Modular software implementation of power oscillation fast discrimination algorithm for thermal power units

Abstract

The application discloses a software modularization implementation method of a thermal power unit power oscillation rapid discrimination algorithm, and particularly relates to the field of smart grids, and comprises the following steps: S1, original oscillation data is acquired to perform double-path parallel caching and cubic spline interpolation resampling, time base alignment and per-unit normalization processing are completed, and time axis synchronous standardized time series data sequences are generated; S2, a three-stage pipeline architecture is adopted to perform Hilbert transform demodulation on the standardized electrical side time series data, electrical oscillation instantaneous features are extracted, and oscillation event basic identification is completed; and S8, oscillation discrimination results and characteristic parameters are standardized data encapsulation and cyclic history caching, structured discrimination information is output, and oscillation event backtracking analysis is supported.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and more specifically, to a software modular implementation method for a fast power oscillation detection algorithm for thermal power units. Background Technology

[0002] With the continuous advancement of smart grid construction and the ongoing expansion of synchronous power grids, thermal power units, as the main frequency regulation and active power support sources in the power grid, directly affect the power angle stability and frequency stability of the power system. Power oscillation is a common dynamic stability problem in power system operation, especially low-frequency oscillations, which are prone to propagate and spread between regional power grids, and can even lead to grid disconnection accidents in severe cases. Currently, the identification of power oscillations in thermal power units mostly relies on electrical quantity data collected by synchronous phasor measurement devices, and uses time-domain waveform analysis and frequency-domain feature extraction algorithms to identify oscillation events. Some solutions adopt a software-based deployment approach to integrate the identification algorithm into the power grid monitoring system, providing oscillation early warning information for power grid operators.

[0003] Existing power oscillation discrimination technologies for thermal power units can identify basic oscillation events based on electrical quantity data, and have certain applicability in grid-side oscillation discrimination scenarios under symmetrical operating conditions. However, there are still several shortcomings in practical applications: First, the discrimination dimension is singular. Although it can rely on grid-side electrical quantity data for analysis, it lacks cross-verification by integrating mechanical operating parameters from the prime mover side. This easily leads to misjudging power fluctuations caused by mechanical disturbances such as turbine speed regulation system fluctuations and steam pressure pulsations as grid-side oscillations, making it difficult to guarantee discrimination accuracy and to achieve side-by-side location of the oscillation source. Second, the software architecture mostly adopts a serial processing mode, with data acquisition, feature calculation, and discrimination output executed sequentially. This results in long data transmission and processing delays, and the sliding window repeatedly calculates all data. The method consumes a lot of computing resources, making it difficult to meet the real-time requirements of smart grids for rapid power oscillation identification and low resource consumption. Third, the identification function is limited to the identification of the presence or absence of oscillation events, making it difficult to quantify and analyze the deep attributes of oscillations such as electromechanical coupling strength, transmission characteristics, and damping level, thus failing to support the refined requirements of grid stability analysis. Fourth, it does not adequately consider the impact of three-phase asymmetrical operating conditions. When the grid experiences non-full-phase operation or asymmetrical faults, negative sequence components will be superimposed on the oscillation signal, causing deviations in the identification results of oscillation intensity and frequency, thus limiting the applicable scenarios. Summary of the Invention

[0004] To overcome the aforementioned deficiencies in existing technologies, this invention provides a software modular implementation method for a fast power oscillation discrimination algorithm for thermal power units. By constructing a parallel bus-style software module architecture, coupled with a software implementation method that utilizes dual-path ring buffering, pipelined feature calculation, and sliding window incremental recursion, and integrating multi-source operating parameters from both the electrical and mechanical sides for coupled analysis of oscillation characteristics, this method achieves rapid and accurate discrimination of oscillation events while simultaneously locating the oscillation source and subdividing the disturbance type. It can also analyze the coupling strength, damping characteristics, and dominant type of the oscillation, and possesses the ability to correct results under asymmetric operating conditions, thus meeting the stable operation monitoring requirements of smart grids and solving the problems mentioned in the background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a software modular implementation method for a fast power oscillation discrimination algorithm for thermal power units, comprising: S1. Obtain the raw oscillation data, perform dual-path parallel buffering and cubic spline interpolation resampling, complete time base alignment and per-unit value normalization processing, and generate a standardized time-series data sequence synchronized with the time axis. S2. A three-stage pipeline architecture is used to perform Hilbert transform demodulation on standardized electrical side time-series data, extract instantaneous features of electrical oscillations, and complete the basic identification of oscillation events. S3. A Hilbert transform demodulation is performed on the standardized mechanical side timing data using a co-source clock triggering mechanism to extract the instantaneous oscillation features of the mechanical side that are aligned with the electrical side timing. S4. The instantaneous characteristics of bilateral oscillations are calculated in the time domain and fused with frequency consistency by using the sliding window incremental recursive algorithm to generate a continuously quantized electromechanical coupling index. S5. Using electromechanical coupling degree as the weight, phase energy transfer factor and amplitude dominance are integrated to carry out hierarchical progressive oscillation source localization, and the prime mover side disturbance type is subdivided through condition triggering mechanism. S6. Based on the numerical range of electromechanical coupling, the corresponding analytical unit is adaptively loaded to perform multi-dimensional characteristic analysis of the oscillation; S7. Based on the positive and negative sequence voltage oscillation amplitude ratio, the asymmetric component of the active oscillation amplitude is stripped, and the synchronous correction of the link amplitude data is achieved through a double buffer reverse broadcast mechanism. S8. Standardize the data encapsulation and cyclic history cache of the oscillation discrimination results and characteristic parameters, output structured discrimination information and support the backtracking analysis of oscillation events.

[0006] Preferably, the raw oscillation data includes electrical-side data and mechanical-side data; the electrical-side data originates from a synchronous phasor measurement device and includes the instantaneous low-frequency oscillation values ​​of the generator's three-phase total active power, the generator terminal fundamental positive-sequence voltage amplitude, the generator terminal fundamental negative-sequence voltage amplitude, and the generator terminal frequency; the mechanical-side data originates from a distributed control system and includes the instantaneous low-frequency oscillation values ​​of the steam pressure after the turbine regulating stage, the turbine speed, and the turbine governor valve opening index. Let the instantaneous value of the low-frequency oscillation be defined. The dual-channel parallel buffer and cubic spline interpolation resampling are implemented using a dual-channel independent ring buffer architecture, corresponding to the electrical side data channel and the mechanical side data channel respectively. The two data channels can be written to the buffer in parallel without blocking or interfering with each other. The interpolation resampling uses the GPS synchronization time scale as a unified reference and performs cubic spline interpolation on the two types of data with different sampling rates to unify the data with different sampling rates into a sequence of synchronization points with the same time interval. The time base alignment eliminates the time base deviation between the two types of data through interpolation resampling. The unified synchronization sampling interval is denoted as . The per-unit normalization is based on the rated values ​​of each parameter, and all time-series data are converted to per-unit values ​​to eliminate the influence of different dimensions on subsequent coupled calculations.

[0007] Preferably, the three-stage pipeline architecture is sequentially divided into a data framing unit, an orthogonal transformation unit, and a feature calculation unit. These three units operate sequentially point-to-point. Under steady-state continuous operation, the overall processing delay of a single sampling point does not exceed one synchronous sampling interval. The method for demodulating the standardized electrical side time-series data using Hilbert transform is as follows: based on the Hilbert transform, an analytical signal is constructed for each electrical oscillation signal to complete the demodulation and extraction of instantaneous features; the instantaneous value of the low-frequency oscillation of the generator's three-phase total active power at time k is... Performing a Hilbert transform yields the active oscillation Hilbert quadrature components. This leads to the construction of an active oscillation analytical signal: Where j is the imaginary unit; the instantaneous characteristics of the electrical oscillation include the instantaneous amplitude, instantaneous phase, and instantaneous frequency characteristics of four types of signals: active oscillation, positive-sequence voltage oscillation, negative-sequence voltage oscillation, and frequency oscillation; the instantaneous amplitude of the active oscillation is obtained based on the analytical signal. The instantaneous phase of the active oscillation is obtained using the four-quadrant arctangent function. Further obtain the instantaneous oscillation frequency of the active oscillation. ,in For synchronous sampling interval, The instantaneous phase of the active power oscillation at time k-1; simultaneously, the instantaneous low-frequency oscillation value of the generator terminal fundamental positive sequence voltage amplitude at time k. By performing a Hilbert transform, the corresponding orthogonal components are obtained, and a positive-sequence voltage oscillation analytic signal is constructed, thereby obtaining the instantaneous amplitude of the positive-sequence voltage oscillation. Instantaneous phase of positive sequence voltage oscillation ; the instantaneous value of the low-frequency oscillation of the fundamental negative sequence voltage amplitude at the generator terminals at time k. By performing a Hilbert transform, the corresponding quadrature components are obtained, and a negative-sequence voltage oscillation analytic signal is constructed, thereby obtaining the instantaneous amplitude of the negative-sequence voltage oscillation. ; for the instantaneous value of low-frequency oscillation at generator terminal frequency at time k. Perform a Hilbert transform to obtain the corresponding orthogonal components and construct a frequency oscillation analytic signal, thereby obtaining the instantaneous phase of the frequency oscillation. The oscillation event identification is based on the instantaneous amplitude of the active oscillation to generate a discrimination result. When the instantaneous amplitude continuously exceeds the set oscillation start amplitude threshold, it is determined that a power oscillation event has occurred.

[0008] Preferably, the method for demodulating the standardized mechanical side time-series data using Hilbert transform is as follows: Hilbert transform, analytical signal construction, and feature quantity calculation are performed sequentially on each mechanical oscillation signal to extract the corresponding instantaneous features; the low-frequency oscillation instantaneous value of the steam pressure after the turbine regulating stage at time k is then analyzed. Performing a Hilbert transform yields the Hilbert orthogonal components of the steam pressure oscillation. Constructing an analytical signal of vapor pressure oscillation The instantaneous amplitude of the steam pressure oscillation is obtained based on the analytical signal. Instantaneous phase of steam pressure oscillation The instantaneous oscillation frequency of the steam pressure oscillation is obtained by measuring the phase difference between adjacent sampling points. ,in For synchronous sampling interval, The instantaneous phase of the steam pressure oscillation at time k-1; the instantaneous value of the low-frequency oscillation of the turbine speed at time k. A Hilbert transform is performed to obtain the corresponding orthogonal components, and an analytical signal of the rotational speed oscillation is constructed to obtain the instantaneous phase of the rotational speed oscillation. The instantaneous value of low-frequency oscillation of the turbine governor valve opening command at time k. Perform a Hilbert transform to obtain the corresponding orthogonal components and construct the valve opening oscillation analytical signal, thereby obtaining the instantaneous phase of the valve opening oscillation. .

[0009] Preferably, the core mechanism of the sliding window incremental recursive algorithm is to pre-store and maintain five statistical cumulative quantities within the sliding window, namely, the cumulative sum of active power oscillation amplitude, the cumulative sum of steam pressure oscillation amplitude, the cumulative sum of cross-multiplication of the two types of amplitudes, the cumulative sum of the squares of active power amplitude, and the cumulative sum of the squares of steam pressure amplitude. When a new sampling point enters the analysis window, only the corresponding contribution value of the earliest exiting sampling point is removed from the five cumulative quantities, and the corresponding contribution value of the latest entering sampling point is superimposed. There is no need to repeatedly perform summation operations on all sampling points in the entire window, reducing the single-step computation complexity from linear to constant order. The time-domain correlation calculation is carried out based on the incrementally updated cumulative quantities. Since the input signal is a low-frequency oscillation component extracted by bandpass filtering, the DC component is approximately zero, which satisfies the premise for calculating the time-domain coherence coefficient. The time-domain coherence coefficient between the instantaneous amplitude of active power oscillation and the instantaneous amplitude of steam pressure oscillation within the current window is calculated. The frequency consistency fusion is used to calculate the instantaneous frequency deviation between the instantaneous oscillation frequency of the active power oscillation and the instantaneous oscillation frequency of the steam pressure oscillation at the current moment. ; and combined with the time-domain coherence coefficient and instantaneous frequency deviation, the computer electrical coupling degree is fused. .

[0010] Preferably, the phase energy transfer factor is calculated based on the two-sided phase difference, and the amplitude dominance is calculated based on the relative strength of the two-sided amplitudes; wherein, before calculating the phase difference, all instantaneous phase sequences are first processed. Phase dewinding: Compare the phase values ​​of adjacent sampling points point by point. If the absolute value of the phase difference between the current sampling point and the previous sampling point is greater than... Then the phase of the current sampling point will be added or subtracted accordingly. This makes the absolute value of the adjacent phase difference less than To eliminate numerical abrupt changes caused by periodic phase transitions; the above procedures are performed on the instantaneous phases of active power oscillation, steam pressure oscillation, frequency oscillation, and speed oscillation, respectively. After phase dewinding, the energy transfer direction factor is obtained. Simultaneously obtain the dominance of unilateral amplitude. The hierarchical progressive oscillation source positioning adopts a three-layer computing unit cascaded deployment in a pipeline. After the first layer energy transfer direction factor calculation is completed, the second layer single-sided amplitude dominance calculation is automatically triggered. After the second layer calculation is completed, the third layer comprehensive bias coefficient fusion calculation is automatically triggered. Electromechanical coupling degree is used as the basis for this calculation. Using weights, the energy transfer direction factor and the dominance of one-sided amplitude are combined to obtain the comprehensive oscillation source bias coefficient. The method for subdividing the disturbance type on the prime mover side is as follows: a conditional triggering mechanism is adopted for oscillation scenarios dominated by the prime mover side. The disturbance type subdivision operator is only loaded and executed when the comprehensive oscillation source bias coefficient is negative. In the non-triggered state, the subdivision operator is in a dormant state to save computing power. After triggering, the disturbance subdivision coefficient on the prime mover side is analyzed. .

[0011] Preferably, the numerical range of the electromechanical coupling degree is divided into a low-value range and a high-value range, which serve as the basis for determining adaptive loading. The adaptive loading corresponding parsing unit refers to the mapping relationship between the electromechanical coupling degree and the parsing dimension pre-established by the parsing engine. When the electromechanical coupling degree is in the low-value range, the engine only loads the oscillation damping ratio calculation unit and the electrical oscillation dominance coefficient calculation unit. When the electromechanical coupling degree is in the high-value range, the engine automatically loads the electromechanical oscillation effective coupling complex gain calculation unit. The multi-dimensional characteristic parsing is used to obtain the electromechanical oscillation effective coupling complex gain, oscillation damping ratio, and electrical oscillation dominance coefficient: when the electromechanical coupling degree is in the high-value range, the coupling complex gain calculation unit is loaded, and the electromechanical oscillation effective coupling complex gain is calculated. ,in This represents the maximum instantaneous amplitude of the active power oscillation within the current sliding analysis window. This represents the maximum instantaneous amplitude of the steam pressure oscillation within the current sliding analysis window. The value of j represents the average difference between the instantaneous phase of the active oscillation and the instantaneous phase of the steam pressure oscillation within the current sliding analysis window, where j is the imaginary unit. Based on the oscillation amplitude variation within the continuous sliding window, the oscillation damping ratio is calculated recursively. ,in Let W be the instantaneous amplitude of the active oscillation after an interval of W sampling points. The time interval between two amplitude sampling points, where For synchronous sampling interval, The average value of the instantaneous oscillation frequency of the active power oscillation within the current sliding analysis window; calculate the dominant electrical oscillation coefficient. Identify the dominant type of oscillation: .

[0012] Preferably, the asymmetric component stripping is based on the negative sequence voltage ratio to generate an asymmetric correction coefficient, and interference components are removed from the active power oscillation amplitude; the asymmetric correction coefficient at the current time is calculated. Based on the asymmetric correction coefficient, the asymmetric component of the instantaneous amplitude of the active power oscillation is removed to obtain the corrected instantaneous amplitude of the active power oscillation. The dual-buffer reverse broadcast mechanism refers to the broadcasting of the corrected data through the reverse dedicated channel of the internal parallel bus. All upstream computing modules complete the in-situ replacement of the corresponding data according to the sampling point sequence number. The module has a built-in dual buffer. During the correction process, the new amplitude data is first written to the backup buffer. After the timing of the full sampling point verification is consistent, it is switched to the main working buffer at one time.

[0013] Preferably, the oscillation discrimination results and characteristic parameters include oscillation event discrimination results, comprehensive oscillation source bias coefficient, prime mover-side disturbance subdivision coefficient, electromechanical coupling degree, effective coupling complex gain of electromechanical oscillation, oscillation damping ratio, electrical oscillation dominance coefficient, asymmetry correction coefficient, and active oscillation dominance frequency, input to the result output module; the standardized data encapsulation encapsulates all information in a structured manner according to the general communication protocol format of the power system; the cyclic historical data cache adopts a built-in cyclic historical database, using a first-in-first-out mechanism to cache original sampling data, intermediate feature data, and final result data, supporting the retrospective playback and offline analysis of oscillation events; the structured discrimination information is output to the power grid stability monitoring system through a standard communication interface; the oscillation event retrospective analysis is realized based on historical cached data and can be used for offline fault review and algorithm iteration optimization.

[0014] The technical effects and advantages of this invention are as follows: 1. This invention integrates multi-source operating parameters from both the electrical and mechanical sides, and uses a hierarchical fusion method combining energy transfer direction factors and single-side amplitude dominance to quantify and calculate the comprehensive oscillation source bias coefficient. This changes the traditional discrimination mode that relies solely on electrical quantities, effectively distinguishing between mechanical-side disturbances and grid-side oscillations, and avoiding misjudgment problems caused by single-side data discrimination. At the same time, it can further subdivide the disturbance sources on the prime mover side, significantly improving the accuracy of power oscillation discrimination and providing source-side location support for power system stability analysis. 2. This invention utilizes a parallel bus-based modular software architecture, combined with dual-path independent ring buffer access, three-stage pipeline feature extraction, and sliding window incremental recursive statistical calculation techniques to achieve a real-time computing link with low latency and low computational consumption. This solves the problems of long data transmission and processing latency, high computational resource consumption, and difficulty in meeting real-time requirements in the serial processing architecture of the background technology. It reduces the complexity of single-step statistical calculation from linear to constant order, and the processing latency of a single sampling point in steady state does not exceed one synchronous sampling interval, which is suitable for the real-time requirements of power grid operation monitoring scenarios. At the same time, the modular architecture has good scalability, which facilitates the subsequent upgrade and iteration of power grid stability monitoring functions. 3. This invention, through an adaptive analytical engine based on electromechanical coupling degree mapping and a dynamic module loading mechanism, obtains multi-dimensional quantitative analytical results of the effective coupling complex gain of electromechanical oscillations, oscillation damping ratio, and electrical oscillation dominance coefficient. This solves the problem in the background technology that although the presence or absence of oscillation events can be identified, it is difficult to quantify and analyze the deep attributes of oscillation electromechanical coupling strength, transmission characteristics, and damping level, and it is difficult to support refined stability analysis. It realizes the effective characterization of oscillation transmission characteristics, attenuation trend, and dominance type, and can provide refined data support for power system frequency stability assessment and low-frequency oscillation suppression strategy formulation. 4. This invention uses an asymmetric component stripping method based on the positive and negative sequence voltage amplitude ratio, combined with a full-link synchronous correction mechanism of dual-buffer atomic switching and reverse broadcasting, to obtain calibrated active power oscillation amplitude consistent with the full-link under three-phase asymmetric operating conditions. This solves the problems in the background technology of insufficient consideration of the impact of three-phase asymmetric operating conditions, deviation in oscillation intensity and frequency discrimination results caused by the superposition of negative sequence components, and limited applicable scenarios. It also solves the interference of non-full-phase operation and asymmetric faults on oscillation discrimination, broadens the applicability of power oscillation discrimination algorithm under complex power grid operating conditions, and improves the reliability of power grid power quality monitoring and stability analysis. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0016] 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.

[0017] The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units relies on a software system that includes a multi-source data access module, an electrical oscillation characteristic calculation module, a mechanical oscillation characteristic calculation module, an electro-mechanical coupling degree calculation module, an oscillation source location module, an oscillation characteristic analysis module, an asymmetry correction module, and a result output module. Each module interacts with data through an internal parallel data bus. The electrical oscillation characteristic calculation module and the mechanical oscillation characteristic calculation module are deployed in parallel and operate synchronously. Specifically, the method includes the following steps: As attached Figure 1 The software modular implementation method of the fast power oscillation detection algorithm for thermal power units shown includes: S1. Obtain the raw oscillation data, perform dual-path parallel buffering and cubic spline interpolation resampling, complete time base alignment and per-unit value normalization processing, and generate a standardized time-series data sequence synchronized with the time axis. In this embodiment, it should be noted that: the original oscillation data includes electrical-side data and mechanical-side data; the electrical-side data comes from a synchronous phasor measurement device and includes the instantaneous value of low-frequency oscillation of the generator's three-phase total active power (unit: MW), the instantaneous value of low-frequency oscillation of the generator terminal fundamental positive-sequence voltage amplitude (unit: kV), the instantaneous value of low-frequency oscillation of the generator terminal fundamental negative-sequence voltage amplitude (unit: kV), and the instantaneous value of low-frequency oscillation of the generator terminal frequency (unit: Hz), with a sampling rate of 100Hz; the mechanical-side data comes from a distributed control system and includes the instantaneous value of low-frequency oscillation of the steam pressure after the turbine regulating stage (unit: MPa), the instantaneous value of low-frequency oscillation of the turbine speed. The time value (unit: r / min) and the instantaneous value of low-frequency oscillation of the turbine governor valve opening command (unit: %) are sampled at a rate of 50Hz. The dual-channel parallel buffer and cubic spline interpolation resampling are implemented using a dual-channel independent ring buffer architecture, corresponding to the electrical side data channel and the mechanical side data channel respectively. The two data channels can be written to the buffer in parallel without blocking and without interference. The interpolation resampling uses the GPS synchronization time scale as a unified reference and performs cubic spline interpolation on the two types of data with different sampling rates to unify the data with different sampling rates into a synchronization point sequence with the same time interval. The time base alignment eliminates the time base deviation of the two types of data through interpolation resampling. The unified synchronization sampling interval is denoted as . The unit is seconds, which serves as the unified time reference for all subsequent time-domain difference and sliding window time span calculations. The per-unit value normalization is based on the rated value of each parameter, and all time-series data are converted to per-unit values ​​to eliminate the influence of different dimensions on subsequent coupled calculations. The standardized time-series data sequence includes two types: standardized electrical data sequence and standardized mechanical data sequence. All sampling points in the sequence are identified by a unified sequence number, and all time-series data are synchronously output to the internal parallel data bus point by point according to the sequence number.

[0018] In detail: During the data preprocessing stage of online monitoring of power oscillation in thermal power units, the time base synchronization scheduling unit and the dual-path independent ring buffer in the multi-source data access module jointly perform data synchronization processing. When the original operating data from the electrical side and the mechanical side are uploaded in real time according to their respective sampling rates, the module receives the data from both sides through the two independent ring buffers to achieve non-blocking parallel writing without interference. Subsequently, using the GPS synchronization time scale as a unified reference, cubic spline interpolation resampling is performed on the two types of data with different sampling rates to unify the data from both sides to the same sampling interval and complete the time base alignment. Then, using the rated values ​​of each parameter as a reference, all time series data are normalized by per-unit value to eliminate the influence of different dimensions on subsequent calculations. Finally, a standardized time series data sequence with time axis alignment is generated and output to the internal parallel data bus point by point according to the globally unified sampling point sequence number.

[0019] S2. A three-stage pipeline architecture is used to perform Hilbert transform demodulation on standardized electrical side time-series data, extract instantaneous features of electrical oscillations, and complete the basic identification of oscillation events. In this embodiment, it should be noted that: the three-level pipeline architecture is sequentially divided into a data framing unit, an orthogonal transformation unit, and a feature calculation unit. These three units operate sequentially point-to-point. Under steady-state continuous operation, the overall processing delay of a single sampling point does not exceed one synchronous sampling interval. The standardized electrical-side time-series data is provided by the electrical-side portion of the standardized time-series data sequence generated by S1, and is input point-to-point into the electrical oscillation feature calculation module according to the global synchronous sampling point number. The Hilbert transform demodulation of the standardized electrical-side time-series data involves constructing analytical signals based on the Hilbert transform for each electrical oscillation signal, completing the demodulation and extraction of instantaneous features. The instantaneous value of the low-frequency oscillation of the generator's three-phase total active power at time k is... Performing a Hilbert transform yields the active oscillation Hilbert quadrature components. This leads to the construction of an active oscillation analytical signal: Where j is the imaginary unit; the instantaneous characteristics of the electrical oscillation include the instantaneous amplitude, instantaneous phase, and instantaneous frequency characteristics of four types of signals: active oscillation, positive-sequence voltage oscillation, negative-sequence voltage oscillation, and frequency oscillation; the instantaneous amplitude of the active oscillation is obtained based on the analytical signal. The instantaneous phase of the active oscillation is obtained using the four-quadrant arctangent function. ,in This is a four-quadrant arctangent function. With the quadrature component as the vertical axis input and the original signal as the horizontal axis input, it can directly output the phase value within a complete cycle, eliminating the quadrant ambiguity of single-quadrant arctangent; further, it can obtain the instantaneous oscillation frequency of the active oscillation. ,in For synchronous sampling interval, The instantaneous phase of the active power oscillation at time k-1; simultaneously, the instantaneous low-frequency oscillation value of the generator terminal fundamental positive sequence voltage amplitude at time k. By performing a Hilbert transform, the corresponding orthogonal components are obtained, and a positive-sequence voltage oscillation analytic signal is constructed, thereby obtaining the instantaneous amplitude of the positive-sequence voltage oscillation. Instantaneous phase of positive sequence voltage oscillation ; the instantaneous value of the low-frequency oscillation of the fundamental negative sequence voltage amplitude at the generator terminals at time k. By performing a Hilbert transform, the corresponding quadrature components are obtained, and a negative-sequence voltage oscillation analytic signal is constructed, thereby obtaining the instantaneous amplitude of the negative-sequence voltage oscillation. ; for the instantaneous value of low-frequency oscillation at generator terminal frequency at time k. Perform a Hilbert transform to obtain the corresponding orthogonal components and construct a frequency oscillation analytic signal, thereby obtaining the instantaneous phase of the frequency oscillation. The oscillation event identification is based on the instantaneous amplitude of the active oscillation to generate a discrimination result. When the instantaneous amplitude continuously exceeds the set oscillation start amplitude threshold, it is determined that a power oscillation event has occurred. All electrical side oscillation instantaneous features extracted in this step are broadcast to downstream modules in real time through a parallel data bus.

[0020] In detail: In the electrical side oscillation feature extraction stage, the feature calculation is performed by the three-level pipeline unit of the electrical oscillation feature calculation module. This stage is carried out synchronously and in parallel with the mechanical side feature extraction. When the standardized electrical side time-series data is input into the module point by point, the module performs point-by-point pipeline operation according to the three-level architecture of data framing, orthogonal transformation, and feature calculation. Under steady-state continuous operation, the overall processing delay of a single sampling point does not exceed one sampling interval. During the processing, Hilbert transform is first performed on each electrical oscillation signal to construct an analytical signal. Then, the instantaneous amplitude, instantaneous phase, and instantaneous oscillation frequency of the corresponding signal are demodulated. At the same time, the basic identification of oscillation events is completed based on the instantaneous amplitude of active power oscillation. When the amplitude continuously exceeds the set start threshold, a power oscillation event is determined to have occurred. Finally, all extracted electrical side oscillation instantaneous features and basic discrimination results are broadcast to downstream modules in real time through a parallel data bus.

[0021] S3. A Hilbert transform demodulation is performed on the standardized mechanical side timing data using a co-source clock triggering mechanism to extract the instantaneous oscillation features of the mechanical side that are aligned with the electrical side timing. In this embodiment, it should be noted that: the co-source clock triggering mechanism refers to the mechanical oscillation feature calculation module and the electrical side module using the same clock source to trigger sampling calculation, ensuring that the sampling point numbers of the feature outputs on both sides are strictly aligned, eliminating the timing deviation of the calculations on both sides; the standardized mechanical side timing data is provided by the mechanical side part of the standardized timing data sequence generated by S1, and is input into the mechanical oscillation feature calculation module point by point according to the global synchronous sampling point number; the Hilbert transform demodulation of the standardized mechanical side timing data: the Hilbert transform, analytical signal construction and feature quantity calculation are performed on each mechanical oscillation signal in sequence to extract the corresponding instantaneous features; the low-frequency oscillation instantaneous value of the steam pressure after the turbine regulating stage at time k is... Performing a Hilbert transform yields the Hilbert orthogonal components of the steam pressure oscillation. Constructing an analytical signal of vapor pressure oscillation The instantaneous amplitude of the steam pressure oscillation is obtained based on the analytical signal. Instantaneous phase of steam pressure oscillation The instantaneous oscillation frequency of the steam pressure oscillation is obtained by measuring the phase difference between adjacent sampling points. ,in For synchronous sampling interval, The instantaneous phase of the steam pressure oscillation at time k-1; the instantaneous value of the low-frequency oscillation of the turbine speed at time k. A Hilbert transform is performed to obtain the corresponding orthogonal components, and an analytical signal of the rotational speed oscillation is constructed to obtain the instantaneous phase of the rotational speed oscillation. The instantaneous value of low-frequency oscillation of the turbine governor valve opening command at time k. Perform a Hilbert transform to obtain the corresponding orthogonal components and construct the valve opening oscillation analytical signal, thereby obtaining the instantaneous phase of the valve opening oscillation. The instantaneous characteristics of mechanical-side oscillation include the instantaneous amplitude and phase data of steam pressure oscillation, rotational speed oscillation, and valve opening oscillation. All characteristics maintain a one-to-one correspondence with the sampling point numbers of electrical-side characteristics and are broadcast to downstream modules in real time via a parallel data bus.

[0022] In detail: In the mechanical side oscillation feature extraction stage, the mechanical oscillation feature calculation module performs feature calculation, which is deployed synchronously and in parallel with the electrical side feature extraction. The module uses a clock-triggered sampling calculation that is completely identical to that of the electrical side to ensure that the sampling point numbers of the feature outputs from both sides are strictly aligned. When the standardized mechanical side timing data is input into the module point by point, Hilbert transform, analytical signal construction, and feature quantity calculation are performed sequentially on each mechanical oscillation signal to extract the instantaneous amplitude, instantaneous phase, and instantaneous oscillation frequency of the corresponding signal. Finally, instantaneous mechanical side oscillation feature data that is strictly aligned with the electrical side timing is generated and broadcast to downstream modules in real time through a parallel data bus.

[0023] S4. The instantaneous characteristics of bilateral oscillations are calculated in the time domain and fused with frequency consistency by using the sliding window incremental recursive algorithm to generate a continuously quantized electromechanical coupling index. In this embodiment, it should be noted that: the instantaneous features of the dual-sided oscillation include the instantaneous features of the electrical side oscillation output by S2 and the instantaneous features of the mechanical side oscillation output by S3, which are synchronously input into the electromechanical coupling degree calculation module; the core mechanism of the sliding window incremental recursive algorithm is to pre-store and maintain five statistical cumulative quantities within the sliding window, namely, the cumulative sum of active power oscillation amplitude, the cumulative sum of steam pressure oscillation amplitude, the cumulative sum of the cross-multiplication of the two types of amplitudes, the cumulative sum of the square of active power amplitude, and the cumulative sum of the square of steam pressure amplitude; when a new sampling point enters the analysis window, only the corresponding contribution value of the earliest exiting sampling point is removed from the five cumulative quantities, and the corresponding contribution value of the latest entering sampling point is superimposed, without repeating the summation operation for all sampling points in the entire window, reducing the single-step calculation complexity from linear to constant order; at the same time, a sliding analysis window with a length of W sampling points is set, where W is a positive integer representing the number of sampling points contained in the sliding window; five cumulative statistics of the sliding window are predefined: the cumulative sum of active power oscillation amplitude. Cumulative sum of steam pressure oscillation amplitude The sum of the two types of amplitude cross-products ; cumulative sum of squared active oscillation amplitudes ; Cumulative sum of squared amplitudes of steam pressure oscillations ;in Let k be the sampling point number within the sliding window, and k is a unique sampling time corresponding to each number. Let be the instantaneous amplitude of the active oscillation corresponding to the l-th sampling point within the sliding window. This represents the instantaneous amplitude of the steam pressure oscillation corresponding to the l-th sampling point within the sliding window; when the sliding window moves one sampling point to the right, the incremental update of the cumulative amount is performed according to the following rules: ;in , To determine the amplitude corresponding to the earliest sampling point of the sliding window, The amplitude corresponding to the latest sampling point in the sliding window is used. The time-domain correlation calculation is based on the cumulative value after incremental update. Since the input signal is a low-frequency oscillation component extracted by bandpass filtering, the DC component is approximately zero, which satisfies the premise for calculating the time-domain coherence coefficient. The time-domain coherence coefficient between the instantaneous amplitude of the active power oscillation and the instantaneous amplitude of the steam pressure oscillation within the current window is calculated. The frequency consistency fusion is used to calculate the instantaneous oscillation frequency of the active power oscillation at the current moment. Instantaneous oscillation frequency with steam pressure oscillation instantaneous frequency deviation ; and combined with the time-domain coherence coefficient and instantaneous frequency deviation, the computer electrical coupling degree is fused. The electromechanical coupling degree index has a value range of [0,1]. The higher the value, the stronger the time-domain correlation and the higher the frequency consistency between electrical oscillation and mechanical oscillation, that is, the more significant the electromechanical coupling effect. The electromechanical coupling degree serves as the quantitative weight basis for subsequent oscillation source localization and oscillation characteristic analysis. The calculation results and corresponding characteristic parameters are transmitted to the oscillation source localization module and the oscillation characteristic analysis module.

[0024] In detail: In the electromechanical coupling characteristic quantification stage, the electromechanical coupling degree calculation module performs the coupling degree calculation. This stage is initiated after the two-sided oscillation feature extraction is completed. After the instantaneous feature data of the two-sided oscillation are synchronously input into the module, the module uses a sliding window incremental recursive algorithm to perform the calculation. A fixed-length sliding analysis window is pre-set, and five statistical cumulative quantities are pre-stored and maintained within the window: the cumulative sum of active power oscillation amplitude, the cumulative sum of steam pressure oscillation amplitude, the cumulative sum of the cross-multiplication of the two types of amplitudes, the cumulative sum of the squares of active power oscillation amplitude, and the cumulative sum of the squares of steam pressure oscillation amplitude. Whenever the sliding window moves one sampling point to the right, only the five cumulative quantities are calculated. The contribution value of the earliest sample point removed from the window is removed from the cumulative value, while the contribution value of the latest sample point moved into the window is added. The cumulative value can be updated without repeating the summation for all sample points in the window. Then, the time-domain coherence coefficient of the oscillation amplitude on both sides is calculated based on the updated cumulative value. At the same time, the instantaneous frequency deviation of the oscillation on both sides is calculated. The electromechanical coupling degree index is obtained by combining the time-domain coherence coefficient and the frequency consistency penalty term. The index ranges from zero to one and can reflect the tightness of the coupling of the oscillation on both sides. Finally, the calculation results and corresponding characteristic parameters are transmitted to the oscillation source localization module and the oscillation characteristic analysis module.

[0025] Table 1 shows the measured data for the selection of the sliding analysis window length parameter. This table is based on measured low-frequency oscillation data from a 300MW thermal power unit. The actual electromechanical coupling degree obtained through offline global identification is used as the accuracy benchmark. The computation time was measured on an industrial-grade embedded processor with a main frequency of 2.0GHz to verify the rationality of the sliding window length parameter selection. The data shows that the single-step time under the incremental recursive algorithm does not increase significantly with the increase of the window length. A 100-point window length can match the typical time scale of low-frequency oscillations in thermal power units while maintaining engineering estimation accuracy.

[0026] Table 1. Actual Measured Data for Sliding Analysis Window Length Parameter Selection

[0027] S5. Using electromechanical coupling degree as the weight, phase energy transfer factor and amplitude dominance are integrated to carry out hierarchical progressive oscillation source localization, and the prime mover side disturbance type is subdivided through condition triggering mechanism. In this embodiment, it should be noted that: the electromechanical coupling degree is a continuous quantization index generated by S4, which is used as the weighting basis for hierarchical positioning and input to the oscillation source positioning module; the phase energy transfer factor is calculated based on the two-sided phase difference, and the amplitude dominance is calculated based on the relative strength of the two-sided amplitudes; wherein, before calculating the phase difference, all instantaneous phase sequences are first processed. Phase dewinding: Compare the phase values ​​of adjacent sampling points point by point. If the absolute value of the phase difference between the current sampling point and the previous sampling point is greater than... Then the phase of the current sampling point will be added or subtracted accordingly. This makes the absolute value of the adjacent phase difference less than To eliminate numerical abrupt changes caused by periodic phase transitions; the above procedures are performed on the instantaneous phases of active power oscillation, steam pressure oscillation, frequency oscillation, and speed oscillation, respectively. After phase dewinding, the energy transfer direction factor is obtained. Characterizes the oscillatory energy transfer trend reflected by the two-sided phase difference: ,in The instantaneous phase of the active oscillation after unwinding Instantaneous phase with steam pressure oscillation The difference, The instantaneous phase of the frequency oscillation after unwinding The instantaneous phase of the rotational speed oscillation The difference, the factor takes values ​​in the range [-1, 1], positive values ​​correspond to the transfer of oscillating energy from the grid side to the prime mover side, negative values ​​correspond to the transfer of oscillating energy from the prime mover side to the grid side, and the larger the absolute value, the more obvious the direction of transfer; calculate the dominance of the single-sided amplitude. Characterizes the relative strength of the amplitude of the two-sided oscillations: The index ranges from -1 to 1, with positive values ​​corresponding to dominant electrical-side oscillation amplitude and negative values ​​corresponding to dominant mechanical-side oscillation amplitude. The hierarchical progressive oscillation source positioning employs a three-layer cascaded computational unit deployment. After the first layer's energy transfer direction factor calculation is completed, the second layer's single-side amplitude dominance calculation is automatically triggered. After the second layer's calculation is completed, the third layer's comprehensive bias coefficient fusion calculation is automatically triggered. Electromechanical coupling degree... Using weights, the energy transfer direction factor and the dominance of one-sided amplitude are combined to obtain the comprehensive oscillation source bias coefficient. The comprehensive oscillation source bias coefficient ranges from -1 to 1. A positive value indicates that the oscillation source is biased towards the grid side, while a negative value indicates that the oscillation source is biased towards the prime mover side. The larger the absolute value, the stronger the side dominance of the oscillation source. The method for subdividing the disturbance type on the prime mover side is as follows: a conditional triggering mechanism is adopted for oscillation scenarios dominated by the prime mover side. The disturbance type subdivision operator is only loaded and executed when the comprehensive oscillation source bias coefficient is negative. In the non-triggered state, the subdivision operator is in a dormant state to save computing power. After triggering, the prime mover side disturbance subdivision coefficient is analyzed. ,in The difference between the instantaneous phase of the valve opening oscillation and the instantaneous phase of the steam pressure oscillation is calculated. When this coefficient is positive, the disturbance originates from the speed control system; when this coefficient is negative, the disturbance originates from the steam flow path. The larger the absolute value, the more specific the direction of the disturbance source. All the quantitative results of the oscillation source location are transmitted to the oscillation characteristic analysis module and the result output module.

[0028] In detail: In the oscillation source localization and disturbance subdivision stage, the localization calculation is performed by the three-layer pipeline computing unit of the oscillation source localization module. This stage is initiated after the electromechanical coupling degree calculation is completed. Once the electromechanical coupling degree and bilateral oscillation characteristic data are all ready, the module first performs phase period dewinding processing on all instantaneous phase sequences, comparing the phase values ​​of adjacent sampling points point by point. If the absolute value of the difference between adjacent phases exceeds the phase angle corresponding to half a cycle, the phase of the current sampling point is adjusted by adding or subtracting the corresponding whole cycle to keep the difference between adjacent phases within half a cycle, eliminating the numerical abrupt changes caused by periodic phase jumps. Subsequently, according to the three-layer pipeline cascade scheduling, the first layer calculates the energy transfer direction factor to reflect the transfer trend of oscillation energy. The second layer calculates the dominance of the amplitude on one side to reflect the relative strength of the oscillations on both sides. The third layer uses electromechanical coupling degree as a dynamic weight and integrates the calculation results of the first two layers to obtain the comprehensive oscillation source bias coefficient. This coefficient can clearly determine whether the oscillation source is biased towards the grid side or the prime mover side. For the oscillation scenario dominated by the prime mover side, a condition triggering mechanism is adopted. Only when the comprehensive biasing coefficient determines that the oscillation source is biased towards the prime mover side will the disturbance type subdivision operator be loaded and executed. The phase relationship between valve opening and steam pressure is used to distinguish whether the disturbance comes from the speed control system control link or the steam flow link. In the non-triggered state, the subdivision operator is in a dormant state to save computing power. Finally, all oscillation source location and quantization results are transmitted to the oscillation characteristic analysis module and the result output module.

[0029] S6. Based on the numerical range of electromechanical coupling, the corresponding analytical unit is adaptively loaded to perform multi-dimensional characteristic analysis of the oscillation; In this embodiment, it should be noted that: the numerical range of the electromechanical coupling degree is divided into a low-value range and a high-value range, which serve as the basis for adaptive loading; the adaptive loading corresponding parsing unit refers to the mapping relationship between the electromechanical coupling degree and the parsing dimension pre-established by the parsing engine. When the electromechanical coupling degree is in the low-value range, the engine only loads the oscillation damping ratio calculation unit and the electrical oscillation dominance coefficient calculation unit; when the electromechanical coupling degree is in the high-value range, the engine automatically loads the electromechanical oscillation effective coupling complex gain calculation unit to realize the dynamic allocation of computing resources according to the coupling strength; and all characteristic quantities (the three types of characteristic parameters: electromechanical oscillation effective coupling complex gain, oscillation damping ratio, and electrical oscillation dominance coefficient) are updated point by point using a sliding window recursive method to ensure the real-time performance of the parsing results; the multi-dimensional characteristic parsing is used to obtain the electromechanical oscillation effective coupling complex gain, oscillation damping ratio, and electrical oscillation dominance coefficient: when the electromechanical coupling degree is in the high-value range, the coupling complex gain calculation unit is loaded, and the electromechanical oscillation effective coupling complex gain is calculated. Characterizes the amplitude amplification effect and phase shift characteristics of the oscillating signal transmitted from the source side to the other side under coupling: ,in This represents the maximum instantaneous amplitude of the active power oscillation within the current sliding analysis window. This represents the maximum instantaneous amplitude of the steam pressure oscillation within the current sliding analysis window. The value of j represents the average difference between the instantaneous phase of the active oscillation and the instantaneous phase of the steam pressure oscillation within the current sliding analysis window, where j is the imaginary unit. Based on the oscillation amplitude variation within the continuous sliding window, the oscillation damping ratio is calculated recursively. This characterizes the decay or divergence trend of oscillations; this formula is applicable to typical low-frequency oscillation scenarios with low damping in power systems, i.e., underdamped oscillations with a damping ratio much less than 1. ;in Let W be the instantaneous amplitude of the active oscillation after an interval of W sampling points. The time interval between two amplitude sampling points, where For synchronous sampling interval, The average value of the instantaneous oscillation frequency of the active power oscillation within the current sliding analysis window; calculate the dominant electrical oscillation coefficient. Identify the dominant type of oscillation: ,in The coefficient is the difference between the instantaneous phase of the active power oscillation after unwinding and the instantaneous phase of the positive sequence voltage oscillation. The closer this coefficient is to 1, the more dominant the oscillation is in power angle swing; the closer this coefficient is to 0, the more dominant the oscillation is in voltage amplitude fluctuation. Among them, the coupling complex gain, oscillation damping ratio and electrical oscillation dominance coefficient are all characteristic analytical parameters, which are transmitted to the result output module.

[0030] In detail: In the oscillation multi-dimensional characteristic analysis stage, the adaptive analysis engine of the oscillation characteristic analysis module performs the analytical calculations. This stage can be carried out simultaneously with the oscillation source location. The module pre-establishes a mapping relationship between the electromechanical coupling degree numerical range and the analysis dimension, and adopts a dynamic module loading mechanism to realize the on-demand allocation of computing power: when the electromechanical coupling degree is in the low value range, the engine only loads the oscillation damping ratio calculation unit and the electrical oscillation dominance coefficient calculation unit; when the electromechanical coupling degree is in the high value range, the engine automatically loads the electromechanical oscillation effective coupling complex gain calculation unit; all characteristic parameters are updated point by point using a sliding window recursive method to obtain the oscillation damping ratio that reflects the oscillation decay or divergence trend, the electrical oscillation dominance coefficient that reflects the oscillation dominance type, and the electromechanical coupling complex gain that reflects the oscillation transmission amplitude amplification and phase shift characteristics in high coupling scenarios. Finally, all characteristic analysis parameters are transmitted to the result output module.

[0031] Table 2 shows the adaptive resolution coupling threshold verification data. This table is based on 16 typical electromechanical coupled oscillation and pure grid-side oscillation samples from the field. The accuracy was judged using the offline identified electromechanical transfer gain as the benchmark. Time consumption tests were completed on the same type of industrial processor to verify the rationality of the selected coupling threshold for adaptive resolution. The data shows that a threshold of 0.5 can effectively reduce invalid calculations in low-coupling scenarios while ensuring resolution accuracy in high-coupling scenarios, thus balancing resolution accuracy and computational efficiency.

[0032] Table 2 Adaptive Analytical Coupling Degree Threshold Validation Data Table

[0033] S7. Based on the positive and negative sequence voltage oscillation amplitude ratio, the asymmetric component of the active oscillation amplitude is stripped, and the synchronous correction of the link amplitude data is achieved through a double buffer reverse broadcast mechanism. In this embodiment, it should be noted that: the ratio of positive to negative sequence voltage oscillation amplitudes is the ratio of the instantaneous amplitude of the negative sequence voltage oscillation to the instantaneous amplitude of the positive sequence voltage oscillation, which serves as the basis for calculating asymmetric component stripping; the active power oscillation amplitude is the instantaneous amplitude of the original active power oscillation, which serves as the object of asymmetric correction; the asymmetric component stripping generates an asymmetric correction coefficient based on the proportion of negative sequence voltage, and removes interference components from the active power oscillation amplitude; the asymmetric correction coefficient at the current moment is calculated. This is used to characterize the proportion of interference of negative-sequence components on the amplitude of active oscillations. ,in This represents the instantaneous amplitude of the negative-sequence voltage oscillation. The instantaneous amplitude of the positive-sequence voltage oscillation is given; based on the asymmetric correction coefficient, the asymmetric component of the instantaneous amplitude of the active power oscillation is removed to obtain the corrected instantaneous amplitude of the active power oscillation. The dual-buffer reverse broadcast mechanism refers to the broadcasting of corrected data through a dedicated reverse channel of the internal parallel bus. All upstream computing modules complete the in-situ replacement of the corresponding data based on the sampling point sequence number. The module has a built-in dual buffer. During the correction process, the new amplitude data is first written to the backup buffer. After the timing of all sampling points is consistent, it is switched to the main working buffer at once to avoid data inconsistency during the correction. The synchronous correction of the link amplitude data refers to the synchronous replacement of the original amplitude of the corresponding sampling point in the electromechanical coupling degree calculation, oscillation source location, and oscillation characteristic analysis modules after the correction is completed, so as to achieve consistent updates of the entire link data. The corrected active oscillation amplitude serves as the unified basic input data for all computing links.

[0034] In detail: In the asymmetrical operating condition correction stage, the asymmetrical correction module performs full-link amplitude correction. This stage can play a role in three-phase asymmetrical operating conditions such as non-full-phase operation and asymmetrical faults in the power grid. The module generates asymmetrical correction coefficients based on the ratio of positive and negative sequence voltage oscillation amplitudes, removes the asymmetrical component from the active power oscillation amplitude, and obtains the corrected amplitude after eliminating negative sequence interference. The correction process adopts a dual-buffer mechanism. First, the corrected new amplitude data is written to the backup buffer. After the timing of the full sampling point verification is consistent, it is switched to the main working buffer at once to avoid data inconsistency during the correction. Then, the corrected data is broadcast through the reverse dedicated channel of the internal parallel bus. All upstream calculation modules complete the in-situ replacement of the corresponding data according to the sampling point sequence number, and synchronously update the corresponding original amplitudes in the electromechanical coupling degree calculation, oscillation source location, and oscillation characteristic analysis stages to achieve consistent correction of the full-link data.

[0035] S8. Standardize the data encapsulation and cyclic history cache of the oscillation discrimination results and characteristic parameters, output structured discrimination information and support the backtracking analysis of oscillation events.

[0036] In this embodiment, it should be noted that: the oscillation discrimination results and characteristic parameters include oscillation event discrimination results, comprehensive oscillation source bias coefficient, prime mover-side disturbance subdivision coefficient, electromechanical coupling degree, effective coupling complex gain of electromechanical oscillation, oscillation damping ratio, electrical oscillation dominance coefficient, asymmetry correction coefficient, and active oscillation dominance frequency, which are input to the result output module; the standardized data encapsulation encapsulates all information in a structured manner according to the general communication protocol format of the power system; the cyclic historical data cache uses a built-in cyclic historical database to cache the original sampling data, intermediate feature data, and final result data in a first-in-first-out mechanism, supporting the retrospective playback and offline analysis of oscillation events; the structured discrimination information is output to the power grid stability monitoring system through a standard communication interface; the oscillation event retrospective analysis is realized based on the historical cached data and can be used for offline fault review and algorithm iteration optimization.

[0037] In detail: In the result output and data retention stage, the result output module performs information encapsulation and data caching. This stage is the final output stage of the entire process. Once all oscillation discrimination results and characteristic parameters are ready, the module encapsulates all information, including oscillation event discrimination results, oscillation source location parameters, electromechanical coupling index, oscillation characteristic analysis parameters, and asymmetry correction coefficient, according to the general communication protocol format of the power system. This information is then output to the power grid stability monitoring system through a standard communication interface. Simultaneously, the module has a built-in cyclic historical database that continuously caches original sampling data, intermediate feature data, and final result data using a first-in-first-out mechanism. This supports the retrospective playback and offline analysis of oscillation events, providing a data foundation for subsequent fault review and algorithm optimization.

[0038] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A software modular implementation method for a fast power oscillation detection algorithm for thermal power units, characterized in that, include: S1. Obtain the raw oscillation data, perform dual-path parallel buffering and cubic spline interpolation resampling, complete time base alignment and per-unit value normalization processing, and generate a standardized time-series data sequence synchronized with the time axis. S2. A three-stage pipeline architecture is used to perform Hilbert transform demodulation on standardized electrical side time-series data, extract instantaneous features of electrical oscillations, and complete the basic identification of oscillation events. S3. A Hilbert transform demodulation is performed on the standardized mechanical side timing data using a co-source clock triggering mechanism to extract the instantaneous oscillation features of the mechanical side that are aligned with the electrical side timing. S4. The instantaneous characteristics of bilateral oscillations are calculated in the time domain and fused with frequency consistency by using the sliding window incremental recursive algorithm to generate a continuously quantized electromechanical coupling index. S5. Using electromechanical coupling degree as the weight, phase energy transfer factor and amplitude dominance are integrated to carry out hierarchical progressive oscillation source localization, and the prime mover side disturbance type is subdivided through condition triggering mechanism. S6. Based on the numerical range of electromechanical coupling, the corresponding analytical unit is adaptively loaded to perform multi-dimensional characteristic analysis of the oscillation; S7. Based on the positive and negative sequence voltage oscillation amplitude ratio, the asymmetric component of the active oscillation amplitude is stripped, and the synchronous correction of the link amplitude data is achieved through a double buffer reverse broadcast mechanism. S8. Standardize the data encapsulation and cyclic history cache of the oscillation discrimination results and characteristic parameters, output structured discrimination information and support the backtracking analysis of oscillation events.

2. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 1, characterized in that: The raw oscillation data includes electrical-side data and mechanical-side data. The electrical-side data comes from a synchronous phasor measurement device and includes the instantaneous low-frequency oscillation values ​​of the generator's three-phase total active power, the generator terminal fundamental positive-sequence voltage amplitude, the generator terminal fundamental negative-sequence voltage amplitude, and the generator terminal frequency. The mechanical-side data comes from a distributed control system and includes the instantaneous low-frequency oscillation values ​​of the steam pressure after the turbine regulating stage, the turbine speed, and the turbine governor valve opening command. The instantaneous value of the frequency oscillation; the dual-channel parallel buffer and cubic spline interpolation resampling are implemented using a dual-channel independent ring buffer architecture, corresponding to the electrical side data channel and the mechanical side data channel respectively. The two data channels can be written to the buffer in parallel without blocking and without interference; the interpolation resampling uses the GPS synchronization time scale as a unified reference, and performs cubic spline interpolation operation on the two types of data with different sampling rates to unify the data with different sampling rates into a synchronization point sequence with the same time interval; the time base alignment eliminates the time base deviation of the two types of data through interpolation resampling, and the unified synchronization sampling interval is denoted as . The per-unit normalization is based on the rated values ​​of each parameter, and all time-series data are converted to per-unit values ​​to eliminate the influence of different dimensions on subsequent coupled calculations.

3. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 2, characterized in that: The three-stage pipeline architecture is sequentially divided into a data framing unit, an orthogonal transformation unit, and a feature calculation unit. These three units operate sequentially, and under steady-state continuous operation, the overall processing delay of a single sampling point does not exceed one synchronous sampling interval. The method for demodulating the standardized electrical side time-series data using Hilbert transform is as follows: based on the Hilbert transform, an analytical signal is constructed for each electrical oscillation signal to complete the demodulation and extraction of instantaneous features; the instantaneous value of the low-frequency oscillation of the generator's three-phase total active power at time k is... Performing a Hilbert transform yields the active oscillation Hilbert quadrature components. This leads to the construction of an active oscillation analytical signal: Where j is the imaginary unit; the instantaneous characteristics of the electrical oscillation include the instantaneous amplitude, instantaneous phase, and instantaneous frequency characteristics of four types of signals: active oscillation, positive-sequence voltage oscillation, negative-sequence voltage oscillation, and frequency oscillation; the instantaneous amplitude of the active oscillation is obtained based on the analytical signal. The instantaneous phase of the active oscillation is obtained using the four-quadrant arctangent function. Further obtain the instantaneous oscillation frequency of the active oscillation. ,in For synchronous sampling interval, The instantaneous phase of the active power oscillation at time k-1; simultaneously, the instantaneous low-frequency oscillation value of the generator terminal fundamental positive sequence voltage amplitude at time k. By performing a Hilbert transform, the corresponding orthogonal components are obtained, and a positive-sequence voltage oscillation analytic signal is constructed, thereby obtaining the instantaneous amplitude of the positive-sequence voltage oscillation. Instantaneous phase of positive sequence voltage oscillation ; the instantaneous value of the low-frequency oscillation of the fundamental negative sequence voltage amplitude at the generator terminals at time k. By performing a Hilbert transform, the corresponding quadrature components are obtained, and a negative-sequence voltage oscillation analytic signal is constructed, thereby obtaining the instantaneous amplitude of the negative-sequence voltage oscillation. ; for the instantaneous value of low-frequency oscillation at generator terminal frequency at time k. Perform a Hilbert transform to obtain the corresponding orthogonal components and construct a frequency oscillation analytic signal, thereby obtaining the instantaneous phase of the frequency oscillation. The oscillation event identification is based on the instantaneous amplitude of the active oscillation to generate a discrimination result. When the instantaneous amplitude continuously exceeds the set oscillation start amplitude threshold, it is determined that a power oscillation event has occurred.

4. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 3, characterized in that: The method for demodulating standardized mechanical side time-series data by Hilbert transform is as follows: perform Hilbert transform, analytical signal construction and feature quantity calculation on each mechanical oscillation signal in sequence, and extract the corresponding instantaneous features. The instantaneous value of low-frequency oscillation of steam pressure after the regulating stage of the turbine at time k. Performing a Hilbert transform yields the Hilbert orthogonal components of the steam pressure oscillation. Constructing an analytical signal of vapor pressure oscillation The instantaneous amplitude of the steam pressure oscillation is obtained based on the analytical signal. Instantaneous phase of steam pressure oscillation The instantaneous oscillation frequency of the steam pressure oscillation is obtained by measuring the phase difference between adjacent sampling points. ,in For synchronous sampling interval, The instantaneous phase of the steam pressure oscillation at time k-1; the instantaneous value of the low-frequency oscillation of the turbine speed at time k. Perform a Hilbert transform to obtain the corresponding orthogonal components and construct the speed oscillation analytical signal, thereby obtaining the instantaneous phase of the speed oscillation. ; the instantaneous value of low-frequency oscillation of the turbine governor valve opening command at time k. Perform a Hilbert transform to obtain the corresponding orthogonal components and construct the valve opening oscillation analytical signal, thereby obtaining the instantaneous phase of the valve opening oscillation. .

5. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 4, characterized in that: The core mechanism of the sliding window incremental recursive algorithm is to pre-store and maintain five statistical cumulative quantities within the sliding window, namely, the cumulative sum of active power oscillation amplitude, the cumulative sum of steam pressure oscillation amplitude, the cumulative sum of cross-multiplication of the two types of amplitudes, the cumulative sum of the squares of active power amplitude, and the cumulative sum of the squares of steam pressure amplitude. When a new sampling point enters the analysis window, only the corresponding contribution value of the earliest exiting sampling point is removed from the five cumulative quantities, and the corresponding contribution value of the latest entering sampling point is superimposed. There is no need to repeatedly perform summation operations on all sampling points in the entire window, reducing the single-step computation complexity from linear to constant order. The time-domain correlation calculation is carried out based on the incrementally updated cumulative quantities. Since the input signal is a low-frequency oscillation component extracted by bandpass filtering, the DC component is approximately zero, which meets the premise for calculating the time-domain coherence coefficient. The time-domain coherence coefficient between the instantaneous amplitude of active power oscillation and the instantaneous amplitude of steam pressure oscillation within the current window is calculated. The frequency consistency fusion is used to calculate the instantaneous frequency deviation between the instantaneous oscillation frequency of the active power oscillation and the instantaneous oscillation frequency of the steam pressure oscillation at the current moment. ; and combined with the time-domain coherence coefficient and instantaneous frequency deviation, the computer electrical coupling degree is fused. .

6. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 5, characterized in that: The phase energy transfer factor is calculated based on the two-sided phase difference, and the amplitude dominance is calculated based on the relative strength of the two-sided amplitudes; wherein, before calculating the phase difference, all instantaneous phase sequences are first processed. Phase dewinding: Compare the phase values ​​of adjacent sampling points point by point. If the absolute value of the phase difference between the current sampling point and the previous sampling point is greater than... Then the phase of the current sampling point will be added or subtracted accordingly. This makes the absolute value of the adjacent phase difference less than To eliminate numerical abrupt changes caused by periodic phase transitions; the above procedures are performed on the instantaneous phases of active power oscillation, steam pressure oscillation, frequency oscillation, and speed oscillation, respectively. After phase dewinding, the energy transfer direction factor is obtained. Simultaneously obtain the dominance of one-sided amplitude. The hierarchical progressive oscillation source positioning adopts a three-layer computing unit cascaded deployment in a pipeline. After the first layer energy transfer direction factor calculation is completed, the second layer single-sided amplitude dominance calculation is automatically triggered. After the second layer calculation is completed, the third layer comprehensive bias coefficient fusion calculation is automatically triggered. Electromechanical coupling degree is used as the basis for this calculation. Using weights, the energy transfer direction factor and the dominance of one-sided amplitude are combined to obtain the comprehensive oscillation source bias coefficient. The method for subdividing the disturbance type on the prime mover side is as follows: a conditional triggering mechanism is adopted for oscillation scenarios dominated by the prime mover side. The disturbance type subdivision operator is only loaded and executed when the comprehensive oscillation source bias coefficient is negative. In the non-triggered state, the subdivision operator is in a dormant state to save computing power. After triggering, the disturbance subdivision coefficient on the prime mover side is analyzed. .

7. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 5, characterized in that: The electromechanical coupling degree is divided into low and high value ranges, serving as the basis for adaptive loading. The adaptive loading corresponding analytical unit refers to the mapping relationship between the electromechanical coupling degree and the analytical dimension pre-established by the analytical engine. When the electromechanical coupling degree is in the low value range, the engine only loads the oscillation damping ratio calculation unit and the electrical oscillation dominance coefficient calculation unit. When the electromechanical coupling degree is in the high value range, the engine automatically loads the electromechanical oscillation effective coupling complex gain calculation unit. The multi-dimensional characteristic analysis is used to obtain the electromechanical oscillation effective coupling complex gain, oscillation damping ratio, and electrical oscillation dominance coefficient: when the electromechanical coupling degree is in the high value range, the coupling complex gain calculation unit is loaded, and the electromechanical oscillation effective coupling complex gain is calculated. ,in This represents the maximum instantaneous amplitude of the active power oscillation within the current sliding analysis window. This represents the maximum instantaneous amplitude of the steam pressure oscillation within the current sliding analysis window. The value of j represents the average difference between the instantaneous phase of the active oscillation and the instantaneous phase of the steam pressure oscillation within the current sliding analysis window, where j is the imaginary unit. Based on the oscillation amplitude variation within the continuous sliding window, the oscillation damping ratio is calculated recursively. ,in Let W be the instantaneous amplitude of the active oscillation after an interval of W sampling points. The time interval between two amplitude sampling points, where For synchronous sampling interval, The average value of the instantaneous oscillation frequency of the active power oscillation within the current sliding analysis window; calculate the dominant electrical oscillation coefficient. Identify the dominant type of oscillation: .

8. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 3, characterized in that: The asymmetric component stripping is based on the negative sequence voltage ratio to generate an asymmetric correction coefficient, and interference components are removed from the active oscillation amplitude. Calculate the asymmetric correction coefficient at the current time. Based on the asymmetric correction coefficient, the asymmetric component of the instantaneous amplitude of the active power oscillation is removed to obtain the corrected instantaneous amplitude of the active power oscillation. The dual-buffer reverse broadcast mechanism refers to the broadcasting of the corrected data through the reverse dedicated channel of the internal parallel bus. All upstream computing modules complete the in-situ replacement of the corresponding data according to the sampling point sequence number. The module has a built-in dual buffer. During the correction process, the new amplitude data is first written to the backup buffer. After the timing of the full sampling point verification is consistent, it is switched to the main working buffer at one time.

9. The software modular implementation method of the fast power oscillation discrimination algorithm for thermal power units according to claim 1, characterized in that: The oscillation discrimination results and characteristic parameters include oscillation event discrimination results, comprehensive oscillation source bias coefficient, prime mover-side disturbance subdivision coefficient, electromechanical coupling degree, effective coupling complex gain of electromechanical oscillation, oscillation damping ratio, electrical oscillation dominance coefficient, asymmetry correction coefficient, and active oscillation dominance frequency, which are input to the result output module; the standardized data encapsulation encapsulates all information in a structured manner according to the general communication protocol format of the power system. The cyclic historical data cache uses a built-in cyclic historical database to cache the original sampling data, intermediate feature data and final result data in a first-in-first-out mechanism, supporting the retrospective playback and offline analysis of oscillation events; the structured discrimination information is output to the power grid stability monitoring system through a standard communication interface; the oscillation event retrospective analysis is realized based on the historical cached data and can be used for offline fault review and algorithm iteration optimization.

Images