Electric power digital anomaly identification method and system based on digital twinning

Abstract

The invention discloses an electric power digital anomaly identification method and system based on digital twinning, and relates to the technical field of electric power. Comprising the following steps: constructing a digital twinborn model according to power data of a plurality of time periods in a current detection period; obtaining multi-source operation data in the power system, and preprocessing the multi-source operation data; according to actually collected data, inputting the data into a digital twin model for simulation, and generating virtual operation data; calculating a data deviation value between the virtual data and a normal state according to a dynamic deviation coefficient formula; on the basis of the deviation value, abnormal features are extracted, and the abnormal features are judged; and generating a corresponding early warning response according to a judgment result. According to the method, through accurate simulation of the digital twin model and comprehensive analysis of multi-source data, abnormal feature extraction and grading judgment are combined, the abnormal condition of the power system can be identified more accurately, and misjudgment and missed judgment are reduced.

Description

Technical Field

[0001] This invention relates to the field of power technology, specifically to a method and system for identifying digital anomalies in power systems based on digital twins. Background Technology

[0002] Based on relevant technologies, a power system can be constructed based on a target digital twin model, thereby enabling the management and optimization of power system equipment.

[0003] However, when the power system is in operation, abnormal situations such as power system failures, power distribution system failures, and electricity theft may occur, affecting users' normal electricity use.

[0004] According to relevant technologies, identifying abnormal conditions requires a step-by-step investigation of the power generation system, distribution system, and power consumption system within the power system. This process consumes a significant amount of manpower and resources, makes it difficult to guarantee repair time, and affects the normal operation of the power system. Therefore, this paper proposes a digital anomaly identification method and system for power systems based on digital twins. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for identifying digital power anomalies based on digital twins, so as to solve the problems mentioned in the background art.

[0006] According to a first aspect of the present invention, a method is provided, comprising:

[0007] A digital twin model is constructed based on power data from multiple time periods within the current monitoring cycle.

[0008] Acquire multi-source operational data from the power system and preprocess it;

[0009] Based on the actual collected data, it is input into the digital twin model for simulation to generate virtual running data;

[0010] The deviation value between the virtual data and the data under normal conditions is calculated based on the dynamic deviation coefficient formula.

[0011] Based on the deviation value, its abnormal features are extracted and the abnormal features are judged.

[0012] Based on the judgment results, a corresponding early warning response will be generated.

[0013] According to a second aspect of the present invention, a power digital anomaly identification system based on digital twins is provided, comprising:

[0014] The data acquisition unit is used to collect multi-source operational data in real time and complete timestamp alignment to ensure the spatiotemporal correlation of the data;

[0015] The model building unit generates a digital twin model by training a spatiotemporal fusion algorithm based on the acquired multi-source operational data.

[0016] The simulation analysis unit, based on the received real-time data, uses the real-time interaction interface between the physical device and the virtual model to drive the digital twin model to perform simulation at a frequency of seconds.

[0017] The anomaly identification unit calculates the deviation between the virtual data and the normal baseline data according to the dynamic deviation coefficient formula, and extracts the anomaly features using the isolated forest algorithm.

[0018] The early warning response unit triggers the corresponding early warning response based on the determination of abnormal characteristics.

[0019] According to a third aspect of the present invention, a power digital anomaly identification device based on digital twins is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to invoke the instructions stored in the memory to execute the power digital anomaly identification method based on digital twins.

[0020] According to a fourth aspect of the present invention, a computer-readable storage medium is provided having computer program instructions stored thereon, which, when executed by a processor, implement the digital twin-based power digitization anomaly identification method.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] This digital twin-based power anomaly identification method and system, through precise simulation of the digital twin model and comprehensive analysis of multi-source data, combined with anomaly feature extraction and hierarchical judgment, can more accurately identify anomalies in the power system, reducing false positives and false negatives. Utilizing 5G edge computing, it achieves rapid data transmission and processing. The digital twin model synchronizes data at a second-level frequency, ensuring the real-time nature of anomaly identification and early warning response. It can promptly detect and handle problems. Through hierarchical response to anomalies, minor anomalies are promptly inspected, moderate anomalies are quickly repaired, and severe anomalies are handled urgently, effectively preventing the escalation of faults and ensuring the reliable operation of the power system. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the process of the present invention;

[0024] Figure 2 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0027] like Figure 1 The flowchart shown is a schematic diagram of a power digital anomaly identification method based on digital twin according to an embodiment of the present invention. The method includes:

[0028] S1: Construct a digital twin model based on power data from multiple time periods within the current detection cycle;

[0029] S2: Acquire multi-source operating data from the power system and preprocess it;

[0030] S3: Based on the actual collected data, input it into the digital twin model for simulation to generate virtual running data;

[0031] S4: Calculate the deviation value between the virtual data and the data under normal conditions based on the dynamic deviation coefficient formula;

[0032] S5: Based on the deviation value, extract its abnormal features and determine the abnormal features;

[0033] S6: Based on the judgment result, generate a corresponding early warning response.

[0034] According to an embodiment of the present invention, in step S1, a digital twin model is constructed based on power data from multiple time periods within the current detection cycle. The digital twin model is a virtual mapping of the physical power system. By fusing power data from different time periods, the operating state of the power system is simulated. The model is trained using a spatiotemporal fusion algorithm, which can simultaneously process the time series characteristics (such as voltage change trends at different times) and spatial distribution characteristics (such as current distribution differences at different nodes) of power data.

[0035] According to the formula:

[0036]

[0037] in, For parameters The spatiotemporal fusion function (fusion of time series features and spatial distribution features) uses power data samples from different time periods as input. By performing fusion calculations, digital twin models are generated for power data samples from different time periods. The power data includes parameters such as voltage, current, and power. This data is historically accumulated and has undergone preliminary sorting of structured data. It serves as the basic data for model training and is used to build a digital twin model that can accurately reflect the actual operating characteristics of the power system. This provides a virtual scene basis for subsequent simulation and anomaly identification. The parameter θ will be continuously optimized during the model training process to improve the model's simulation accuracy of the actual power system.

[0038] According to an embodiment of the present invention, in step S2, multi-source operating data of the power system is acquired. The multi-source data acquisition aims to comprehensively acquire various types of data that affect the operation of the power system. Internal sensors are directly connected to the equipment and lines of the power system to perceive the core operating parameters of the power system in real time. External sensors are used to capture environmental factors that may affect the operation of the power system, because environmental conditions (such as high temperature may cause line resistance to increase) will indirectly affect the performance of the power system. Therefore, when acquiring data, internal sensors are used to acquire real-time power data, such as voltage data of each node acquired by voltage sensors, line current data acquired by current sensors, and active power and reactive power data acquired by power sensors. External sensors are used to acquire relevant impact data, such as environmental parameters that may affect the operation of the power system, such as temperature, humidity, and wind speed at different time periods. The combination of the two provides a comprehensive and real-time data source for the state analysis and anomaly identification of the power system, and is the input basis for subsequent data processing and simulation.

[0039] 5G edge computing nodes feature low latency and high bandwidth, enabling timestamp alignment of multi-source data. By synchronizing data from different sources and at different collection times, the spatiotemporal correlation between data is ensured, providing a consistent time reference for subsequent data processing and analysis.

[0040] According to an embodiment of the present invention, in step S2, the acquired power data is preprocessed, wherein impulse noise in parameters such as voltage and current is identified and removed by the Raida criterion (3σ principle). This criterion states that in a normal distribution, approximately 99.7% of the data will fall within the range of the mean plus or minus 3 times the standard deviation. Data outside this range can be regarded as outliers (impulse noise). By calculating the mean and standard deviation of the data, data outside the 3σ range are removed to ensure the accuracy of the data.

[0041] LSTM (Long Short-Term Memory) networks can capture long-term dependencies in time series data. By learning from historical data, they can predict missing data values ​​with an imputation error rate of less than 3%, ensuring data integrity.

[0042] Because different parameters have different units (such as voltage units) Power unit Direct comparison and analysis of these parameters would be affected; therefore, according to the formula:

[0043]

[0044] in This is the original data. These are the minimum and maximum values ​​in the historical data, respectively. The data is then transformed to the [0,1] interval to eliminate the influence of units and make different parameters comparable.

[0045] The preprocessed time-domain data is converted to the frequency domain, and harmonic component features are extracted. Harmonics in the power system can affect the stable operation of the system. Harmonic component features can be used as auxiliary features for anomaly identification to improve the accuracy of anomaly identification.

[0046] After preprocessing, the data eliminates impulse noise, fills in missing values, and completes normalization and Fourier transform, becoming clean, unified, and structured data containing more feature information. This provides high-quality data for inputting into the digital twin model for simulation, reduces the impact of data noise and inconsistencies on simulation results, and the extracted harmonic features provide supplementary basis for anomaly identification.

[0047] According to an embodiment of the present invention, in S3, the data actually collected is input into the digital twin model for simulation to generate virtual operation data. In this process, a real-time interactive interface between the physical device and the virtual model is established using edge computing nodes, which realizes low-latency data transmission between the physical device and the virtual model. The preprocessed power data is synchronized with the digital twin model according to the second-level frequency value to ensure that the virtual model can track the state changes of the physical system in real time and realize the dynamic alignment between the physical state and the virtual state.

[0048] Digital twin models perform multi-dimensional simulations of input power data. They simulate the operating mechanisms of equipment in a power system, the dynamic changes in network topology, the energy flow paths, and the evolution of potential faults, thereby generating virtual operating data.

[0049] By comparing virtual data with actual data, calculating the deviation between the two, and dynamically adjusting the model parameters based on the deviation, the model is continuously optimized to ensure that the virtual operating data can accurately reflect the actual operating state of the physical system.

[0050] The input is preprocessed real-time power data, and the output is virtual operating data, including equipment operating parameters such as virtual voltage and virtual current, as well as information such as network topology changes, energy flow paths, and potential fault evolution trends. This data is virtual data that corresponds to the physical system in real time. The virtual operating data serves as a mirror image of the physical system and is used to compare with the data under normal conditions, providing a virtual reference for subsequent deviation calculations and anomaly identification.

[0051] According to an embodiment of the present invention, in step S4, the deviation value between the virtual data and the data under normal conditions is calculated according to the dynamic deviation coefficient formula, wherein,

[0052] According to the formula:

[0053]

[0054] To quantify the difference between virtual operation data and normal state data, for The dynamic deviation coefficient at time t, This is virtual operational data. Rt is the baseline data under normal conditions (the average of historical periods without anomalies), ϵ is the minimum value to avoid a denominator of 0, and αt is the time decay factor (linearly decreasing with the detection cycle, ranging from 0.8 to 1.0), which reflects the timeliness of the data. As time goes on, the reference value of historical data gradually decreases. Adjusting αt makes the deviation calculation more consistent with the actual situation.

[0055] The inputs in this step are the virtual running data Vt and the baseline data Rt under normal conditions. The output is the dynamic deviation coefficient Kt at time t, which is a quantitative deviation index. Kt reflects the degree of deviation between the virtual running data and the normal data and is the core basis for subsequent abnormal feature extraction and judgment.

[0056] According to an embodiment of the present invention, in step S5, abnormal features are extracted based on the deviation value. The isolated forest algorithm is a highly efficient anomaly detection algorithm. It constructs multiple isolated trees to perform feature dimensionality reduction on the deviation coefficient sequence [K1, K2, ..., Kn]. This algorithm can quickly identify abnormal data that deviates from the normal pattern, thereby extracting three core features: deviation mutation rate (e.g., Kt increases by more than 50% within 5 minutes), duration (number of consecutive samples where Kt exceeds the threshold), and spatial correlation (multiple nodes in the same area simultaneously exceeding the threshold). The deviation mutation rate reflects the magnitude of change of the deviation coefficient in a short period of time (e.g., Kt increases by more than 50% within 5 minutes), the duration reflects the duration for which the deviation coefficient exceeds the threshold, and spatial correlation indicates the situation where multiple nodes in the same area simultaneously exceed the deviation threshold.

[0057] Determining abnormal features includes:

[0058] Three threshold levels are set (Kth1=0.1, Kth2=0.3, Kth3=0.5). When Kt∈(Kth1,Kth2], it is judged as a slight anomaly; when Kt∈(Kth2,Kth3], it is judged as a moderate anomaly; and when Kt>Kth3, it is judged as a severe anomaly. The degree of anomaly is judged according to the interval in which Kt is located, so as to realize the classification of anomalies. The anomaly characteristics and judgment results provide a clear basis for subsequent early warning response and determine what level of response measures to take.

[0059] According to an embodiment of the present invention, in step S6, a corresponding early warning response is generated based on the determination result, wherein, when determining the abnormal result,

[0060] When a minor anomaly is detected, a yellow alert is pushed to the operation and maintenance terminal through the system backend, along with a deviation data trend chart and the location information of related equipment, prompting that the inspection be completed within 12 hours;

[0061] When a moderate anomaly is detected, the audible and visual alarm device is triggered, a maintenance work order is automatically generated and assigned to the nearest maintenance team, and the simulation parameters of the digital twin model are adjusted to focus on monitoring the abnormal area.

[0062] When a serious anomaly is determined, immediately disconnect the electrical connection between the abnormal area and the main grid (e.g., trigger the circuit breaker to trip), send a red warning signal to the dispatch center, simultaneously start the backup power switching procedure, and record complete data for 10 minutes before and after the anomaly for post-event analysis.

[0063] The input is the anomaly determination result, and the output includes various early warning signals, work order information, operation instructions, and recorded data before and after the anomaly. The early warning response measures are used to promptly handle abnormal situations in the power system and ensure the safe and stable operation of the power system. The recorded data provides a basis for subsequent system optimization and fault analysis.

[0064] The power digital anomaly identification method based on digital twin according to an embodiment of the present invention is based on digital twin technology. It achieves real-time simulation of the power system's operating status by constructing a virtual model that precisely corresponds to the physical power system. First, power data from multiple time periods within the current detection cycle is collected. A spatiotemporal fusion algorithm is used to construct a digital twin model, which is a mapping of the power system. Next, multi-source operating data is acquired using sensors. After timestamp alignment by 5G edge computing nodes, the alignment undergoes preprocessing, such as noise removal, missing value filling, normalization, and harmonic feature extraction, to obtain high-quality data. The preprocessed data is then input into the digital twin model and synchronized at a second-level frequency. Multi-dimensional simulation is performed using the model to generate virtual operating data, which is then compared with actual data. Subsequently, the deviation between the virtual data and normal baseline data is calculated using a dynamic deviation coefficient formula. The isolated forest algorithm is used to extract abnormal features from the deviation coefficient sequence, and the abnormality level is determined based on a three-level threshold. Finally, a corresponding early warning response is generated according to the abnormality level to ensure the safe operation of the power system.

[0065] like Figure 2 As shown, this is a schematic diagram of a power digital anomaly identification system based on digital twins according to an embodiment of the present invention. The system includes:

[0066] The data acquisition unit is used to collect multi-source operational data in real time and complete timestamp alignment to ensure the spatiotemporal correlation of the data;

[0067] The model building unit generates a digital twin model by training a spatiotemporal fusion algorithm based on the acquired multi-source operational data.

[0068] The simulation analysis unit, based on the received real-time data, uses the real-time interaction interface between the physical device and the virtual model to drive the digital twin model to perform simulation at a frequency of seconds.

[0069] The anomaly identification unit calculates the deviation between the virtual data and the normal baseline data according to the dynamic deviation coefficient formula, and extracts the anomaly features using the isolated forest algorithm.

[0070] The early warning response unit triggers the corresponding early warning response based on the determination of abnormal characteristics.

[0071] According to an embodiment of the present invention, a power digital anomaly identification device based on digital twin is provided, comprising: a processor and a memory for storing processor-executable instructions;

[0072] The processor is configured to invoke instructions stored in memory to perform the method as described in any one of claims 1-8.

[0073] According to one embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored, wherein the computer program instructions, when executed by a processor, implement the power digital anomaly identification method based on digital twin.

[0074] In summary, this invention discloses a digital twin-based method and system for identifying digital anomalies in the power system. It includes: constructing a digital twin model based on power data from multiple time periods within the current detection cycle; acquiring and preprocessing multi-source operational data from the power system; inputting the actually collected data into the digital twin model for simulation to generate virtual operational data; calculating the deviation value between the virtual data and the data under normal conditions using a dynamic deviation coefficient formula; extracting abnormal features based on the deviation value and determining the abnormal features; and generating a corresponding early warning response based on the determination result. This invention, through precise simulation of the digital twin model and comprehensive analysis of multi-source data, combined with anomaly feature extraction and hierarchical determination, can more accurately identify abnormal situations in the power system, reducing false positives and false negatives. Utilizing 5G edge computing enables rapid data transmission and processing. The digital twin model synchronizes data at a second-level frequency, ensuring the real-time nature of anomaly identification and early warning response, enabling timely detection and handling of problems. Through hierarchical responses to anomalies—timely inspection of minor anomalies, rapid repair of moderate anomalies, and emergency handling of severe anomalies—it effectively prevents fault escalation and ensures the reliable operation of the power system.

[0075] The present invention may be a method, apparatus, system and / or computer program product, and the computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the present invention.

[0076] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. A method and system for identifying digital power anomalies based on digital twins, characterized in that, include: A digital twin model is constructed based on power data from multiple time periods within the current monitoring cycle. Acquire multi-source operational data from the power system and preprocess it; Based on the actual collected data, it is input into the digital twin model for simulation to generate virtual running data; The deviation value between the virtual data and the data under normal conditions is calculated based on the dynamic deviation coefficient formula. Based on the deviation value, its abnormal features are extracted and the abnormal features are judged. Based on the judgment results, a corresponding early warning response will be generated.

2. The power digital anomaly identification method and system based on digital twin as described in claim 1, characterized in that: Based on power data from multiple time periods within the current monitoring cycle, a digital twin model is constructed, including: Model training is performed using a spatiotemporal fusion algorithm. According to the formula: Generate digital twin models ,in, For parameters The spatiotemporal fusion function fuses time series features and spatial distribution features. This consists of power data samples from different time periods. The power data includes parameters such as voltage, current, and power.

3. The power digital anomaly identification method and system based on digital twin as described in claim 1, characterized in that: Acquire multi-source operational data from the power system, including: Real-time power data is acquired using internal sensors, such as voltage data of each node collected by voltage sensors, line current data collected by current sensors, and active and reactive power data collected by power sensors. External sensors are used to acquire relevant impact data, such as environmental parameters that may affect the operation of the power system, such as temperature, humidity, and wind speed at different times. By leveraging 5G edge computing nodes, timestamp alignment of multi-source data can be achieved, ensuring the spatiotemporal correlation of data.

4. The power digital anomaly identification method and system based on digital twin as described in claim 1, characterized in that: The acquired power data is preprocessed, including: Impulse noise in parameters such as voltage and current is identified and removed using the Raida criterion. Missing data is then filled in using an LSTM-based time series prediction model, with an infilling error rate of less than 3%. Parameters of different quantities of steel, such as voltage units. Power unit Transform to the [0,1] interval, according to the formula: in This is the original data. These are the minimum and maximum values ​​in the historical data, respectively. Fourier transform is performed on the preprocessed data to extract harmonic component features, which are used as auxiliary feature dimensions for anomaly identification.

5. The power digital anomaly identification method and system based on digital twin as described in claim 1, characterized in that: Based on the actual collected data, it is input into the digital twin model for simulation, generating virtual running data including: By using edge computing nodes to establish a real-time interactive interface between physical devices and the virtual model, and by using a digital twin model of preprocessed power data with second-level frequency synchronization values, dynamic alignment between the physical and virtual states can be achieved. Multi-dimensional simulations are performed on the input power data to generate equipment operating parameters, such as virtual voltage, virtual current, network topology changes, energy flow paths, and potential fault evolution trends. By comparing virtual data with actual data and dynamically adjusting model parameters, the reliability of virtual running data can be ensured.

6. The power digital anomaly identification method and system based on digital twin as described in claim 1, characterized in that: Based on the dynamic deviation coefficient formula, the deviation value between the virtual data and the data under normal conditions is calculated, including: According to the formula: in, for The dynamic deviation coefficient at time t, The data is for virtual operation. Rt is the baseline data under normal conditions, which is the average of the same period in history without abnormalities. ϵ is the minimum value to avoid the denominator being 0. αt is the time decay factor, which decreases linearly with the increase of the detection cycle, ranging from 0.8 to 1.

0.

7. The power digital anomaly identification method and system based on digital twin as described in claim 1, characterized in that: Based on the deviation value, its abnormal features are extracted, including: The Isolation Forest algorithm is used to perform feature dimensionality reduction on the bias coefficient sequence [K1,K2,...,Kn], and three core features are extracted: bias mutation rate, duration, and spatial correlation. Determining abnormal features includes: Three threshold levels are set (Kth1=0.1, Kth2=0.3, Kth3=0.5). When Kt∈(Kth1,Kth2], it is judged as a slight abnormality; when Kt∈(Kth2,Kth3], it is judged as a moderate abnormality; and when Kt>Kth3, it is judged as a severe abnormality.

8. The power digital anomaly identification method and system based on digital twin as described in claim 1, characterized in that: Based on the judgment results, the corresponding early warning response includes: When a minor anomaly is detected, a yellow alert is pushed to the operation and maintenance terminal through the system backend, along with a deviation data trend chart and the location information of related equipment, prompting that the inspection be completed within 12 hours; When a moderate anomaly is detected, an audible and visual alarm device is triggered, a maintenance work order is automatically generated and assigned to the nearest maintenance team, and the simulation parameters of the digital twin model are adjusted to focus on monitoring the abnormal area. When a serious anomaly is determined, the electrical connection between the abnormal area and the main grid is immediately cut off. If the circuit breaker trips, a red warning signal is sent to the dispatch center, and the backup power supply switching procedure is started simultaneously. Complete data for 10 minutes before and after the anomaly occurs is recorded for post-event analysis.

9. A power digital anomaly identification system based on digital twins, characterized in that, include: The data acquisition unit is used to collect multi-source operational data in real time and complete timestamp alignment to ensure the spatiotemporal correlation of the data; The model building unit generates a digital twin model by training a spatiotemporal fusion algorithm based on the acquired multi-source operational data. The simulation analysis unit, based on the received real-time data, uses the real-time interaction interface between the physical device and the virtual model to drive the digital twin model to perform simulation at a frequency of seconds. The anomaly identification unit calculates the deviation between the virtual data and the normal baseline data according to the dynamic deviation coefficient formula, and extracts the anomaly features using the isolated forest algorithm. The early warning response unit triggers the corresponding early warning response based on the determination of abnormal characteristics.

10. A power digital anomaly identification device based on digital twins, characterized in that, include: The processor and memory used to store instructions that can be executed by the processor; The processor is configured to invoke instructions stored in memory to execute the method as described in any one of claims 1-8.

Images