bus B2, and 3 is the circuit breaker Q3, then the adjacency matrix is:

[0106]

[0107] The node feature matrix is a collection of device state information used for GNN to fuse neighborhood features. The dimension is "total number of associated devices (N) x feature dimension (D)", and it contains the core state features of the device. The following features are extracted from sensors, SCADA, and PMS systems for each device: real-time operating characteristics, historical state characteristics, and corrected characteristics. Let the feature dimension D=5 (real-time load rate, real-time temperature, cumulative full load proportion, load loss coefficient, and aging correction coefficient). Then the feature matrix is:

[0108]

[0109] Preprocess the adjacency matrix: add a self-loop (A'=A+I, I is the identity matrix) to allow the node to retain its own features when fused. Perform neighborhood feature fusion through 1-2 layer graph convolution layers, and the propagation formula is simplified as:

[0110]

[0111] where, is the fused feature, is the activation function, is the normalized matrix A, is the node feature of the l-th layer, is the trainable weight of the l-th layer, is the bias term.

[0112] With a full connection layer and a Sigmoid activation function, the health of each node is output (0-1, the closer to 1, the healthier), and with a full connection layer and a Softmax activation function, the probability of each node failure is output to all other nodes;

[0113] For example, a certain substation has 3 core associated devices (main transformer T1, bus B2, circuit breaker Q3), and traditional monitoring only collects individual device parameters: main transformer T1 monitors current and temperature, bus B2 monitors voltage and load, and circuit breaker Q3 monitors tripping state and operating mechanism temperature. Problems in actual operation: first, when the temperature of main transformer T1 is abnormal, it is misjudged as a minor abnormality based on T1's own parameters without combining the voltage fluctuation data of bus B2, and the potential overload risk is not timely investigated; second, it cannot predict the cascading impact of device failure, and in a certain short-circuit fault of main transformer T1, the probability of its transmission to bus B2 and circuit breaker Q3 is not known in advance, resulting in delay in tripping of bus B2 protection device and expansion of power outage range;

[0114] According to the connection type between devices, the edge weight is determined, main transformer T1 and bus B2 are hard connected (weight 0.9), bus B2 and circuit breaker Q3 are semi-hard connected (closing state, weight 0.8), and the weight between devices without direct connection is set to 0, forming a 3x3 adjacency matrix, clearly presenting the connection relationship and strength between nodes (devices); 5-dimensional core features (real-time load rate, real-time temperature, cumulative full load proportion, load loss coefficient, and aging correction coefficient) of each device are extracted from sensors, SCADA, and PMS systems to form a 3x5 node feature matrix, which stores real-time, historical, and corrected state information of devices, and a self-loop is added to the adjacency matrix (i.e., superimposing the identity matrix) to obtain a new matrix, ensuring that each device can obtain the state of adjacent devices while not losing its own state information (e.g., main transformer T1 retains its own temperature feature when fused);

[0115] A simplified graph convolution propagation formula is used, the normalized adjacency matrix and the initial node feature matrix (i.e., the above node feature matrix) are substituted into the calculation, and through 1-layer graph convolution, the neighborhood feature fusion is realized, main transformer T1 fuses the features of bus B2, bus B2 fuses the features of main transformer T1 and circuit breaker Q3, and circuit breaker Q3 fuses the features of bus B2, breaking the limitations of single-device isolated monitoring, and through a full connection layer combined with a Sigmoid activation function, the health of each device is output (value 0-1, the closer to 1, the healthier); through a full connection layer combined with a Softmax activation function, the probability of each device failure is output to other devices, providing a quantitative basis for risk judgment and early warning;

[0116] Through the above steps, after fusing the adjacent device state, the device health degree is no longer dependent on its own parameters. For example, the T1 health degree of the main transformer combined with the state of the bus B2 can more accurately determine the risk level, avoid the misjudgment problem of traditional isolated monitoring, and clearly output the fault propagation probability between devices (such as the probability of the main transformer T1 fault to the bus B2). The protection strategy of the associated device can be adjusted in advance (such as shortening the bus B2 trip delay), and the protection action delay can be avoided.

[0117] Considering that the traditional associated device monitoring only analyzes the state of a single device in isolation, the connection weight between devices (such as the difference between hard connection and semi-hard connection) is not quantified, and the state information of adjacent devices is not fused, wherein the associated device monitoring unit 220 divides the connection into hard connection, semi-hard connection and soft connection according to the functional attributes of electrical connection, determines the basic weight of the edge between devices according to the connection type, and modifies the basic weight through the key parameters of different connections;

[0118] The basic weight is determined according to the fault conduction ability, and the directness, speed and intensity of connection are classified and valued according to the fault conduction ability, wherein the hard connection can directly and immediately transmit fault current (such as short circuit), has the strongest conduction ability, and is given high weight. The semi-hard connection (closed state) can transmit fault as hard connection, and has the second strongest conduction ability and the second highest weight. The soft connection only indirectly transmits fault through signal (with delay and uncertainty), has the weakest conduction ability, and is given low weight. The weight quantifies the inherent influence of connection in fault propagation. The more direct the conduction, the greater the influence, and the higher the weight.

[0119] For hard connection, the contact resistance is modified (the greater the resistance, the lower the weight). For semi-hard connection, the switch state is first considered (the weight is 0 when the switch is open), and the action number is modified when the switch is closed (the more the number, the lower the weight). For soft connection, the signal-to-noise ratio is modified (the lower the signal-to-noise ratio, the lower the weight). The key parameters reflect the real-time state of the connection. The worse the state, the more the weight is down-regulated. The weight of the edge is valued through the above modifications.

[0120] Collect historical fault cases, use synthetic minority over-sampling technique, generate new synthetic samples in the feature space of existing minority (fault) samples through linear interpolation, collect real device state data from SCADA and PMS systems to form an initial training set, over-sample fault class samples until the data volume of each class is balanced, retain high-fidelity feature data, and form complete synthetic training data samples.

[0121] The above data samples are divided into training set and validation set, the actual health degree and actual fault propagation path of each device are labeled, and the model is trained. The adjacency matrix and node feature matrix are input into the GNN global propagation model. Through the graph convolution layer, each node fuses its own and adjacent device state information, and outputs the node health degree and fault propagation probability of the associated device.

[0122] In addition, the hierarchical adjustment module 300, according to the node health degree and the fault propagation probability, classifies the associated device abnormal high-risk anomaly, medium-risk anomaly and low-risk anomaly, and takes adjustment measures aiming at suppressing faults and reducing the impact on the normal operation of the power grid.

[0123] From the above training samples, the minimum value of the health degree of the fault device and the minimum value of the pre-fault propagation probability are counted as the preliminary threshold of high risk, and the range below the threshold is the high-risk range. The health degree and the propagation probability of the non-fault but hidden danger device are counted as the medium-risk threshold, and the range above the preliminary threshold and below the medium-risk threshold is the medium-risk range. The range above the medium-risk threshold is the low-risk range.

[0124] Among them, the hierarchical adjustment module 300 aims at minimizing the outage range by quickly isolating faults for high-risk anomalies, aims at preventing fault escalation and fine-tuning stability for medium-risk anomalies, and aims at minimizing intervention and maintaining normal operation for low-risk anomalies.

[0125] For high-risk anomalies, isolation and load shedding, standby operation, and power supply are avoided to prevent fault propagation. If the device has a standby (such as a double main transformer and a standby line), the standby device is first put into operation to carry the load, and then the high-risk device is shut down for maintenance. If there is no standby device, the device load is temporarily limited to reduce the probability of fault triggering, and emergency maintenance (within 24 hours) is arranged. Associated device protection: Adjust the protection threshold of the associated device in advance (such as shorten the circuit breaker trip delay) to prevent the fault from being conducted and expanded.

[0126] For medium-risk anomalies, intensive monitoring and preventive maintenance are adopted to eliminate hidden dangers in advance. The monitoring frequency is encrypted: the device real-time parameter collection interval is reduced from 15 minutes to 5 minutes, and key parameters such as temperature and current are tracked. Planned maintenance: Preventive maintenance (such as replacing aged parts of circuit breakers) is carried out during the low load period of the power grid (such as at night) to avoid affecting the power peak. Associated device warning: Mark the associated device in the SCADA system. If the parameter of the medium-risk device is abnormal, the warning prompt (not tripping) of the associated device is triggered immediately.

[0127] For low-risk anomalies, parameter optimization and planned maintenance are adopted to avoid excessive intervention. Optimize operating parameters: Fine-tune device operating thresholds to strengthen early warning. Include routine maintenance: Include device anomalies in the next quarter's planned maintenance list without the need for emergency treatment. Associated device adjustment exemption: Only record low-risk anomalies without changing the operating state and protection settings of the associated device.

[0128] Considering that the associated device will have an impact on the single-device threshold after taking adjustment measures, the hierarchical adjustment module 300 determines the direction of threshold adjustment in the single device according to the abnormality level of the associated device after taking adjustment measures, determines the threshold adjustment amplitude according to the node health degree and the fault propagation probability, and adjusts the threshold of the single device;

[0129] The higher the risk of the associated device, the more sensitive the local threshold of the single device needs to be (down-regulated, triggering an early warning earlier); the lower the risk, the more relaxed the threshold can be (up-regulated, reducing false positives), the higher the risk of the associated device, the threshold is significantly down-regulated, the medium risk is slightly down-regulated, and the low risk is appropriately up-regulated. The adjustment method is as follows:

[0130] First, determine the adjustment direction: if the associated device is a high-risk abnormality, it means that the environment of the device is easy to be affected by faults, and the local threshold needs to be significantly down-regulated to make the warning more sensitive to avoid false negatives; if the associated device is a medium-risk abnormality, the environment has some hidden dangers, and only the threshold needs to be slightly down-regulated to balance the sensitivity of the warning and the normal operation of the power grid; if the associated device is a low-risk abnormality, the environment risk is low, and the threshold can be appropriately up-regulated to reduce the interference of false positives on the power grid;

[0131] Then calculate the adjustment amplitude: in the high-risk scenario, the adjustment amplitude is obtained by weighted superposition of the reverse index of the node health degree and the fault propagation probability. First, take the reverse value of the node health degree (i.e., the index reflecting the degree of deterioration of the device state) as the basis, and calculate the first part of the amplitude according to the high weight of the node health degree; then take the fault propagation probability (i.e., the index reflecting the possibility of fault propagation to the single device) as the basis, and calculate the second part of the amplitude according to the high weight of the fault propagation probability; finally, add the two parts of the results to obtain the overall down-regulation amplitude. The worse the state of the device itself and the easier the fault is to conduct, the greater the down-regulation amplitude, ensuring that the threshold is sensitive enough;

[0132] Among them, the high weight of the node health degree and the high weight of the fault propagation probability are allocated with basic weights according to the influence proportion of the two indexes in the high-risk data sample on the fault result, for example, the driving effect of health degree deterioration on the fault is stronger, and the weight is set to 10%, and the weight of the propagation probability is set to 5% in proportion (the proportion of the two is about 2:1, which is consistent with the case influence proportion);

[0133] In the medium-risk scenario, the adjustment amplitude is obtained by weighted superposition based on the difference between the node health degree and the upper limit of the medium risk, and the difference between the fault propagation probability and the lower limit of the medium risk. First, calculate the difference between the node health degree and the highest critical value of the medium risk, and obtain the first part of the amplitude according to the lower weight of the node health degree; then calculate the part of the fault propagation probability exceeding the lowest critical value of the medium risk, and obtain the second part of the amplitude according to the lower weight of the fault propagation probability; the sum of the two parts is the overall down-regulation amplitude, which is smaller than that in the high-risk scenario, covering the potential risk and avoiding excessive adjustment to cause false positives;

[0134] Wherein, the lower weight of node health and the lower weight of fault propagation probability, through the influence proportion of the two indicators of medium risk in the statistical data sample on the fault result, allocate the basic weight according to the influence proportion, for example, because the influence amplitude of medium risk fault is only 1 / 2-2 / 3 of high risk, the health weight is reduced from 10% to 5%, and the propagation probability weight is reduced from 5% to 3% (maintaining the same influence proportion ratio as high risk);

[0135] In the low risk scenario, the adjustment amplitude is obtained by subtracting the weighted sum of the part of node health exceeding the low risk lower limit and the part of fault propagation probability below the low risk upper limit, first calculate the part of node health higher than the low risk minimum threshold, and obtain the first part amplitude according to the low weight of node health; Then calculate the second part amplitude according to the low weight of fault propagation probability, and finally subtract the second part result from the first part result to obtain the overall adjustment amplitude. The better the device state is, the more difficult the fault is to conduct, and the more reasonable the adjustment amplitude is, reducing the invalid false alarm in normal operation;

[0136] Wherein, the lower weight of node health and the lower weight of fault propagation probability, through the influence proportion of the two indicators of medium risk in the statistical data sample on the fault result, allocate the basic weight according to the influence proportion, for example, because the influence amplitude of medium risk fault is only 1 / 2-2 / 3 of high risk, the health weight is reduced from 10% to 5%, and the propagation probability weight is reduced from 5% to 3% (maintaining the same influence proportion ratio as high risk);

[0137] Finally, verify the adjustment effect: after adjustment, observe the single device warning situation, if there is a missed judgment problem that the actual exception does not trigger the warning, further reduce the threshold; If there are many false alarms in normal state, slightly increase the amplitude, and dynamically iterate and optimize the threshold every day according to the latest health and fault propagation probability data of the associated device, to ensure that it always adapts to the current risk;

[0138] Considering that frequent adjustment of the threshold may lead to frequent false alarms, which will affect the stability of the system, therefore, set the minimum adjustment interval to limit the adjustment frequency, according to the characteristics of real-time data collection (time granularity 15 minutes) of SCADA system, the minimum interval of single device threshold adjustment is greater than 1 hour, and the adjustment is triggered only when the risk level of associated device changes across levels (such as from medium risk to high risk, from low risk to medium risk); If the risk level only fluctuates within the same level, do not trigger adjustment to avoid adjusting the threshold due to slight fluctuations;

[0139] In the filtering and denoising preprocessing of the data by the difference analysis module 200, it is required that when adjusting the threshold, the real-time instantaneous value of the associated device is not used, but the average value of the node health degree and fault propagation probability within 15 minutes is used to ensure stable risk state adjustment and exclude false risk signals caused by sensor instantaneous errors and short-term fluctuations of the power grid.

[0140] For example, in a certain substation, there is a hard connection relationship between the core associated device and the single device: when the associated device enters a high-risk state due to overload, the original local threshold of the single device is not adjusted in time, even if the operating parameters of the single device have approached the safety threshold, the early warning is not triggered, and there is a risk of missing the fault; when the associated device returns to a low-risk state, the threshold of the single device is not relaxed, and the frequent fluctuations of the parameters in the normal operation frequently trigger invalid early warnings, increasing the burden of the operation and maintenance personnel; and when the associated device fluctuates slightly within the same risk level, the system still frequently adjusts the threshold of the single device, causing the early warning state to switch repeatedly, affecting the stability of the power grid operation.

[0141] According to the risk level of the associated device, the adjustment direction of the local threshold of the single device is determined. If the associated device is in a high-risk abnormality, it indicates that the environment of the single device is easily affected by the fault, and the local threshold needs to be significantly lowered to improve the early warning sensitivity and avoid missing the fault. If the associated device is in a medium-risk abnormality, the environment has some hidden dangers, and only the threshold is slightly lowered to balance the early warning sensitivity and the normal operation of the power grid. If the associated device is in a low-risk abnormality, the environmental risk is low, and the threshold is appropriately raised to reduce the interference of false alarms on the power grid. The adjustment amplitude is calculated by combining the two indicators of the node health degree and the fault propagation probability of the associated device. When the associated device is in a high-risk abnormality, the adjustment amplitude is determined by weighting and superimposing the reverse indicator of the node health degree and the fault propagation probability. The lower the health degree of the associated device and the higher the fault propagation probability, the greater the adjustment amplitude. When the associated device is in a medium-risk abnormality, the slight adjustment amplitude is calculated based on the difference between the medium interval of the node health degree and the fault propagation probability with a lower weight to avoid excessive adjustment. When the associated device is in a low-risk abnormality, the reasonable adjustment amplitude is determined by weighting the part of the node health degree exceeding the low-risk threshold and the part of the fault propagation probability below the low-risk threshold. The higher the health degree of the associated device and the lower the fault propagation probability, the more adaptive the adjustment amplitude to the actual demand.

[0142] After adjusting the threshold, the early warning of the single device is observed. If there is a missing problem of not triggering early warning in actual abnormality, the threshold is further lowered. If the early warning is frequently triggered in the normal operation state, the adjustment amplitude is slightly adjusted. The threshold is dynamically iteratively optimized according to the latest health degree and fault propagation probability data of the associated device every day to ensure that the threshold is always adapted to the current risk state.

[0143] Through the above steps, the single-device early warning is sensitive when the associated device is at high risk, and there is no fault missed judgment; when the associated device is at low risk, the early warning is loose, there is no frequent false alarm, the single-device abnormality determination accuracy is significantly improved, the adjustment frequency is greatly reduced, the system oscillation caused by frequent threshold changes is avoided, the number of false alarms is reduced, and the power grid operation is not disturbed by invalid early warnings.

[0144] As shown in Figure 3 A substation monitoring and maintenance method, characterized in that it comprises the following steps:

[0145] S1, collecting electrical connection data and parameter association data, and quantifying the fault influence range of the device to obtain a comprehensive score, comparing the comprehensive score with a preset threshold, and dividing the device type into single-device type and associated device type;

[0146] S2, using a local threshold diagnosis model to monitor the state of the single device to determine the abnormality of the single device, and using a GNN global propagation model to monitor the state of the associated device, taking the node feature matrix and the adjacency matrix as input, using the graph convolution layer to aggregate information, and outputting the node health degree and the fault propagation probability;

[0147] S3, according to the node health degree and the fault propagation probability, the associated device is divided into high-risk abnormality, medium-risk abnormality and low-risk abnormality, and aiming at the abnormality level, taking reducing the influence on the normal operation of the power grid while suppressing the fault as the goal, and taking adjustment measures.

[0148] In summary, the working principle of the present scheme is as follows:

[0149] The substation monitoring and maintenance system and method, the quantification classification module 100 classifies the devices in the substation based on electrical connection and operation parameter quantification, avoids artificial subjective bias, sets the initial threshold value combined with historical fault data, and dynamically adjusts the threshold value according to the device load rate, adapts to different operation scenarios of the substation, and ensures that the classification fits the actual fault influence of the device;

[0150] The difference analysis module 200 introduces correction parameters such as aging and load loss for single-device monitoring, combines multi-level threshold values, reduces the misjudgment and missed judgment caused by changes in the state of the device, uses a GNN model for modeling of associated devices, aggregates neighbor states, outputs device health degree, and quantifies fault propagation probability, early warning of cascading failure risk, realizes full-coverage accurate monitoring, and takes into account local and global risks;

[0151] The hierarchical adjustment module 300 aims at high-risk rapid isolation of faults to protect core power supply, fine-tuning to prevent escalation for medium-risk, and reducing intervention for low-risk, avoiding excessive operation disturbance to the power grid, while taking risk response measures, adjusting the threshold value of the single device, so that the single-device monitoring and the risk state of the associated device are adaptively matched.

[0152] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A monitoring and maintenance system of a substation, characterized by, The system comprises: A quantitative classification module (100) collects electrical connection data and parameter correlation data, calculates the electrical connection degree and data correlation degree of the equipment, quantifies the fault influence range of the equipment, obtains a comprehensive score, compares the comprehensive score with a preset threshold, and divides the equipment type into a single equipment type and an associated equipment type; The difference analysis module (200) adopts a local threshold contrast method to perform state monitoring on a single device, acquires function parameters of the single device in real time, compares the function parameters with rated parameters of the single device, and determines abnormal conditions of the single device; the rated parameters are corrected based on a cumulative running time length of the single device, a load loss coefficient is calculated according to a cumulative full load time length proportion, and a final correction parameter is determined: the load loss coefficient is calculated . If the current, power load related threshold is corrected, the final correction parameter = K 损 × aging coefficient; if only the load loss related threshold is corrected, directly take K 损 as the final correction parameter; A GNN global propagation model is used for state monitoring of the associated equipment, the associated equipment and its connection relationship are modeled as a graph structure, a node feature matrix and an adjacency matrix are used as input, information aggregation is performed by using a graph convolution layer, and a node health degree and a fault propagation probability are output; A hierarchical adjustment module (300) divides the abnormality of the associated equipment into high-risk abnormality, medium-risk abnormality and low-risk abnormality according to the node health degree and the fault propagation probability, and takes adjustment measures aiming at the abnormality level and the goal of suppressing faults while reducing the impact on the normal operation of the power grid.

2. The monitoring maintenance system of a substation according to claim 1, characterized in that: The quantitative classification module (100) comprises: A collection calculation unit (110) extracts the static electrical connection relationship and the dynamic switch state between the equipment from the design drawings of the substation, calculates the electrical connection degree by counting the number of directly and indirectly affected equipment and the connection strength through the electrical connection, collects the equipment operating parameters from sensors and intelligent instruments, and selects strong correlation parameter pairs, and calculates the data correlation degree by counting the number of strong correlation parameter pairs; A quantitative comparison unit (120) quantifies the fault influence range of the equipment by weighted summation, calculates the comprehensive score of the equipment, compares the comprehensive score of the equipment with a preset threshold, and if the comprehensive score is lower than the preset threshold, it is a single equipment type, and if the comprehensive score exceeds the preset threshold, it is an associated equipment type.

3. The monitoring maintenance system of a substation according to claim 2, characterized in that: The quantitative comparison unit (120) collects historical operation data of the substation equipment, calculates the fault influence range of each equipment, and draws a distribution histogram of equipment fault influence by using a data statistical analysis method, and takes the inflection point of the distribution curve as the initial preset threshold.

4. The monitoring maintenance system of a substation according to claim 3, characterized in that: The quantitative comparison unit (120) collects the actual operating load of the equipment in the substation in real time, calculates the load rate of the equipment, classifies the operating conditions of the equipment, determines the adjustment range of the initial threshold according to the amplitude of the equipment fault influence of the operating condition level, and establishes a rule mapping table of operating conditions and thresholds.

5. The monitoring maintenance system of a substation according to claim 1, characterized by: The difference analysis module (200) comprises: A single equipment monitoring unit (210) determines a correction parameter according to the cumulative running time of the single equipment, corrects the rated parameter, compares the real-time functional parameter of the single equipment with the corrected rated parameter, outputs the abnormality of the single equipment, and sends a warning signal; An associated equipment monitoring unit (220) takes each associated equipment as a node, takes the electrical connection relationship between the equipment as an edge and assigns a weight, and then converts the node and the edge into an adjacency matrix, inputs the adjacency matrix and the node feature matrix into the GNN global propagation model, and fuses the state information of each node and its neighbor equipment through the graph convolution layer, and outputs the node health degree and the fault propagation probability of the associated equipment.

6. The monitoring maintenance system of a substation according to claim 5, characterized in that: The single-device monitoring unit (210) acquires the cumulative running time of the single device, calculates a load loss coefficient according to the proportion of the cumulative full-load time, and determines the final correction parameter.

7. The monitoring maintenance system of a substation according to claim 5, characterized by: The associated device monitoring unit (220) divides the connection into hard connection, semi-hard connection and soft connection according to the functional attribute of electrical connection, determines the basic weight of the edge between devices according to the connection type, and corrects the basic weight through the key parameters of different connections.

8. The monitoring maintenance system of a substation according to claim 1, characterized by: The hierarchical adjustment module (300) aims to minimize the outage range for high-risk abnormalities, to prevent fault escalation for medium-risk abnormalities, and to minimize intervention for low-risk abnormalities.

9. The monitoring maintenance system of a substation according to claim 8, characterized in that: The hierarchical adjustment module (300) determines the adjustment direction of the threshold in the single device according to the abnormal level of the associated device after taking the adjustment measures, determines the threshold adjustment amplitude according to the node health degree and the fault propagation probability, and adjusts the threshold of the single device.

10. A method of monitoring maintenance of a substation, characterized by, The method comprises the following steps: S1, collecting electrical connection data and parameter association data, and quantifying the fault influence range of the device to obtain a comprehensive score, comparing the comprehensive score with a preset threshold, and dividing the device type into single device type and associated device type; S2, using threshold comparison method to monitor the state of single device, by real-time acquisition of single device function parameters, and compared with the rated parameters of single device, to determine the abnormal situation of single device, using GNN global propagation model to monitor the state of associated device, taking node feature matrix and adjacency matrix as input, using graph convolution layer to aggregate information, outputting node health degree and fault propagation probability; S3, according to the node health degree and the fault propagation probability, the associated device abnormality is divided into high-risk abnormality, medium-risk abnormality and low-risk abnormality, according to the abnormal level, taking the target of inhibiting the fault and reducing the influence on the normal operation of power grid, taking the adjustment measures.

Images