Centralized control center-based signal automatic patrolling and abnormal intelligent analysis method and device

Abstract

The application proposes a signal automatic patrol and abnormal intelligent analysis method and device based on a centralized control center. By constructing a patrol logic control mechanism, combining real-time data acquisition and operation trend analysis, identifying abnormal states and potential fault signs of power equipment during operation, periodically or instantaneously analyzing the operation of target objects, and automatically generating patrol reports and abnormal level division results. This method can realize intelligent identification and response of characteristic problems such as ultra-low frequency oscillation and signal fluctuation in new energy access and other new power system scenarios, help to improve the intelligent level of operation monitoring of the centralized control center, improve the early warning ability and accident response efficiency of equipment failure, and avoid operation risks and system disturbances caused by monitoring omissions.

Description

Technical Field

[0001] This invention relates to the field of power system monitoring and intelligent operation and maintenance, and in particular to a method and device for automatic signal inspection and intelligent anomaly analysis based on a centralized control center. Background Technology

[0002] With the accelerated large-scale integration of new energy sources, the new power system is exhibiting increasingly complex operational characteristics.

[0003] As the core platform for power system operation coordination and decision-making, the centralized control center bears the important responsibility of real-time perception and intelligent judgment of the operating status of power grid equipment. Related technologies have constructed a technical system for power equipment status monitoring through the collaborative operation of modules such as data acquisition, trend analysis, and anomaly identification. Specifically, this system covers the entire process from signal acquisition and data preprocessing to trend modeling, disturbance identification, and anomaly response, including key aspects such as periodic inspections, manual triggering mechanisms, multi-parameter analysis, and hierarchical push notifications. However, existing centralized control center monitoring methods directly employ fixed-period scanning or simple rule-triggered inspection strategies without dynamically adjusting task priorities based on equipment operating status. This may lead to high-risk equipment not being identified and addressed in a timely manner, or insufficient ability to capture slow-changing disturbance characteristics such as ultra-low frequency oscillations, thereby affecting the safety and stability of power grid operation. Especially in asynchronous interconnected regions, traditional methods are insufficient to effectively identify such disturbances, increasing the risk of potential fault evolution and system instability. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems in the related art.

[0005] Therefore, the first objective of this invention is to propose an automatic signal inspection and intelligent anomaly analysis method based on a centralized control center. This method aims to dynamically generate inspection tasks through a logic control module, combined with trend analysis and disturbance identification of real-time operational data, to achieve intelligent judgment, automatic classification, and result push of abnormal states. This method is applicable to both periodic inspections and manually triggered scenarios, and features automatic generation of analysis reports and a classification-based push mechanism. It helps reduce the workload of operators, improves the early detection and rapid response capabilities of equipment anomalies, and avoids equipment failures, system disturbances, or even large-scale power outages caused by missed or misjudged abnormal signals.

[0006] The second objective of this invention is to propose an automatic signal inspection and intelligent anomaly analysis device based on a centralized control center.

[0007] To achieve the above objectives, a first aspect of the present invention proposes a method for automatic signal inspection and intelligent anomaly analysis based on a centralized control center, comprising:

[0008] S1. Construct a set of patrol task triggering factors based on equipment operating status, historical anomaly records and periodic strategies, and generate a priority queue according to the set of triggering factors to dynamically schedule patrol tasks.

[0009] S2, collect real-time and historical operating data of the target device, perform missing data completion, anomaly removal and time sequence alignment processing on the data to obtain a structured data stream;

[0010] S3, use a sliding window to calculate the rate of change of key parameters, and perform a fast Fourier transform on the structured data stream to extract frequency domain disturbance features and identify whether there is ultra-low frequency oscillation;

[0011] S4 constructs a weighted scoring model based on voltage offset, frequency fluctuation rate, and signal duration, calculates anomaly scores, classifies anomaly levels, generates structured anomaly reports, and pushes information according to level.

[0012] In one embodiment of the present invention, S1 includes:

[0013] S11. Based on at least three indicators from equipment commissioning time, load change range, scheduling strategy and historical anomaly records, a weighted scoring algorithm is used to calculate the comprehensive risk index of the equipment, which serves as the quantitative basis for triggering factors.

[0014] S12 compares the comprehensive risk index with the preset threshold range. If the threshold is exceeded, the device is automatically added to the high-priority queue, and the task execution interval is dynamically adjusted according to the queue length.

[0015] In one embodiment of the present invention, S2 includes:

[0016] S21, a sliding window algorithm is used to perform linear interpolation or fill in missing data based on the mean of adjacent time periods to ensure data continuity;

[0017] S22 utilizes a rule-based anomaly detection algorithm, combined with a set threshold and a signal mutation identification model, to remove or replace outliers and align timestamps of multi-source data to construct a structured data stream with a unified time sequence.

[0018] In one embodiment of the present invention, S4 includes:

[0019] S41, calculate the correlation between voltage and frequency signals according to the Spearman correlation coefficient formula, and increase the weight of the anomaly score if the correlation exceeds the preset threshold.

[0020] S42 classifies anomalies into three levels based on anomaly scores: Level 1, Level 2, and Level 3. Level 1 anomalies trigger immediate alarms and are pushed to the duty personnel's terminals. Level 2 anomalies generate summary reports and are pushed periodically. Level 3 anomalies are only recorded in the system log for subsequent analysis.

[0021] In one embodiment of the present invention, it further includes:

[0022] S5. Perform pattern matching analysis on the identified ultra-low frequency oscillation signal, compare the signal with a preset disturbance template library to determine whether it belongs to a known disturbance type, and generate corresponding handling suggestions based on the matching results.

[0023] To achieve the above objectives, a second aspect of the present invention provides a signal automatic inspection and anomaly intelligent analysis device based on a centralized control center, comprising:

[0024] The task scheduling module is used to construct a set of patrol task triggering factors based on equipment operating status, historical anomaly records and periodic strategies, and generate a priority queue according to the set of triggering factors to dynamically schedule patrol tasks.

[0025] The data acquisition and preprocessing module is used to acquire real-time and historical operating data of the target device, and to perform missing data completion, anomaly removal and time sequence alignment processing on the data to obtain a structured data stream.

[0026] The time-frequency feature extraction module is used to calculate the rate of change of key parameters using a sliding window, and to perform a fast Fourier transform on the structured data stream to extract frequency domain perturbation features and identify whether ultra-low frequency oscillations exist.

[0027] The anomaly scoring and grading module is used to construct a weighted scoring model based on voltage offset, frequency fluctuation rate and signal duration, calculate anomaly scores and classify anomaly levels, generate structured anomaly reports and push information according to level.

[0028] The methods and apparatus of this invention can perform periodic or real-time inspections of the operating status of target objects, automatically analyze their historical and current operating data, identify potential anomalies, and take strategies such as immediate alarms or hierarchical summary push according to the anomaly level, so as to support the intelligent operation and maintenance needs under the new power system.

[0029] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0030] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0031] Figure 1 This is a flowchart of a method for automatic signal inspection and intelligent anomaly analysis based on a centralized control center, according to an embodiment of the present invention.

[0032] Figure 2 This is an architecture diagram of a signal automatic inspection and anomaly intelligent analysis method based on a centralized control center according to an embodiment of the present invention;

[0033] Figure 3 This is a flowchart illustrating the key technical route according to an embodiment of the present invention;

[0034] Figure 4 This is a structural diagram of a signal automatic inspection and anomaly intelligent analysis device based on a centralized control center, according to an embodiment of the present invention. Detailed Implementation

[0035] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0036] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. 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 should fall within the scope of protection of the present invention.

[0037] The following describes, with reference to the accompanying drawings, an embodiment of the present invention, a method and apparatus for automatic signal inspection and intelligent anomaly analysis based on a centralized control center.

[0038] Example 1

[0039] Figure 1 This is a flowchart of a signal automatic inspection and anomaly intelligent analysis method based on a centralized control center according to an embodiment of the present invention, such as... Figure 1 As shown, it includes:

[0040] S1 constructs a set of trigger factors for inspection tasks based on equipment operating status, historical anomaly records, and periodic strategies, and generates a priority queue based on the set of trigger factors to dynamically schedule inspection tasks.

[0041] Specifically, the step in this invention of "constructing a set of patrol task triggering factors based on equipment operating status, historical anomaly records, and periodic strategies, and generating a priority queue according to the set of triggering factors to dynamically schedule patrol tasks" is the core logic control mechanism for achieving efficient operation of the automatic signal patrol and intelligent anomaly analysis system. Its technical implementation principle is based on multi-source data fusion and task scheduling optimization algorithms, aiming to improve the response speed and resource utilization of patrol tasks.

[0042] In some implementations, this step first constructs a multi-dimensional set of trigger factors by collecting real-time operating status of the equipment (such as commissioning time, load variation, and telemetry signals), historical anomaly records (such as fault type, occurrence time, and handling results), and preset periodic strategies (such as daily, weekly, and monthly inspection plans). Each factor corresponds to a quantitative indicator, such as a load variation rate exceeding 10%, more than 3 consecutive anomalies, or the time since the last inspection exceeding a set threshold (such as 72 hours). These factors are combined and judged according to preset logical rules to form trigger conditions, determining whether to generate a new inspection task.

[0043] Furthermore, the system employs a priority queue algorithm to sort triggered inspection tasks. The priority evaluation model comprehensively considers parameters such as the safety importance of the equipment (e.g., whether it is core equipment), the scope of the anomaly's impact (e.g., whether it involves multiple lines or critical nodes), and the equipment's risk level (e.g., risk levels 1-5 based on historical failure frequencies), calculating a task priority index through a weighted scoring function. For example, the safety importance weight is 0.4, the impact scope weight is 0.3, and the risk level weight is 0.3, with a total score of 100 points. Tasks are queued according to their scores from highest to lowest.

[0044] At the application level, this step is widely applicable to the intelligent inspection and scheduling of asynchronous networked areas, new energy access nodes, and key power transmission and transformation equipment in centralized control centers. Especially in the identification of slowly changing anomalies such as ultra-low frequency oscillations and signal drift, dynamically adjusting the inspection frequency and task priority can significantly improve the system's ability to detect potential faults.

[0045] The technical effect of this step is that by constructing a multi-dimensional set of triggering factors and a priority scheduling mechanism, the intelligent generation of patrol tasks and the optimized allocation of resources are realized. This effectively solves the problems of response lag and scheduling conflicts in traditional fixed-cycle or simple rule triggering methods, and improves the real-time performance and reliability of the system in complex operating environments.

[0046] Furthermore, S1 includes:

[0047] S11. Based on at least three indicators from equipment commissioning time, load change range, scheduling strategy, and historical anomaly records, a weighted scoring algorithm is used to calculate the comprehensive risk index of the equipment, which serves as the quantitative basis for triggering factors.

[0048] Specifically, in some implementations, the present invention employs a weighted scoring algorithm to quantify the comprehensive risk index of the equipment, serving as a factor for triggering inspection tasks. This algorithm integrates at least three indicators from equipment commissioning time, load variation amplitude, scheduling strategy, and historical anomaly records. By setting reasonable weighting coefficients, it achieves a multi-dimensional assessment of equipment operational risk. Specifically, this step first extracts equipment operational data from systems such as SCADA and EMS, including the cumulative operating time since commissioning, the amplitude of load fluctuations (e.g., active power change rate ΔP / P0, reactive power change rate ΔQ / Q0), the frequency of scheduling strategy changes, and the number and severity of historical anomaly events. Regarding parameter settings, the load variation amplitude is typically calculated using a sliding window of 15 minutes or 1 hour; the scheduling strategy change frequency is based on the execution frequency of scheduling instructions; and historical anomaly records are weighted and scored according to anomaly type (e.g., tripping, protection action, signal loss). For example, a tripping event may be assigned 3 points, and a signal anomaly 1 point. Anomalies are classified according to the IEC 60870-5-104 standard.

[0049] Furthermore, the weighting coefficients of each indicator in the weighted scoring algorithm can be dynamically adjusted according to the equipment type, operating environment, and system stability requirements. For example, in asynchronous networking areas, the weights of scheduling strategies and load variation can be appropriately increased to enhance sensitivity to system disturbances.

[0050] This step is widely applicable in practical scenarios for intelligent monitoring of critical equipment in centralized control centers, such as hydropower stations and new energy power plants. By introducing a multi-parameter weighted scoring mechanism, the system can effectively identify potential risks in equipment operation, improve the accuracy and timeliness of anomaly warnings, and thus provide reliable data support for subsequent trend analysis and anomaly response. Its technical value lies in realizing the transformation from experience-driven to data-driven intelligent operation and maintenance, significantly reducing false alarm and false negative rates, and improving the safety and stability of system operation.

[0051] S12 compares the comprehensive risk index with the preset threshold range. If the threshold is exceeded, the device is automatically added to the high-priority queue, and the task execution interval is dynamically adjusted according to the queue length.

[0052] Specifically, in some implementations, comparing the comprehensive risk index with a preset threshold range and automatically adding the device to a high-priority queue based on the comparison result is a core logic control step in the graded response and report push module of this invention. This step, based on the comprehensive risk index calculated by a multi-parameter weighted scoring model, combined with preset graded threshold ranges (e.g., low risk: 0-30; medium risk: 31-70; high risk: 71-100), uses a real-time comparison mechanism to determine whether the device's current status has entered the high-risk category. If the comprehensive risk index exceeds the high-risk threshold (e.g., 71), the system will automatically identify the device as a high-priority task and add it to the high-priority queue to ensure priority scheduling of subsequent analysis and handling processes.

[0053] Furthermore, the comprehensive risk index calculation model used in this step is a multi-parameter linear weighted function. Its input variables include key operating parameters such as voltage offset, frequency fluctuation rate, signal duration, and load change rate. The weights of each parameter can be adaptively adjusted according to equipment type, operating environment, and historical fault data. For example, in asynchronous network areas, the weight of frequency fluctuation rate can be set to 0.4, voltage offset to 0.3, signal duration to 0.2, and load change rate to 0.1, to highlight the sensitivity to slow-changing disturbance characteristics.

[0054] Optionally, the system supports dynamically adjusting the task execution interval. This means that, based on the length of the high-priority queue and the system resource load, the system can automatically shorten or extend the task execution interval using a time-slice round-robin or preemptive scheduling strategy. For example, when the queue length exceeds a set threshold (e.g., 10 tasks), the system can adjust the task execution interval from the default 5 minutes to 1 minute to improve response speed and system real-time performance.

[0055] This step is widely applicable in practical scenarios for intelligent monitoring of large-scale power equipment in centralized control centers, especially in asynchronous grid areas with high renewable energy integration and frequent signal changes. It can effectively identify potential fault symptoms such as ultra-low frequency oscillations and voltage drift, and ensure timely handling of high-risk equipment through a priority scheduling mechanism. Its technical effect lies in significantly improving the timeliness and accuracy of anomaly response, reducing false alarm and missed alarm rates, thereby enhancing the operational stability and security of new power systems.

[0056] S2, collect real-time and historical operating data of the target device, and perform missing data completion, anomaly removal and time sequence alignment processing on the data to obtain a structured data stream.

[0057] Specifically, this step is the core processing stage of the "Real-time Data Acquisition and Preprocessing Module," and its technical implementation principle is based on the integrated acquisition and standardized processing mechanism of multi-source heterogeneous data. First, the system interfaces with mainstream power monitoring systems such as SCADA, EMS, and D5000 through standardized interfaces, employing power industry communication protocols such as IEC 61850 and DL / T 860 to achieve high-frequency sampling and synchronous acquisition of real-time telemetry data (such as voltage, current, active power, reactive power, frequency, etc.) and remote signaling status, control feedback, and protection signals from target equipment. In some implementations, the system supports data granularity from minutes to seconds, ensuring that the time resolution meets the needs of trend analysis and disturbance identification.

[0058] Secondly, the system calls historical database interfaces (such as Oracle, MySQL, or the time-series database InfluxDB) to obtain historical operational data of the target equipment within a specified inspection period, constructing a complete dataset containing both real-time and historical information. To improve data quality, the system employs a sliding window algorithm and rule-based anomaly detection mechanisms (such as the 3σ principle and threshold judgment) to perform anomaly removal and missing data completion. Missing data can be completed using methods such as linear interpolation and time series prediction (such as ARIMA and LSTM), while abnormal data is identified and replaced by setting upper and lower thresholds or using a dynamic baseline model to ensure the continuity and reliability of the data stream.

[0059] Furthermore, the system performs timing alignment operations, unifying the time base of different data sources through timestamp calibration and data resampling (such as using Resample or Interpolate methods), eliminating timing deviations caused by inconsistent sampling frequencies or transmission delays. The final output structured data stream contains metadata information such as time series, signal type, numerical range, and sampling frequency, conforming to power system data modeling standards such as IEC 61970 / 61968.

[0060] This step plays a crucial role in providing fundamental data support within the system, offering high-quality, standardized data input for subsequent trend analysis and anomaly identification. Its technical value lies in effectively improving the accuracy and robustness of signal analysis through multi-dimensional data fusion and preprocessing algorithms. It is particularly suitable for identifying slowly varying disturbance characteristics in asynchronously networked areas, laying a solid foundation for intelligent inspection and fault early warning.

[0061] Furthermore, S2 includes:

[0062] S21. A sliding window algorithm is used to perform linear interpolation or fill in missing data based on the mean of adjacent time periods to ensure data continuity.

[0063] Specifically, in the real-time data acquisition and preprocessing module of this invention, employing a sliding window algorithm to perform linear interpolation or mean-based imputation on missing data is a crucial step in ensuring data continuity and analytical accuracy. This step identifies missing points in time series data and uses a sliding window mechanism combined with linear interpolation or mean-based imputation methods to reasonably fill data gaps, thereby providing a high-quality data foundation for subsequent trend analysis and anomaly identification.

[0064] In some implementations, the sliding window algorithm performs local scanning of the target signal's time series in fixed time windows (such as 5 minutes, 10 minutes, or 30 minutes). When missing data points are detected within the window, the system performs interpolation based on the known data points within the window.

[0065] Optionally, when adjacent data points fluctuate significantly or experience abrupt changes, the system can employ a mean-based completion strategy within a sliding window. This strategy calculates the arithmetic mean of the valid data within the window by setting a window size (e.g., the first 3 and last 3 data points) and uses this mean as a replacement value for missing points. Furthermore, to improve data reliability, the system can combine data quality labels (e.g., "valid data," "suspicious data," and "invalid data" statuses in a SCADA system) for interpolation decisions, completing only points that are "suspicious data" or "invalid data" but have recoverable conditions.

[0066] In practical applications, this step is widely used in centralized control centers for signal acquisition and processing of equipment in hydropower stations, wind farms, and photovoltaic power stations. This is particularly relevant in asynchronous network areas where data loss is more pronounced due to lower signal sampling frequencies (e.g., 1 second / time or lower). This method effectively eliminates breakpoints caused by communication interruptions, sensor malfunctions, or abnormal data storage, ensuring the continuity and stability of trend analysis and disturbance identification algorithms.

[0067] The technical benefits of this step are that it significantly improves data integrity and continuity, reduces the misjudgment rate caused by missing data, and provides a reliable basis for subsequent anomaly identification and tiered response. Simultaneously, by flexibly selecting interpolation strategies, the system can adapt to the data repair needs of different signal characteristics, enhancing the robustness and practicality of the overall analysis.

[0068] S22 utilizes a rule-based anomaly detection algorithm, combined with a set threshold and a signal mutation identification model, to remove or replace outliers and align timestamps of multi-source data to construct a structured data stream with a unified time sequence.

[0069] Specifically, this step involves the collaborative application of rule-based anomaly detection algorithms and signal mutation identification models. The aim is to remove or replace outliers in multi-source power operation signals, align timestamps, and construct a structured data stream with a unified time sequence. In some implementations, this step first performs preliminary screening of the raw data by setting thresholds (such as upper and lower voltage limits, frequency fluctuation range, power mutation thresholds, etc.) to remove outliers exceeding reasonable ranges. For example, the normal range for voltage signals can be set to ±5% of the rated value, and for frequency signals, it can be set to 50Hz ±0.2Hz. Data points exceeding these ranges will be marked as anomaly candidates.

[0070] Furthermore, the system introduces a sliding window mechanism and abrupt change identification models (such as difference-based abrupt change detection or Hampel filter-based noise identification) to dynamically analyze continuous data sequences. The sliding window length can be set to 10–30 sampling points depending on the sampling frequency, with a window step size of 1 sampling point to ensure real-time response to signal changes. The abrupt change identification model calculates the deviation between the current point and the median within the window. If the deviation exceeds a set multiple threshold (such as 3 times the median absolute deviation MAD), it is identified as an outlier and removed or replaced by interpolation (such as linear interpolation, spline interpolation, or replacement based on the mean of historical data).

[0071] Meanwhile, the system performs timestamp alignment processing on multi-source data from heterogeneous systems such as SCADA, EMS, and D5000. In some implementations, time synchronization protocols (such as IEEE 1588v2 or NTP) are used to ensure that the time bases of each system are consistent, with errors controlled within ±1ms. For data points with inconsistent timestamps, the system aligns them through time interpolation or resampling methods (such as linear interpolation or nearest neighbor interpolation) to ensure that the data is comparable on a unified time axis.

[0072] This step plays a crucial role in data preprocessing in this invention, providing a high-quality, continuous, and structured time-series data foundation for subsequent trend analysis and anomaly identification. By removing noise and outliers, the robustness and accuracy of the analysis model can be effectively improved, especially in asynchronous network regions, where it is of great significance for identifying slowly varying disturbances such as ultra-low frequency oscillations.

[0073] S3. The sliding window is used to calculate the rate of change of key parameters, and a fast Fourier transform is performed on the structured data stream to extract frequency domain disturbance features and identify whether ultra-low frequency oscillations exist.

[0074] Specifically, this step uses a sliding window to calculate the rate of change of key parameters and performs a Fast Fourier Transform (FFT) on the structured data stream to extract frequency domain disturbance features in the 0.05–0.3 Hz range, thereby identifying the presence of ultra-low frequency oscillations. This method plays a core role in the trend analysis and anomaly identification module and is a key technical path for realizing the detection of slowly varying disturbance features in asynchronous networked areas.

[0075] At the technical implementation level, the sliding window algorithm uses a fixed-length time window (e.g., 10–30 minutes) to perform point-by-point calculations on key operating parameters (such as voltage, frequency, and active power), extracting their rate of change over time. The rate of change is typically calculated using the difference method: the difference between the current signal value and the signal value at the previous moment within the window is divided by the time interval Δt to obtain the change per unit time. This method effectively captures short-term trend changes in signals, providing a foundation for subsequent disturbance identification.

[0076] Furthermore, a Fast Fourier Transform (FFT) is performed on the structured data stream after sliding window processing to convert the time-domain signal into a frequency-domain signal. The sampling frequency of the FFT needs to be set according to the sampling period of the system signal, typically 1Hz to 10Hz, to meet the real-time requirements of the power system signal. By setting the frequency domain analysis range to 0.05 to 0.3Hz, the characteristic frequency of ultra-low frequency oscillations can be focused on. This frequency band is usually closely related to electromechanical coupling oscillations and system stability issues in asynchronous interconnected areas.

[0077] At the parameter level, the sliding window length, step size, sampling frequency, and frequency domain analysis range (0.05–0.3 Hz) are all adjustable parameters and need to be optimized based on the specific equipment type, signal characteristics, and system operating environment. For example, in a hydropower station scenario, the sliding window length can be set to 15 minutes and the step size to 5 minutes to balance computational efficiency and trend sensitivity. In FFT analysis, the spectral resolution should be no less than 0.01 Hz to ensure accurate identification of ultra-low frequency disturbances.

[0078] In application scenarios, this step is suitable for centralized control centers to monitor the operational status of key equipment such as asynchronous networked areas and new energy access nodes. By identifying oscillation characteristics in the 0.05–0.3 Hz frequency band, potential stability issues in the system can be effectively detected, such as inter-regional power oscillations and excitation system instability, thereby achieving early warning and risk control.

[0079] The technical advantage of this step lies in its enhanced ability to identify slowly varying disturbances through joint analysis of time-domain change rate and frequency-domain characteristics, particularly in the detection of ultra-low frequency oscillations. Compared to traditional rule-based anomaly identification methods, this solution offers higher sensitivity and accuracy, contributing to improved intelligent monitoring capabilities in centralized control centers and enabling in-depth perception and proactive early warning of the complex operating status of power systems.

[0080] S4 constructs a weighted scoring model based on multiple parameters such as voltage offset, frequency fluctuation rate, and signal duration, calculates anomaly scores and classifies anomaly levels, generates structured anomaly reports, and pushes information according to level.

[0081] Specifically, this step, "Anomaly Score Calculation and Classification Based on Multi-Parameter Weighted Scoring Model," is one of the core functions of the graded response and report push module in this invention. Its technical implementation principle involves quantifying the degree of anomaly in key operating parameters such as voltage offset, frequency fluctuation rate, and signal duration to construct a weighted scoring model, thereby achieving intelligent assessment and anomaly level classification of equipment operating status.

[0082] In some implementations, this step first standardizes the acquired voltage, frequency, and other signals by using a normalization formula to map the raw data to the [0,1] interval, eliminating dimensional differences. Then, the system calculates the anomaly score for each parameter according to a preset scoring function. For example, voltage deviation can be scored based on the percentage deviation from the rated voltage; with a threshold range of ±5%, the score increases for every 1% increase in deviation. Frequency volatility is calculated by the ratio of the difference between adjacent sampling points to the average value; the volatility threshold is set at ±0.2Hz, with any excess weighted exponentially. Signal duration is used to assess the persistence of the anomaly; if the anomaly persists for more than a set time window (e.g., 30 seconds or 5 minutes), the score weight increases accordingly.

[0083] Furthermore, the system introduces a weighted scoring mechanism, which determines the weight coefficients of each parameter through expert experience or machine learning methods. For example, the voltage offset has a weight of 0.4, the frequency fluctuation rate has a weight of 0.3, and the signal duration has a weight of 0.3.

[0084] This step is widely used in centralized control centers for real-time monitoring of complex operating scenarios such as asynchronously connected areas and new energy access nodes, and is particularly suitable for identifying slow-changing anomalies such as ultra-low frequency oscillations and slow voltage drift. Through automatic scoring and hierarchical push, the system can achieve immediate response to high-risk anomalies and summarize and analyze low-risk anomalies, thereby improving the decision-making efficiency of operators and the stability of the system. Its technical value lies in constructing a scalable and configurable anomaly assessment system, providing quantitative basis and response mechanism for the intelligent operation and maintenance of new power systems.

[0085] The intelligent identification and hierarchical response method for ultra-low frequency oscillations based on time-frequency fusion analysis in this invention can realize automatic inspection and intelligent identification of anomalies in power equipment operation signals, effectively improve the detection accuracy and response efficiency of slow-changing disturbances such as ultra-low frequency oscillations, and enhance the fault early warning and hierarchical handling capabilities of the centralized control center under the new power system.

[0086] Also includes:

[0087] S5. Perform pattern matching analysis on the identified ultra-low frequency oscillation signal, compare the signal with a preset disturbance template library to determine whether it belongs to a known disturbance type, and generate corresponding handling suggestions based on the matching results.

[0088] Specifically, in some implementations, pattern matching analysis of the identified ultra-low frequency oscillation signals is a key step in the trend analysis and anomaly identification module of this invention. Its core lies in comparing the signal with a pre-set disturbance template library to determine whether the signal belongs to a known disturbance type and generating corresponding handling suggestions accordingly. This step, based on signal processing and pattern recognition technology, combined with the frequency and time domain characteristics of typical disturbances in power systems, achieves accurate identification and intelligent response to slow-changing disturbances in asynchronous interconnected areas.

[0089] At the technical implementation level, this step first inputs the pre-processed signal (such as voltage, frequency, active power, etc.) into the pattern matching algorithm. The system's built-in perturbation template library contains time-frequency feature templates for various known perturbation types, such as ultra-low frequency oscillations, frequency drift, and slow voltage drops in the 0.05–0.3 Hz range. During the comparison process, the frequency domain features of the signal are extracted using Fast Fourier Transform (FFT) or Wavelet Transform, and its time-domain rate of change is calculated through a sliding window to construct a multi-dimensional feature vector. Subsequently, the system uses algorithms such as Euclidean distance, Dynamic Time Warping (DTW), or correlation coefficient to match the current signal features with the standard perturbation patterns in the template library and calculate the matching degree score.

[0090] At the parameter level, key parameters involved in the matching algorithm include: frequency resolution of frequency domain analysis (usually set to 0.01Hz), sliding window length (e.g., 30 seconds to 5 minutes), matching threshold (e.g., correlation coefficient R ≥ 0.85 or Euclidean distance ≤ set threshold), and disturbance duration (e.g., ≥ 2 minutes). These parameters are set according to power system operation standards (e.g., IEEE 112 standard, GB / T 19963-2011 "Technical Regulations for Wind Farm Access to Power Systems") to ensure the accuracy and reliability of the matching results.

[0091] At the application level, this step is widely used in centralized control centers for real-time monitoring of equipment in asynchronous network areas, new energy access nodes, and critical transmission lines. In actual operation, the system can automatically identify potential risk signals such as regional power oscillations, frequency instability, and slow voltage fluctuations, and generate handling suggestions based on the matching results, such as adjusting excitation control, activating damping control strategies, or notifying maintenance personnel to conduct on-site inspections.

[0092] In terms of technical effects, this step effectively improves the system's ability to identify slow-changing disturbances. Especially in the context of a high proportion of new energy integration, it can significantly enhance the early warning level for hidden faults such as ultra-low frequency oscillations, realize the automatic classification of abnormal signals and the generation of response strategies, thereby improving the safety and stability of power grid operation.

[0093] Example 2

[0094] This invention, based on a centralized control center platform, constructs a framework for an automatic signal inspection and anomaly analysis system, integrating multiple functional modules such as data acquisition, logic control, trend analysis, anomaly identification, level classification, report generation, and information push. The system can perform periodic or real-time inspections of the operating status of target objects, automatically analyze their historical and current operating data, identify potential anomalies, and implement strategies such as immediate alarms or tiered summary pushes based on the anomaly level to support the intelligent operation and maintenance needs of new power systems. Figure 2 and Figure 3 As shown.

[0095] In one embodiment of the present invention, a patrol task logic triggering module is included. This module is responsible for the automatic or manual generation and scheduling of patrol tasks. It includes functions such as equipment operating status identification, periodic strategy setting, trigger condition judgment, and a manual triggering interface. This allows for dynamic perception of equipment operating status and scheduling instructions, intelligent matching and filtering of patrol objects and time periods to obtain the optimal patrol task plan, ensuring the timeliness of patrol actions and the accuracy of targets. It includes:

[0096] Collect various operational information, including equipment commissioning time, load variation, historical anomaly records, and scheduling strategies, to comprehensively assess the operational status of the target equipment. This step provides the data foundation for the inspection mission and identifies whether the inspection conditions are met.

[0097] Based on the status analysis results and preset logic, it is determined whether this inspection is automatically triggered by a periodic plan or manually initiated by operators. Periodic inspections are usually triggered based on fixed time intervals or key events; manual triggering is achieved by inputting information such as the inspection object and time period through the interface.

[0098] Determine the scope of equipment to be inspected and the corresponding data time period, conduct a priority assessment of the current inspection tasks, including safety importance, scope of failure impact, equipment risk level, etc., and reasonably arrange the task execution order to avoid resource conflicts and scheduling congestion, and ensure that high-priority tasks are executed first.

[0099] The final confirmed objects, methods, times, and priorities are structured and output as an inspection task list, which is then distributed to the subsequent analysis module. This list supports API calls, task concurrency control, and result feedback identifiers to ensure closed-loop execution of inspection tasks.

[0100] In one embodiment of the present invention, a real-time data acquisition and preprocessing module is included. This module is responsible for real-time acquisition of operational data of the inspected object and retrieval of historical data. This includes reading signals such as voltage, current, active power, reactive power, frequency, remote signaling, and remote control; calling historical data interfaces; and data synchronization and preprocessing functions to build a complete data foundation. Operations such as missing data completion, anomaly removal, and time sequence alignment are performed on the acquired data to obtain a structured and reliable data stream, ensuring the accuracy and continuity of data in the analysis phase. This includes:

[0101] By interfacing with systems such as SCADA, EMS, and D5000, real-time telemetry data (voltage, current, active power, reactive power, frequency, etc.), remote signaling status, control feedback, and protection signals of the inspected objects are collected.

[0102] Call the database interface to retrieve historical operating data of the target device within a specified time period, supporting minute-level and second-level data granularity to ensure that the data covers the time window required for the inspection task.

[0103] Outlier and noise removal utilizes set thresholds, sliding windows, or rule-based algorithms to detect outliers and remove or replace them, reducing the impact of errors on the analysis results.

[0104] Data normalization:

[0105]

[0106] In the formula X g The normalized value; X i X a Let X be the minimum and maximum values ​​of the initial sequence, respectively, and let X be the initial value of the sequence.

[0107] Data relevance:

[0108]

[0109] In the formula, R is the correlation coefficient between the two variables; d i 2R represents the positional difference between two variables after sorting them from smallest to largest in the original data; n is the size of the original dataset; the larger R is, the higher the correlation between the variables.

[0110] In one embodiment of the present invention, a trend analysis and anomaly identification module is included. This module is responsible for trend modeling and anomaly pattern identification of the inspection data. It includes functions such as trend fitting, rate of change analysis, disturbance feature extraction, and ultra-low frequency oscillation detection to understand the changing trends of equipment operation, identify and judge potential fluctuations, drifts, and oscillations, obtain trend-based anomaly results, and ensure the system's early warning capability for slow-evolving faults. This includes:

[0111] Methods such as moving average, linear fitting, and polynomial fitting are used to extract the time variation trend of key parameters; indicators such as rate of change and incremental speed are calculated to preliminarily determine whether there is any deviation from the running trajectory.

[0112] By combining historical data templates, set rules, and disturbance libraries, typical abnormal behaviors such as continuous voltage drop, frequency oscillation, and remote signaling status jitter are identified and marked as suspected fault signals.

[0113] Fluctuation intensity:

[0114]

[0115] In the formula, Δx(t) is the rate of change of the trend upward / downward, x(t) is the signal value at the current moment, and Δt is the time interval.

[0116] By using wavelet analysis, FFT spectrum extraction and other techniques, we can identify whether there are oscillations in the target signal within the range of 0.05 to 0.3 Hz, determine whether they are ultra-low frequency oscillations, and adapt to the disturbance characteristics in asynchronous networking environments.

[0117] Ultra-low frequency oscillation:

[0118]

[0119] In the formula, X(f) is the frequency domain signal and x(n) is the time domain sample value.

[0120] By aggregating and comprehensively analyzing the abnormal behaviors of multiple signals and parameters of a single device over time, false alarms and missed alarms can be avoided, thereby improving the accuracy and completeness of anomaly identification.

[0121] In one embodiment of the present invention, a graded response and report push module is included. This module is responsible for classifying identified anomalies into grades, formulating response strategies, and generating and pushing reports. It includes functions such as risk level assessment, alarm logic setting, report template arrangement, and information distribution path configuration to achieve graded anomaly handling and accurate information push. For anomalies of different grades, it adopts immediate alarms or batch aggregation methods to obtain structured reports and actionable suggestions, ensuring that operators can quickly grasp the equipment status and handle issues efficiently.

[0122] Through the above technical solution, the present invention can realize periodic or real-time inspection of the operating status of the target object, automatically analyze its historical and current operating data, identify potential anomalies, and take strategies such as immediate alarm or hierarchical summary push according to the anomaly level to support the intelligent operation and maintenance needs under the new power system.

[0123] The beneficial effects of this invention are as follows:

[0124] (1) Automatic scheduling method for inspection tasks based on conditional priority queue: This invention constructs a set of task triggering factors (including equipment operating status, fault history, inspection cycle, etc.), sorts them using a priority queue algorithm, and executes inspection scheduling according to dynamic scheduling weights. This method is suitable for high-concurrency inspection scenarios, improves the real-time performance and accuracy of task distribution, and solves the problem of response lag in traditional rule matching methods.

[0125] (2) Trend Disturbance Identification Algorithm Combining Sliding Window and FFT: This invention uses a sliding window to extract the rate of change of operating parameters and introduces Fast Fourier Transform (FFT) to analyze and identify the frequency domain features of disturbances, thereby determining whether there are ultra-low frequency oscillations or potential abnormal trends. This algorithm improves the dual sensitivity of signal analysis in terms of time and frequency, enabling rapid detection of low-frequency disturbances in asynchronous network systems, improving fault prediction capabilities, and reducing false alarm rates.

[0126] (3) Anomaly Level Classification Method Based on Multi-Parameter Weighted Scoring: This invention constructs a multi-parameter scoring function, assigns weights to anomaly characteristic indicators such as voltage offset, frequency fluctuation rate, and signal duration, calculates anomaly scores, and classifies anomalies accordingly. This method has a clear structure, is easy to extend, and can adaptively adjust the scoring model according to the complexity of system operation, which helps to achieve accurate graded response and optimize alarm strategies and information push rhythm.

[0127] Example 3

[0128] To achieve the above embodiments, such as Figure 4 As shown, this embodiment also provides a signal automatic inspection and anomaly intelligent analysis device 10 based on a centralized control center, including:

[0129] The task scheduling module 100 is used to construct a set of patrol task triggering factors based on the equipment operating status, historical anomaly records and periodic strategies, and generate a priority queue according to the set of triggering factors to dynamically schedule patrol tasks.

[0130] The data acquisition and preprocessing module 200 is used to acquire real-time and historical operating data of the target device, and to perform missing data completion, anomaly removal and time sequence alignment processing on the data to obtain a structured data stream.

[0131] The time-frequency feature extraction module 300 is used to calculate the rate of change of key parameters using a sliding window, and to perform a fast Fourier transform on the structured data stream to extract frequency domain disturbance features and identify whether ultra-low frequency oscillations exist.

[0132] The anomaly scoring and grading module 400 is used to construct a weighted scoring model based on voltage offset, frequency fluctuation rate and signal duration, calculate anomaly scores and classify anomaly levels, generate structured anomaly reports and push information according to level.

[0133] Furthermore, the task scheduling module is also used for:

[0134] Based on at least three indicators from equipment commissioning time, load variation range, scheduling strategy, and historical anomaly records, a weighted scoring algorithm is used to calculate the comprehensive risk index of the equipment, which serves as the quantitative basis for triggering factors.

[0135] The comprehensive risk index is compared with a preset threshold range. If the threshold is exceeded, the device is automatically added to a high-priority queue, and the task execution interval is dynamically adjusted according to the queue length.

[0136] Furthermore, the data acquisition and preprocessing module is also used for:

[0137] The sliding window algorithm is used to perform linear interpolation or imputation based on the mean of adjacent time periods to ensure data continuity;

[0138] By using rule-based anomaly detection algorithms, combined with set thresholds and signal mutation identification models, outliers are eliminated or replaced, and timestamps of multi-source data are aligned to construct a structured data stream with a unified time series.

[0139] Furthermore, the anomaly scoring and grading module is also used for:

[0140] The correlation between voltage and frequency signals is calculated based on the Spearman correlation coefficient formula. If the correlation exceeds a preset threshold, the weight of the anomaly score is increased.

[0141] Anomalies are classified into three levels based on their anomaly scores: Level 1, Level 2, and Level 3. Level 1 anomalies trigger immediate alarms and are pushed to the on-duty personnel's terminals. Level 2 anomalies generate summary reports and are pushed out periodically. Level 3 anomalies are only recorded in the system log for subsequent analysis.

[0142] Furthermore, it also includes:

[0143] The pattern matching module is used to perform pattern matching analysis on the identified ultra-low frequency oscillation signal, compare the signal with a preset disturbance template library to determine whether it belongs to a known disturbance type, and generate corresponding handling suggestions based on the matching results.

[0144] The signal automatic inspection and anomaly intelligent analysis device based on the centralized control center of this invention can realize periodic or real-time inspection of the operating status of the target object, automatically analyze its historical and current operating data, identify potential anomalies, and take strategies such as immediate alarm or hierarchical summary push according to the anomaly level to support the intelligent operation and maintenance needs under the new power system.

[0145] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0146] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for automatic signal inspection and intelligent anomaly analysis based on a centralized control center, characterized in that, include: S1. Construct a set of patrol task triggering factors based on equipment operating status, historical anomaly records and periodic strategies, and generate a priority queue according to the set of triggering factors to dynamically schedule patrol tasks. This includes: using a priority evaluation model to comprehensively consider the safety importance of the equipment, the scope of the anomaly impact, and the equipment risk level, calculating the task priority index through a weighted scoring function, and enqueuing tasks in descending order of score. S2, collect real-time and historical operating data of the target device, perform missing data completion, anomaly removal and time sequence alignment processing on the data to obtain a structured data stream; S3, use a sliding window to calculate the rate of change of key parameters, and perform a fast Fourier transform on the structured data stream to extract frequency domain disturbance features and identify whether there is ultra-low frequency oscillation; S4 constructs a weighted scoring model based on voltage offset, frequency fluctuation rate, and signal duration, calculates the abnormal score of the equipment operating status, classifies the abnormality level, generates a structured abnormality report, and pushes information according to the level.

2. The method as described in claim 1, characterized in that, S1 includes: S11. Based on at least three indicators from equipment commissioning time, load change range, scheduling strategy and historical anomaly records, a weighted scoring algorithm is used to calculate the comprehensive risk index of the equipment, which serves as the quantitative basis for triggering factors. S12 compares the comprehensive risk index with the preset threshold range. If the threshold is exceeded, the device is automatically added to the high-priority queue, and the task execution interval is dynamically adjusted according to the queue length.

3. The method as described in claim 1, characterized in that, The S2 includes: S21, a sliding window algorithm is used to perform linear interpolation or fill in missing data based on the mean of adjacent time periods to ensure data continuity; S22 utilizes a rule-based anomaly detection algorithm, combined with a set threshold and a signal mutation identification model, to remove or replace outliers and align timestamps of multi-source data to construct a structured data stream with a unified time sequence.

4. The method as described in claim 1, characterized in that, The S4 includes: S41, calculate the correlation between voltage and frequency signals according to the Spearman correlation coefficient formula, and increase the weight of the anomaly score if the correlation exceeds the preset threshold. S42 classifies anomalies into Level 1, Level 2, and Level 3 based on the anomaly score of the equipment's operating status. Level 1 anomalies trigger an immediate alarm and are pushed to the terminal of the on-duty personnel. Level 2 anomalies generate a summary report and are pushed out periodically. Level 3 anomalies are only recorded in the system log for subsequent analysis.

5. The method as described in claim 1, characterized in that, Also includes: S5. Perform pattern matching analysis on the identified ultra-low frequency oscillation signal, compare the signal with a preset disturbance template library to determine whether it belongs to a known disturbance type, and generate corresponding handling suggestions based on the matching results.

6. A signal automatic inspection and anomaly intelligent analysis device based on a centralized control center, characterized in that, include: The task scheduling module is used to construct a set of patrol task triggering factors based on equipment operating status, historical anomaly records and periodic strategies, and generate a priority queue according to the set of triggering factors to dynamically schedule patrol tasks. This includes: using a priority evaluation model to comprehensively consider the safety importance of the equipment, the scope of the anomaly impact, and the equipment risk level, calculating the task priority index through a weighted scoring function, and enqueuing tasks according to their scores from high to low. The data acquisition and preprocessing module is used to acquire real-time and historical operating data of the target device, and to perform missing data completion, anomaly removal and time sequence alignment processing on the data to obtain a structured data stream. The time-frequency feature extraction module is used to calculate the rate of change of key parameters using a sliding window, and to perform a fast Fourier transform on the structured data stream to extract frequency domain perturbation features and identify whether ultra-low frequency oscillations exist. The anomaly scoring and grading module is used to construct a weighted scoring model based on voltage offset, frequency fluctuation rate and signal duration, calculate the anomaly score of the equipment operating status and classify the anomaly level, generate a structured anomaly report and push information according to the level.

7. The apparatus as claimed in claim 6, characterized in that, The task scheduling module is also used for: Based on at least three indicators from equipment commissioning time, load variation range, scheduling strategy, and historical anomaly records, a weighted scoring algorithm is used to calculate the comprehensive risk index of the equipment, which serves as the quantitative basis for triggering factors. The comprehensive risk index is compared with a preset threshold range. If the threshold is exceeded, the device is automatically added to a high-priority queue, and the task execution interval is dynamically adjusted according to the queue length.

8. The apparatus as claimed in claim 6, characterized in that, The data acquisition and preprocessing module is also used for: The sliding window algorithm is used to perform linear interpolation or imputation based on the mean of adjacent time periods to ensure data continuity; By using rule-based anomaly detection algorithms, combined with set thresholds and signal mutation identification models, outliers are eliminated or replaced, and timestamps of multi-source data are aligned to construct a structured data stream with a unified time series.

9. The apparatus as claimed in claim 6, characterized in that, The anomaly scoring and grading module is also used for: The correlation between voltage and frequency signals is calculated based on the Spearman correlation coefficient formula. If the correlation exceeds a preset threshold, the weight of the anomaly score is increased. Based on the anomaly score of the equipment's operating status, anomalies are divided into three levels: Level 1, Level 2, and Level 3. Level 1 anomalies trigger an immediate alarm and are pushed to the terminal of the on-duty personnel. Level 2 anomalies generate a summary report and are pushed out periodically. Level 3 anomalies are only recorded in the system log for subsequent analysis.

10. The apparatus as claimed in claim 6, characterized in that, Also includes: The pattern matching module is used to perform pattern matching analysis on the identified ultra-low frequency oscillation signal, compare the signal with a preset disturbance template library to determine whether it belongs to a known disturbance type, and generate corresponding handling suggestions based on the matching results.

Images