Intelligent power monitoring method and system based on power big data

Abstract

The invention relates to the technical field of electric power system monitoring and intelligent scheduling, and discloses an electric power intelligent monitoring method and system based on electric power big data, and the method comprises the steps: calibrating a physical degradation model through historical asset big data offline to generate a simple calibration function, and calling the function online to calculate a dynamic safety limit reflecting chronic degradation, meanwhile, high-frequency real-time physical measurement data are obtained in parallel and are subjected to cross comparison with the physical reference model to generate an authentication signal reflecting acute abnormality; according to the method, a technical path for connecting asset health with the scheduling procedure is constructed, a dual-guarantee mechanism of chronic degradation monitoring and acute anomaly authentication is established, the security logic gating is executed, and the scheduling procedure is updated by using the dynamic security limit only when the authentication signal does not exceed the threshold value, otherwise, the updating is stopped and the rollback is carried out. And the adaptability and robustness of the power grid safety guarantee are improved.

Description

Technical Field

[0001] This invention relates to a power intelligent monitoring method and system based on power big data, belonging to the field of power system monitoring and intelligent dispatching technology. Background Technology

[0002] Currently, in large-scale power grid security and intelligent dispatch systems, the core foundation of operation lies in operating procedures built upon corresponding safety criteria. These procedures provide dispatchers with rigid physical constraints, such as specifying clear static transmission limits for critical transmission lines or main transformers. Dispatchers use these limits to optimize dispatching and power flow calculations to maintain the safety and stability of the entire power system. The effectiveness of this static limit-dependent operating mode depends on the premise that the health status of physical assets remains consistent with their factory design ratings. However, in long-term engineering practice, the actual physical health status of energy-carrying physical entities, especially 750kV-level main transformers or reactors, such as insulation performance and thermal characteristics, will dynamically and progressively deteriorate with changes in the operating environment and equipment aging. This creates an inconsistency between the operating procedures upon which dispatchers base their decisions and the dynamic health of the assets. The dispatch system is actually dispatching an asset whose physical state has deteriorated based on potentially outdated static data, which introduces uncontrollable factors into the safe operation of the power grid.

[0003] To address this issue, a common approach is to leverage massive amounts of asset big data, such as monitoring data from dissolved gases in oil or partial discharges, and analyze this data using general data processing platforms or machine learning algorithms to output an equipment health score or comprehensive risk probability. However, this approach creates a mismatch in information types when applied to power grid dispatching. Dispatchers and their intelligent dispatching systems, such as power flow optimization algorithms, cannot directly apply an abstract, non-physical dimension score or percentage. Dispatching systems require execution commands with clear physical dimensions, such as amperes, megawatts, and volts. Monitoring results with uninterpretable internal logic are ineffective for safety assurance and intelligent dispatching systems. Meanwhile, other technological approaches focus on the security management and governance of power big data, but these also fail to bridge the information gap at the physical application level. For example, Chinese invention patent CN120277687A discloses a method and system for power big data security management based on artificial intelligence. This solution uses artificial intelligence to classify, encrypt, and control access to data. Its technical purpose is to ensure the storage security and compliance of the data itself and prevent data leakage or tampering. However, this method is essentially still within the scope of information technology (IT) management of data. It solves the problem of how to store data securely, but fails to address the core technical problem that this invention aims to solve, namely, how to translate and convert the physical health information contained in these data (regardless of their security level) into dynamic operation instructions with clear physical dimensions that can be directly adopted and executed by the dispatching system (OT). This type of method also suffers from the limitation of the aforementioned mismatch of information types and cannot open up the technical path between asset big data and online dispatching procedures.

[0004] Therefore, the technical problem to be solved by this invention is how to provide a new monitoring method that, within the framework of power grid security and dispatch, transforms the physical health information contained in massive asset big data into dynamic operating procedures with clear physical dimensions that can be directly executed by the dispatch system, thereby realizing information exchange between operating data and asset data. Summary of the Invention

[0005] This invention provides a power intelligent monitoring method and system based on power big data. Its main purpose is to solve the problems in the prior art, such as the inability of static operating procedures to reflect the dynamic physical health status of assets, and the inability of the abstract scoring results provided by data analysis schemes to be directly applied by the dispatching system, resulting in a lack of information exchange between asset big data and online dispatching decisions.

[0006] To achieve the above objectives, this invention provides a power intelligent monitoring method based on power big data, applied to a large-scale power grid security and intelligent dispatch system. The method includes: Step 101: On the offline end, aggregate historical asset big data of physical assets in the power grid, retrieve the physical degradation model, and use the historical asset big data to calibrate the parameters of the physical degradation model. Step 102: Based on the calibrated physical degradation model, a simplified calibration function is generated through offline calculation, dimensionality reduction and encapsulation. The input of the simplified calibration function is set to at least one simplified operating parameter that the intelligent scheduling system can obtain in real time, and the output of the simplified calibration function is defined as the dynamic security limit of the physical asset. Step 103: In the online monitoring module, load the simplified calibration function, obtain the simplified operating parameters in real time, and call the simplified calibration function to calculate the dynamic safety limit; Step 104: Obtain high-frequency real-time physical measurement data of the physical assets and retrieve a physical reference model of the physical assets; Step 105: Instantly cross-compare the real-time physical measurement data with the expected output of the physical reference model under the current operating conditions to generate an identification signal characterizing acute physical anomalies. Step 106: Establish and execute a security logic gating rule, wherein when the value of the authentication signal does not exceed the preset security threshold, the static security limit procedure for physical assets in the intelligent scheduling system or security assurance system is automatically updated to the dynamic security limit calculated in step 103; when the value of the authentication signal exceeds the security threshold, the automatic update is prevented, and the operation procedure of the intelligent scheduling system or security assurance system is automatically rolled back to a preset conservative limit.

[0007] Preferably, the historical asset big data in step 101 includes at least one of the following: dissolved gas data in oil, partial discharge data, infrared thermal imaging data, historical load curves and maintenance records; and the physical degradation model in step 101 includes at least one of the following: a hot spot temperature model of equipment based on thermodynamic principles and an aging model of insulation materials based on electrochemical principles.

[0008] Preferably, the simplified calibration function in step 102 is encapsulated as at least one of a lookup table, a polynomial function, and a piecewise linear function; and the simplified operating parameters in step 103 include at least one of ambient temperature and real-time load current.

[0009] Preferably, the high-frequency real-time physical measurement data in step 104 is phasor measurement unit data, which includes voltage phasors and current phasors; and the physical reference model in step 104 is the T-type equivalent circuit impedance model of the physical asset.

[0010] Preferably, the dynamic safety limit updated in step 106 is used as a rigid physical constraint for the operation of at least one of the optimized power flow algorithm and the safety and stability control device.

[0011] Preferably, after step 103 and before step 106, the method further includes: step 601: using the dynamic safety limit calculated in step 103 as a virtual constraint; step 602: running a simulation based on the virtual constraint in the sandbox pre-simulation mode of at least one of the economic scheduling module and the safety verification module to evaluate the economic impact index of applying the dynamic safety limit; step 603: determining whether the economic impact index exceeds a preset economic threshold; and step 106 is modified to: automatically updating only when the value of the authentication signal does not exceed the safety threshold and the economic impact index does not exceed the economic threshold; performing a rollback when the value of the authentication signal exceeds the safety threshold; and preventing automatic updating when the value of the authentication signal does not exceed the safety threshold but the economic impact index exceeds the economic threshold, and instead outputting a collaborative decision-making auxiliary information containing the dynamic safety limit and the economic impact index to the intelligent scheduling system.

[0012] Preferably, after step 102, the method further includes: step 701: on the offline end, independently analyze all historical operational big data of the physical asset to construct a statistical security envelope that characterizes its historical experience security boundary; and in step 106, the automatic update rule is modified to: when the value of the authentication signal does not exceed the security threshold, compare the dynamic security limit with the limit of the statistical security envelope under the current operating conditions, take the smaller value of the two as the final dynamic security limit, and use the final dynamic security limit to perform automatic update.

[0013] Preferably, after step 102, the method further includes: step 801: reusing the calibrated physical degradation model generated in step 101; step 802: obtaining at least one of the typical future load curve and expected operating mode of the physical asset; step 803: performing a simulation based on the calibrated physical degradation model and the typical future load curve to calculate a remaining usable life index characterizing the maintenance cycle, wherein the remaining usable life index is defined as the virtual time corresponding to when the virtual health state of the physical asset deteriorates to a preset maintenance threshold during the simulation; step 804: inputting the remaining usable life index... Before step 103, the method, originating from at least one of the asset management and maintenance decision systems, further includes: step 901: generating an additional function characterizing the relationship between steady-state load and temperature rise based on the asset thermodynamic model contained in the calibrated physical degradation model; step 902: acquiring the real-time load and real-time temperature of the physical asset during online operation; step 903: substituting the real-time load into the function characterizing the relationship between steady-state load and temperature rise to calculate the theoretical temperature rise value; step 904: subtracting the theoretical temperature rise value from the real-time temperature to reverse-calculate an equivalent heat dissipation temperature; and in step 103, the simplified operating parameters are replaced with the equivalent heat dissipation temperature.

[0014] Preferably, the equivalent heat dissipation temperature calculated in reverse in step 904 Its calculation rules follow: ,in, The real-time temperature obtained in step 902. The theoretical temperature rise value calculated in step 903 is based on the real-time load and a function characterizing the relationship between steady-state load and temperature rise.

[0015] A smart power monitoring system based on big data of electricity, the system includes: An offline calibration module is configured to: aggregate historical asset big data of physical assets in the power grid; retrieve physical degradation models; calibrate the parameters of the physical degradation models using historical asset big data; and, based on the calibrated physical degradation models, reduce dimensions, encapsulate, and generate a simplified calibration function through offline calculation. An online monitoring module is configured to: load a simplified calibration function; obtain at least one simplified operating parameter corresponding to the input of the simplified calibration function in real time; and call the simplified calibration function to calculate the dynamic security limit of the physical asset. A real-time authentication module is configured to: acquire high-frequency real-time physical measurement data of physical assets; retrieve a physical reference model of physical assets; and instantaneously cross-compare the high-frequency real-time physical measurement data with the expected output of the physical reference model under the current operating conditions to generate an authentication signal characterizing an acute physical anomaly. A security arbitration module is configured to: establish and execute a security logic gating rule, wherein when the value of the authentication signal does not exceed a preset security threshold, the static security limit procedure for physical assets in the intelligent scheduling system or security assurance system is automatically updated to a dynamic security limit; otherwise, when the value of the authentication signal exceeds the security threshold, the automatic update is prevented, and the operating procedure of the intelligent scheduling system or security assurance system is automatically rolled back to a preset conservative limit.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This method provides an operational approach that connects the offline health status of physical assets with online scheduling procedures. First, at the offline end, historical asset data is used to calibrate the parameters of a recognized physical degradation model. This reflects the complex physical relationships of a specific device's health and is pre-encapsulated into a simplified calibration function that can be called by the online system. During online monitoring, the scheduling system only needs to use the existing simplified operating parameters and call this function to instantly calculate the dynamic safety limit corresponding to the asset. This limit value has a clear physical dimension and can be directly used by the intelligent scheduling system or security assurance system as a rigid physical constraint condition for its operation. Thus, the original static safety procedure based on design values ​​is transformed into an executable online procedure that can dynamically reflect the true physical health status of the asset.

[0017] 2. By introducing a parallel real-time physical verification pathway, the limitations of the basic solution in terms of time scale response are compensated. The dynamic security limit generated by the basic solution is based on the chronic deterioration state of the physical asset. The verification mechanism in this method uses high-frequency real-time physical measurement data to cross-compare with a simplified physical reference model to identify acute physical anomalies that cannot be perceived by chronic data. This logical gating method, which combines the chronic health limit generated based on historical data with the rejection of acute anomalies based on real-time phasor verification, constructs a double insurance system. This ensures that the final security procedures not only reflect the long-term health trend of the asset but also pass the test of the current physical reality.

[0018] 3. This method coordinates the security assessment of physical assets with the economic dispatch of power grid operation. After calculating the dynamic security limit, this method does not immediately implement it as a rigid constraint. Instead, it utilizes existing economic dispatch or security verification modules in the dispatch system to run a simulation in sandbox pre-simulation mode to evaluate the economic impact of applying the limit. Based on this pre-simulation result, this method can perform hierarchical arbitration of security and economy: for adjustments with economic impact within an acceptable range, automatic updates are performed; for adjustments that may cause significant economic costs, automatic execution is prevented, and instead, a collaborative decision-making aid information including security boundaries and economic costs is provided to the dispatcher. This transforms dispatch decision-making from a single physical security execution to a security-economy approach. The method provides collaborative intelligent decision support and establishes an independent post-arbitration mechanism that combines physical model derivation with historical experience induction. While calculating the theoretical dynamic security limit using the calibrated physical model, it independently analyzes all historical operational big data of the physical asset to construct a statistical security envelope that characterizes its empirical security boundary. Before the final update, the method compares these two limit values ​​from different sources and with independent principles, and takes the smaller value as the final dynamic security limit for execution. This approach ensures that when the scheme optimizes the operating space using the physical model, its output is always constrained by the historically verified security boundary, thereby avoiding the risk of artificially inflated levels that may be caused by model defects or offline data contamination. Attached Figure Description

[0019] Figure 1 This is a flowchart of the monitoring method for economic evaluation and collaborative safety arbitration of the present invention; Figure 2 This is a comparison of the dynamic safety limit and static limit of the present invention, as well as a diagram showing the load-limiting effect. Figure 3 This is a diagram showing the collaborative relationship between the four core functional modules of the monitoring method of the present invention; Figure 4 This is an interactive timing diagram of the acute verification and automatic update procedure of this invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments described herein are only for explaining the present invention and are not intended to limit the scope of protection of the present invention.

[0021] This invention discloses a power intelligent monitoring method and system based on power big data. This system is applied to a large-scale power grid security and intelligent dispatching system of 750 kV and above. The system architecture includes an offline calibration module, an online monitoring module, a real-time authentication module, and a security arbitration module. The core task of the offline calibration module is to perform deep calibration of the physical model using massive historical data in a non-real-time environment, and to encapsulate the complex physical relationships it contains into a simplified calibration function. The task of the online monitoring module is to load this simplified calibration function in a real-time dispatching environment (such as an EMS system) and instantly calculate the dynamic safety limit reflecting the chronic health status of assets using very few real-time operating parameters. The real-time authentication module runs in parallel, using high-frequency real-time measurement data to cross-compare with a fixed physical reference model to identify acute physical anomalies that cannot be perceived by the chronic model. The security arbitration module, as the final decision-making and execution unit, establishes a security logic gating rule to collaboratively arbitrate the outputs of the above two channels. Only when both the chronic health and acute safety dimensions pass verification is the use of the dynamic safety limit to update the dispatching procedure approved; otherwise, the update and execution will be prevented. A conservative rollback strategy is adopted to construct a dual guarantee mechanism connecting asset health and scheduling procedures. In a specific implementation scenario, this solution is applied to a 750 kV main transformer, which is a critical physical asset in a large-scale power grid. First, at the offline end, the offline calibration module performs step 101, which gathers all historical asset big data of the transformer within 5 years since its commissioning. This data may include: dissolved gas data (DGA), which reflects whether there is overheating or discharge faults inside the transformer; partial discharge data (PD), which reflects the state of the insulation system; and infrared thermal imaging data and historical load curves, which reflect the thermal characteristics and load history of the equipment. At the same time, the module retrieves the field-recognized physical degradation model for this type of transformer, which may include: equipment hot spot temperature model based on thermodynamic principles, and insulation material aging model based on electrochemical principles, such as the Arrhenius model. Subsequently, the offline calibration module uses historical asset big data to calibrate the parameters of the physical degradation model. This calibration process is not a black-box training, but a calibration with clear physical meaning. For example, analysis of DGA data reveals acetylene ( The content of [unspecified substance] shows a continuous slight increase during specific load periods, indicating the presence of chronic high-temperature overheating. The system then uses this data to adjust the equivalent heat dissipation coefficient, a key parameter in the hotspot temperature model. The value was corrected and calibrated from the factory design value of 1.0 to 0.92. The parameters materialize the chronic deterioration state of a specific asset, characterized by a slight decrease in heat dissipation efficiency due to long-term operation.

[0022] Step 101 uses historical asset big data to analyze physical degradation model parameters, such as the equivalent heat dissipation coefficient in the hotspot temperature model. The calibration procedure is specifically configured as an iterative optimization process: First, the physical degradation model is retrieved and its factory design values ​​are used, such as... First, the model takes historical operating data from the historical asset big data, such as historical load curves and historical ambient temperature data, as input and performs a historical backtracking simulation to calculate theoretical health indicators, such as theoretical hotspot temperature curves or theoretical DGA gas production rate curves, under these initial parameter values. Second, it compares these theoretical health indicators with actual measurement results from the historical asset big data, such as actual DGA data or infrared thermal imaging data, point by point, and establishes a cost function, for example, defined as the root mean square error (RMSE) between the theoretical and actual values. Finally, it uses a numerical optimization algorithm, such as gradient descent or simplex method, to automatically iterate and adjust the parameters offline with the goal of minimizing this cost function. The value of is determined when the cost function converges to a preset minimum or parameter. When the value stabilizes, the iteration stops, and the result is... Value, such as This refers to the final calibrated parameter that is determined to materialize the specific chronic degradation state of the asset; then, the offline calibration module performs step 102, which is based on this... The physical degradation model, after parameter calibration, is then dimensionality-reduced and encapsulated through offline computation. This process involves executing tens of thousands of simulations offline, with the simulation input consisting of simplified operating parameters that the intelligent scheduling system can obtain in real time, such as ambient temperature. Real-time load current Its simulation output is the result under the input condition based on The calibration model ensures that the hotspot temperature does not exceed the limit (e.g., 140°C). The dynamic security limits of physical assets ( ) Finally, the system will process these ( , ) The massive simulation results are fitted, encapsulated, and a simplified calibration function is generated. To ensure absolute simplicity in online calls, this function can be encapsulated as a lookup table that is easy for the scheduling system to call, or a simple polynomial function. Subsequently, during the online operation phase of the system, the online monitoring module executes step 103. This module is deployed at the front end of the intelligent scheduling system (EMS) or safety assurance system (SCS). It first loads the simplified calibration function, i.e., the lookup table. This module obtains simplified operating parameters in real time through the SCADA system. At a specific moment, it obtains that the ambient temperature is 35°C. The real-time load current is The online monitoring module immediately invokes the simplified calibration function (table lookup) to instantly calculate the asset's value at 35. Under the circumstances, considered After reaching a state of chronic degradation, its dynamic safety limit is This value is below its nameplate static limit. .

[0023] Meanwhile, to prevent acute physical anomalies that the LMF model cannot detect, the real-time authentication module executes steps 104 and 105 in parallel. In step 104, the module acquires high-frequency real-time physical measurement data of the physical asset. In a preferred embodiment, this data is data from phasor measurement units (PMUs) deployed on both the high and low voltage sides of the transformer. The phasor measurement unit data provides high-precision voltage and current phasors at a rate of 50 times per second. Simultaneously, the module retrieves a physical reference model of the physical asset. This model is preferably a T-type equivalent circuit impedance model of the physical asset. The impedance parameters of this model are calibrated at the factory and are regarded as the physical benchmark for the healthy state. In step 105, the real-time authentication module performs an instantaneous cross-comparison between the high-frequency real-time physical measurement data (PMU phasors) and the expected output of the physical reference model (T-type model) under the current operating conditions. The specific implementation procedure is as follows: First, using the real-time voltage and current phasors measured by the PMU, the actual equivalent impedance of the asset is directly calculated using Ohm's law. The second step involves taking the operating conditions (such as voltage and load) measured by the PMU as input, substituting them into a fixed T-type equivalent circuit impedance model, and calculating the theoretical equivalent impedance. The third step is to generate a forensic signal characterizing the acute physical abnormality, which is defined as the relative deviation between the two, i.e. Finally, the security arbitration module executes step 106, which establishes and executes a security logic gating rule that relies on a preset security threshold, for example... The preset safety threshold used for acute anomaly identification in step 106; The specific calibration procedure is configured as follows: First, on the offline end, high-frequency real-time physical measurement data samples of the physical asset are screened and retrieved from historical asset big data for all periods in history when the physical asset was confirmed to be in a healthy operating state (i.e., without any acute failure records), such as retrieving 1000 hours of healthy PMU data; Second, the batch of healthy sample data is processed one by one through the instantaneous cross-comparison procedure in step 105, that is, the authentication signal is calculated. The values ​​are used to generate a massive histogram of the statistical distribution of the authentication signal, representing only normal measurement noise and the inherent simplified error of the physical reference model; finally, to achieve high detection sensitivity while ensuring an extremely low false alarm rate, this safety threshold is... It is set as a preset high quantile of the statistical distribution of this health status, for example, by taking the mean of the distribution plus three times the standard deviation. The value corresponding to ) or directly the value corresponding to its 99.9th percentile, such as 3%, is used. Furthermore, this threshold needs to be backtested and verified using known historical acute fault data samples (such as data on inter-turn short circuits or cooling faults that have occurred) before deployment to confirm that the threshold can effectively distinguish between fault states and healthy states. This threshold, calibrated offline, represents the normal deviation range allowed by the PMU measurement error and the T-model simplification. The execution logic of this rule is divided into two typical operating conditions: Operating Condition 1 (chronic degradation, acute normal): in Moment, Authentication Signal The value did not exceed the preset 3% safety threshold, indicating that the 750kV transformer has not experienced any acute physical anomalies such as inter-turn short circuits or sudden cooling system failures. At this point, the safety arbitration module determines that the acute verification has passed and approves the use of the calculation results from the online monitoring module, applying the static safety limit regulations for the transformer in the intelligent dispatch system. ), automatically updated to the dynamic safety limit calculated in step 103 ( Condition 2 (Acute Abnormal Sudden Onset): In At a certain moment, a cooling fan on the transformer suddenly stopped, causing a sudden surge in local temperature and resulting in slight deformation of the windings. The PMU detected this change and... Transient changes make the authentication signal Mutation Its value exceeded the 3% safety threshold; at this point, although the online monitoring module was based on DGA data that had not yet been updated, it still calculated an optimistic dynamic safety limit ( However, if the security arbitration module rejects the authentication due to an acute failure, it will immediately prevent the automatic update from being executed and automatically revert the operating procedures of the intelligent scheduling system to a preset conservative limit, such as reverting to 70% of the nameplate rating (i.e., At the same time, an alarm is sent to the dispatch center to report an acute physical state mismatch with the model.

[0024] The dynamic safety limit successfully updated in operating condition one will serve as a rigid physical constraint when the scheduling system executes the Optimized Power Flow (OPF) algorithm or initiates the Safety and Stability Control (SCS) device, ensuring that scheduling decisions are always anchored within the true physical health boundary of the asset. Based on the above technical solution, this invention may further include optimized configurations to coordinate economic scheduling objectives with safety boundary constraints. In a preferred configuration, the dynamic safety limit is calculated in step 103. Following this, and before performing the update in step 106, the system further includes the following steps: Step 601, [the system will then...] The dynamic safety limit is used as a virtual constraint; in step 602, a simulation is run in the sandbox study mode of the economic dispatch module (OPF) based on the virtual constraint to evaluate the economic impact index brought about by applying this limit. For example, it is calculated that in order to avoid this constraint, the system needs to start an additional standby unit, resulting in an increase of 300,000 yuan / hour in the total power generation cost of the entire network; in step 603, it is determined whether the economic impact index exceeds the preset economic threshold, for example, the threshold is set by the dispatch procedure to 200,000 yuan / hour; accordingly, the safety logic gating rule in step 106 is modified to: when the authentication signal does not exceed the safety threshold (such as Automatic updates are only performed when the economic impact indicators do not exceed the economic threshold (e.g., 150,000 yuan / hour < 200,000 yuan / hour); automatic updates are only performed when the authentication signal exceeds the safety threshold (e.g., ...). Safety always takes precedence, and a rollback should be executed immediately; however, if the authentication signal does not exceed the safety threshold ( However, when the economic impact indicators exceed the economic threshold ( If the automatic update is blocked, the system will instead output a collaborative decision-making support information containing dynamic safety limits and economic impact indicators to the intelligent dispatching system console. This information can be expressed as: "The dynamic limit of asset [750kV main transformer A] is recommended to be lowered to 1900A (safe), with an estimated economic impact of high (300,000 yuan / hour). Please have the dispatcher manually confirm the execution." This transforms rigid safety execution into intelligent decision support that integrates safety and economics. To further enhance the robustness of the system's decision-making and prevent the LMF from calculating artificially high dynamic safety limits due to defects in the physical degradation model or offline data contamination, in another preferred configuration, after generating the simplified calibration function in step 102, the method further includes: Step 701, on the offline end, independently and without relying on any physical model, analyzing all historical operational big data of the physical asset (such as 5 years of SCADA data) using statistical methods to construct a statistical safety envelope characterizing its historical experience safety boundary. This envelope can be constructed as a quantile lookup table, for example... 99.9th percentile of historical maximum load This value represents an empirical boundary that has been verified to operate safely in the past; accordingly, in step 106, the automatic update rule is modified to: when the value of the authentication signal does not exceed the safety threshold ( The dynamic safety limit calculated in step 103 ( ) and the limit of statistical safety envelope under current operating conditions ( Compare the two values ​​and take the smaller one. and using the final dynamic safety limit ( Automatic updates will be performed; in another case, if the LMF calculates an artificially high value due to model defects. The final updated value is .

[0025] To fully utilize the information byproducts generated during the offline calibration phase, namely the calibrated physical degradation model ( This invention can also extend the system's integrated scheduling and maintenance decision-making capabilities in a single configuration. Following step 102, the method further includes: step 801, reusing the calibrated physical degradation model generated in step 101; step 802, obtaining typical future load curves for physical assets, such as the expected standard load pattern for the next 5 years provided by the dispatching department; step 803, based on the calibrated physical degradation model and typical future load curves, performing a simulation on an offline end. In this simulation, the system continuously calculates the virtual insulation aging rate of the asset and calculates a Remaining Usable Life (RUL) index characterizing the maintenance cycle. The RUL index is defined as the virtual health status (e.g., insulation) of the physical asset during the simulation. The virtual time corresponding to when the degree of polymerization (DP value) of the insulating paper deteriorates to a preset maintenance threshold (e.g., DP=200), for example, the calculated result is 6.2 years; Step 804, the remaining usable life index (6.2 years) is output to the Asset Management and Maintenance Decision System (AMS) to provide a basis for the competent department to formulate maintenance strategies; At the same time, in order to solve the problem that the simplified on-site operating parameters (e.g., the ambient temperature measured by the standard weather station) may not accurately reflect the actual heat dissipation environment of the asset due to direct sunlight or wind and cold, before step 103, the method may also include an input compensation procedure: Step 901, reuse the asset thermodynamic model contained in the calibrated physical degradation model to generate an additional function characterizing the relationship between steady-state load and temperature rise. Step 902: During online operation, obtain the real-time load of physical assets. and real-time temperature (e.g., readings from a conventional winding thermometer); Step 903, real-time load Substituting the values ​​into the function representing the relationship between steady-state load and temperature rise, the theoretical temperature rise value is calculated. Step 904: Subtract the theoretical temperature rise from the real-time temperature to calculate an equivalent heat dissipation temperature. In one specific embodiment, the equivalent heat dissipation temperature The calculation rules follow: ,in The real-time temperature obtained in step 902 (e.g., 95°C) ), Step 903 is based on real-time load. Calculated theoretical temperature rise (e.g., 60) Then, the reverse calculation yields... ;this (35) This includes additional thermal effects, such as those caused by direct afternoon sunlight, which may be higher than the standard ambient temperature measured at a weather station 100 meters away at that moment (e.g., 30°C). Finally, in step 103, the simplified operating parameters are replaced with this more accurate equivalent heat dissipation temperature. This enables the online monitoring module to be based on 35 (instead of 30) Go to call the simplified calibration function.

[0026] Example 1: In a regional power grid operating under high summer load conditions, the transmission current of a critical 750 kV transmission section needs to be increased from 1950A to 2050A. The bottleneck of this section is a main transformer that has been in operation for 20 years, with a static safety limit of 2000A. The dispatcher faces the objective constraint that if this static limit is strictly followed, expensive standby units must be started, while forcibly exceeding this limit may cause irreversible damage to this physical asset known to have a chronic deterioration trend. In this scenario, the power intelligent monitoring system of the present invention, firstly, according to step 103 in the specific implementation method, its online monitoring module obtains simplified operating parameters in real time, namely, an ambient temperature of 38°C. With a current load of 1950A, the simplified calibration function generated offline in steps 101 and 102 was invoked. This function encapsulates the chronic thermal degradation state of the transformer as reflected by the DGA data calibration, and calculates its dynamic safety limit. It is 1980A.

[0027] Meanwhile, the real-time authentication module executes steps 104 and 105 in parallel. After cross-referencing the high-frequency PMU data it acquires with the T-type equivalent circuit impedance model, it generates an authentication signal. The value is 1.8%, which is lower than the preset safety threshold of 3%, indicating that the asset has not experienced any acute physical anomalies. The safety arbitration module then executes the safety logic gating rules of step 106, determining that both chronic health and acute verification have passed. The system then automatically updates the operating procedures of the intelligent dispatching system to 1980A. This dynamic safety limit of 1980A provides dispatchers with a previously completely missing decision-making basis. It first rejects dispatching operations that exceed the static limit to 2050A. At the same time, it quantifies the operation within the 1980A limit. This operation is based on the acute safety confirmation of the real-time verification module, thereby eliminating the uncertainty that dispatchers face when executing procedures in the traditional way. The dispatcher's decision-making problem changes from the dilemma of whether to take risks or bear economic losses to a deterministic engineering operation of performing optimization within the 1980A safety boundary and starting standby units to make up for the remaining 70A gap. The coordinated operation of the two mechanisms of chronic degradation monitoring and acute anomaly verification in this system provides a closed-loop control basis for a large-scale power grid security system that takes into account both long-term asset health and instantaneous physical reality.

[0028] Example 2: To verify the response of the power intelligent monitoring method of the present invention to both chronic degradation and acute anomalies of physical assets, a comparative test based on a standard power system simulation platform was conducted. This test simulated the operation of a 750 kV main transformer in a power grid under high temperature and heavy load conditions in summer, and compared the operational results of the method of the present invention with those of the traditional static safety limit procedure. The test platform used a simulation environment containing a standard transformer thermal model and an electromagnetic transient model. This environment simulated the thermal dynamic response of the transformer and generated a high-frequency phasor data stream conforming to PMU specifications, containing Gaussian white noise with a set signal-to-noise ratio of 40 dB. The simulation time step was set to 20 ms. The test object was a 750 kV main transformer with a long service life and exhibiting chronic degradation characteristics. Its simulated historical DGA data was analyzed offline to calibrate the equivalent heat dissipation coefficient in its physical degradation model. The value is 0.85, which is lower than the factory design value of 1.0; based on this... The calibration model, following steps 101 and 102, generated a simplified calibration function (LMF) specifically for the transformer offline, encapsulated as a lookup table; simultaneously, following the procedure in step 105, the verification signal threshold for acute anomaly detection was determined through multiple benchmark simulation calibrations. The percentage was 3%; the simulated acute physical abnormalities in the experiment were set to occur at a specific time. At that time, a cooling fan assembly of the transformer tripped due to a fault, causing a momentary drop in its heat dissipation capacity. The impact of this event was not included in the historical DGA data used to generate the LMF.

[0029] The experiment set up two parallel simulated scenarios. The control group used a traditional scheduling method, controlled according to the transformer's nameplate static safety limit specification 2000A. The experimental group deployed the power intelligent monitoring method of this invention, including an online monitoring module, a real-time authentication module, and a safety arbitration module. Both scenarios operated under the same external conditions, with the ambient temperature set at 40°C. The initial load was 1800A, and in The load was initially increased linearly at a rate of 5 A / s; during the test, the actual winding hot spot temperature of the transformer and the operating limits used by the control system were continuously monitored and recorded under both scenarios; in the initial stage of the test... to During this period, the load gradually increased from 1800A to nearly 2300A; the control group maintained a static limit of 2000A throughout this period; the online monitoring module of the experimental group adjusted the load based on real-time load and 40... Ambient temperature, calculated using LMF. This value decreases as the load increases, until... When the load reaches 1980A, the calculated value is... The value is 1980A, calculated by the real-time authentication module. (For example in) If the percentage (at 1.2%) is below the 3% threshold, the security arbitration module determines that it can pass, the system automatically updates the operating procedure to 1980A, and issues a load limiting command; An acute event, namely a cooling fan failure, was simulated and triggered. At this time, the control group load was 2000A, while the test group load was limited to 1980A. After the event was triggered, the real-time verification module of the test group monitored the changes in the PMU phasor data, and its calculated... exist The percentage increased to 4.5%, exceeding the 3% threshold; at this point, the online monitoring module calculated based on historical data... It remains 1980A, but the security arbitration module, based on gating rules, prevents its use due to failure of acute authentication. The system forcibly reverted the operating procedure to the preset conservative limit of 1400A and issued an alarm. The control group, which relied solely on static limits, did not detect any fan failure events, and the system continued to allow the transformer to carry a load of 2000A. Subsequent data showed that, due to a revert to the original operating procedures, the hot spot temperature in the test group was controlled at 136°C. Below, the control group, due to failure to limit the load, experienced a continuous rise in its hot spot temperature. It reached 152 at that time The experimental process and key data records are shown in Table 1.

[0030] Table 1: Key Data Recording Table for Simulation Experiments

[0031] The results of this simulation test show that the power intelligent monitoring method proposed in this invention can transform the chronic deterioration state of physical assets into dynamic safety limits through offline calibration and online simplified mapping mechanisms, which can be used to adjust the operating load. Its parallel real-time physical verification mechanism can monitor acute physical anomalies that cannot be reflected by historical data, and execute conservative control strategies through safety logic gating rules.

[0032] Example 3: This example combines Figures 1 to 4 This section describes the intelligent power monitoring methods and systems based on big data in the power sector, such as... Figure 1 As shown, the process begins offline, where historical asset big data is used to calibrate the physical degradation model in the offline calibration module, and then encapsulated into a simplified calibration function. On the online side, the online monitoring module calls this function and uses simplified operating parameters to calculate the dynamic safety limit. Meanwhile, the real-time authentication module uses high-frequency real-time physical measurement data to cross-compare with the physical reference model in parallel to generate authentication signals characterizing acute anomalies. At the same time, the dynamic safety limit is also sent to the economic impact assessment module to evaluate its economic cost in the sandbox pre-simulation mode and output economic impact indicators. Finally, the security arbitration module gathers the authentication signals and economic impact indicators and executes collaborative logic gating to make three decisions based on whether the authentication signals and economic indicators exceed the threshold: automatically update the scheduling procedure, prevent the update and rollback, or output collaborative decision support information.

[0033] like Figure 2 As shown in the figure, the horizontal axis represents time (s) and the vertical axis represents current (A). The static safety limit is displayed as a horizontal line that remains constant at 2000A, while the dynamic safety limit is a curve that slowly decreases over time. When the actual load curve increases and reaches the dynamic safety limit at 36s, the system's load limiting action prevents it from rising further and instead follows the constraint of the dynamic safety limit. Figure 3 As shown, it achieves this through the synergy of four major functional modules: offline calibration and encapsulation, online chronic degradation monitoring, parallel real-time acute verification, and secure logic gating arbitration. This establishes a dual-protection mechanism of chronic degradation monitoring and acute anomaly verification. Figure 4As shown, the module interaction sequence is illustrated in operating condition 1: chronic degradation and acute normal scenario. The online monitoring module calculates the dynamic safety limit as 1900A based on the chronic degradation state. The parallel real-time authentication module obtains high-frequency phasor data from the PMU measurement device and generates an authentication signal value of 1.5% after cross-comparison. After receiving these two inputs, the safety arbitration module executes safety logic gating, determines that 1.5% does not exceed the 3% threshold, judges that the authentication is passed, approves the automatic update, and sends an instruction to the intelligent scheduling system to update the operating procedure from the static 2000A to the dynamic 1900A.

[0034] Example 4: This comparative example uses basically the same experimental setup and objects as Example 2 to simulate a machine with an equivalent heat dissipation coefficient. The 750 kV main transformer, rated at 0.85, is at 40 At ambient temperature, the load increases linearly from 1800A at a rate of 5A / s, and... The simulation triggers an acute physical anomaly causing a cooling fan assembly failure. The key difference from Example 2 is that the monitoring method in this comparative example removes the real-time authentication module and the security arbitration logic based on the authentication signal. The system relies solely on the dynamic safety limit calculated by the online monitoring module based on the simplified calibration function (LMF). To update the operating procedures, this setup simulates a technical path that only considers incorporating the chronic deterioration state of assets into scheduling decisions; the experiment was conducted according to the steps of Example 2; in to During this period, the load increased from 1800A to nearly 2300A; the online monitoring module adjusted the load based on real-time load and 40 Ambient temperature, calculated using LMF. This value decreases as the load increases; When the load reaches 1980A, the calculated value is... The system was set to 1980A, and the operating procedures were automatically updated to 1980A. A load limiting command was issued, and the performance during this phase was consistent with the test group in Example 2, indicating that the method of removing some modules can also respond to the chronic degradation of assets; however, in When an acute event, such as a cooling fan failure, is triggered, the system fails to detect this sudden change in physical state due to the lack of a real-time verification mechanism. Therefore, the online monitoring module continues to perform calculations based on unchanged simplified operating parameters and a fixed LMF (Limited Flow Factor), and its output... The value remained near 1980A, and the system did not execute any procedural rollback or alarm actions; the transformer continued to be allowed to carry a load close to 1980A despite the reduced cooling capacity; key data records are shown in Table 2 (see Table 2); the data shows that, Subsequently, the system limit in this comparative example remained at around 1980A, failing to recede to 1400A as in the test group of Example 2; correspondingly, its winding hot spot temperature rose rapidly after the acute event, with a faster rate of increase than the control group in Example 2. The temperature of the hot spot has reached 155 degrees Celsius. and in The time rose further to 161 The test exceeded the common safe operating threshold; at the same time, due to the lack of verification signals, the system failed to issue any alarms regarding the abnormal physical state. The test process and key data records are shown in Table 2.

[0035] Table 2: Key data recording table for simulation experiment of Comparative Example 1.

[0036] The comparative test results show that simply incorporating the chronic degradation assessment results based on historical data into the operating procedures, without a parallel acute anomaly verification mechanism based on real-time physical measurements and corresponding safety arbitration logic, cannot provide an effective safety response when sudden physical anomalies occur. The system will maintain an operating limit that no longer conforms to physical reality due to the time scale limitations of the information source, which may lead to equipment overheating or operational risks.

[0037] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0038] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A power intelligent monitoring method based on power big data, applied to a large-scale power grid security and intelligent dispatch system, characterized in that, The methods include: Step 101: On the offline end, aggregate historical asset big data of physical assets in the power grid, retrieve the physical degradation model, and use the historical asset big data to calibrate the parameters of the physical degradation model. Step 102: Based on the calibrated physical degradation model, a simplified calibration function is generated through offline calculation, dimensionality reduction and encapsulation. The input of the simplified calibration function is set to at least one simplified operating parameter that the intelligent scheduling system can obtain in real time, and the output of the simplified calibration function is defined as the dynamic security limit of the physical asset. Step 103: In the online monitoring module, load the simplified calibration function, obtain the simplified operating parameters in real time, and call the simplified calibration function to calculate the dynamic safety limit; Step 104: Obtain high-frequency real-time physical measurement data of the physical assets and retrieve a physical reference model of the physical assets; Step 105: Instantly cross-compare the real-time physical measurement data with the expected output of the physical reference model under the current operating conditions to generate an identification signal characterizing acute physical anomalies. Step 106: Establish and execute a security logic gating rule, wherein when the value of the authentication signal does not exceed the preset security threshold, the static security limit procedure for physical assets in the intelligent scheduling system or security assurance system is automatically updated to the dynamic security limit calculated in step 103; when the value of the authentication signal exceeds the security threshold, the automatic update is prevented, and the operation procedure of the intelligent scheduling system or security assurance system is automatically rolled back to a preset conservative limit.

2. The intelligent power monitoring method based on big data as described in claim 1, characterized in that, The historical asset big data in step 101 includes at least one of the following: dissolved gas data in oil, partial discharge data, infrared thermal imaging data, historical load curves, and maintenance records; and the physical degradation model in step 101 includes at least one of the following: a hot spot temperature model of equipment based on thermodynamic principles and an aging model of insulation materials based on electrochemical principles.

3. The intelligent power monitoring method based on big data as described in claim 1, characterized in that, The simplified calibration function in step 102 is encapsulated as at least one of a lookup table, a polynomial function, and a piecewise linear function; and the simplified operating parameters in step 103 include at least one of ambient temperature and real-time load current.

4. The intelligent power monitoring method based on big data as described in claim 1, characterized in that, The high-frequency real-time physical measurement data in step 104 is phasor measurement unit data, which includes voltage phasors and current phasors; and the physical reference model in step 104 is the T-type equivalent circuit impedance model of the physical asset.

5. The intelligent power monitoring method based on big data as described in claim 1, characterized in that, The dynamic safety limit updated in step 106 is used as a rigid physical constraint for the operation of at least one of the optimized power flow algorithm and the safety and stability control device.

6. The intelligent power monitoring method based on big data as described in claim 1, characterized in that, After step 103 and before step 106, the method further includes: step 601: using the dynamic safety limit calculated in step 103 as a virtual constraint; step 602: running a simulation based on the virtual constraint in the sandbox pre-simulation mode of at least one of the economic scheduling module and the safety verification module to evaluate the economic impact index of applying the dynamic safety limit; step 603: determining whether the economic impact index exceeds a preset economic threshold; and step 106 is modified to: automatically updating is performed only when the value of the authentication signal does not exceed the safety threshold and the economic impact index does not exceed the economic threshold; when the value of the authentication signal exceeds the safety threshold, a rollback is performed; when the value of the authentication signal does not exceed the safety threshold but the economic impact index exceeds the economic threshold, automatic updating is prevented, and a collaborative decision-making auxiliary information containing the dynamic safety limit and the economic impact index is output to the intelligent scheduling system instead.

7. The intelligent power monitoring method based on big data as described in claim 1, characterized in that, Following step 102, the method further includes: Step 701: On the offline end, independently analyze all historical operational big data of the physical asset to construct a statistical security envelope that characterizes its historical experience security boundary; and in step 106, the automatic update rule is modified to: when the value of the authentication signal does not exceed the security threshold, compare the dynamic security limit with the limit of the statistical security envelope under the current operating conditions, take the smaller value of the two as the final dynamic security limit, and use the final dynamic security limit to perform automatic updates.

8. The intelligent power monitoring method based on big data as described in claim 1, characterized in that, Following step 102, the method further includes: Step 801: reusing the calibrated physical degradation model generated in step 101; Step 802: obtaining at least one of the typical future load curve and expected operating mode of the physical asset; Step 803: performing a simulation based on the calibrated physical degradation model and the typical future load curve to calculate a remaining usable life index characterizing the maintenance cycle, wherein the remaining usable life index is defined as the virtual time corresponding to when the virtual health state of the physical asset deteriorates to a preset maintenance threshold during the simulation; Step 804: outputting the remaining usable life index to... In at least one of the asset management and maintenance decision systems, prior to step 103, the method further includes: step 901: generating an additional function characterizing the relationship between steady-state load and temperature rise based on the asset thermodynamic model contained in the calibrated physical degradation model; step 902: acquiring the real-time load and real-time temperature of the physical asset during online operation; step 903: substituting the real-time load into the function characterizing the relationship between steady-state load and temperature rise to calculate the theoretical temperature rise value; step 904: subtracting the theoretical temperature rise value from the real-time temperature to reverse-calculate an equivalent heat dissipation temperature; and in step 103, the simplified operating parameters are replaced with the equivalent heat dissipation temperature.

9. The intelligent power monitoring method based on big data as described in claim 8, characterized in that, The equivalent heat dissipation temperature calculated in reverse in step 904 Its calculation rules follow: ,in, The real-time temperature obtained in step 902. The theoretical temperature rise value calculated in step 903 is based on the real-time load and a function characterizing the relationship between steady-state load and temperature rise.

10. A power intelligent monitoring system based on power big data, used to implement the power intelligent monitoring method based on power big data as described in claim 1, characterized in that the system... include: An offline calibration module is configured to: aggregate historical asset big data of physical assets in the power grid; retrieve physical degradation models; calibrate the parameters of the physical degradation models using historical asset big data; and, based on the calibrated physical degradation models, reduce dimensions, encapsulate, and generate a simplified calibration function through offline calculation. An online monitoring module is configured to: load a simplified calibration function; obtain at least one simplified operating parameter corresponding to the input of the simplified calibration function in real time; and call the simplified calibration function to calculate the dynamic security limit of the physical asset. A real-time authentication module is configured to: acquire high-frequency real-time physical measurement data of physical assets; retrieve a physical reference model of physical assets; and instantaneously cross-compare the high-frequency real-time physical measurement data with the expected output of the physical reference model under the current operating conditions to generate an authentication signal characterizing an acute physical anomaly. A security arbitration module is configured to: establish and execute a security logic gating rule, wherein when the value of the authentication signal does not exceed a preset security threshold, the static security limit procedure for physical assets in the intelligent scheduling system or security assurance system is automatically updated to a dynamic security limit; otherwise, when the value of the authentication signal exceeds the security threshold, the automatic update is prevented, and the operating procedure of the intelligent scheduling system or security assurance system is automatically rolled back to a preset conservative limit.

Images