Primary frequency modulation control method and system based on big data and signal homology

Abstract

The invention discloses a primary frequency modulation control method and system based on big data and signal homology, and the method comprises the steps: collecting a terminal voltage and a PMU signal, and combining dynamic DFT calculation, multi-source fusion and dynamic calibration to obtain a high-precision power grid real-time frequency; working condition clustering is carried out by adopting an improved K-Medoids algorithm based on the real-time frequency of the power grid, a steam flow characteristic curve is established through an improved Friedel-Gel formula and a characteristic through-flow area method, and a frequency modulation control signal is output. According to the method, traditional rotating speed measurement delay is avoided, and frequency measurement precision and anti-interference capability are improved; through working condition clustering and flow identification, effective disturbance is distinguished, steam admission characteristics of a unit are matched, and invalid adjustment and valve oscillation are reduced. The method gives consideration to the primary frequency modulation performance of the power grid and safe and economical operation of the unit, and meets the stable control requirements of the new energy high-permeability power grid.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system automatic control, in particular to a primary frequency modulation control method and system based on big data and signal homology. BACKGROUND

[0002] In the scenario of high penetration of new energy in power grid, the traditional primary frequency modulation control method has obvious limitations. First, in terms of speed signal acquisition, the traditional method relies on reluctance sensors or speed measuring gears to measure the speed of the turbine, which is affected by mechanical inertia and has a response delay, making it difficult to achieve a fast response in the early stage of power grid frequency disturbance. At the same time, such measurement is easily affected by local interference, installation deviation and gear processing error, resulting in insufficient signal accuracy, further causing misoperation, lag or refusal in the primary frequency modulation process, and thus exacerbating the risk of power grid frequency drop. Second, in terms of control strategy, the existing method mainly improves the frequency modulation performance by reducing the action dead zone, increasing the slope of the speed difference-power function curve, etc. Although this improves the response effect to some extent, it causes frequent oscillation of the turbine regulating valve, generates a large amount of invalid disturbance, and affects the safe and economic operation of the unit, making it difficult to meet the high-precision and low-disturbance requirements of new power grid for primary frequency modulation.

[0003] In view of the above problems existing in the primary frequency modulation process, such as speed signal lag, susceptibility to interference and poor adaptability of control strategy, the existing technology mainly tries to improve it in the following ways: in the signal measurement link, some studies use high-precision photoelectric encoders, laser speed meters, etc. to replace traditional reluctance sensors to improve measurement accuracy and anti-interference ability; there are also technologies that modify the speed signal through digital filtering and signal reconstruction methods to reduce measurement noise and response delay. In the control strategy link, the existing method usually optimizes the "speed difference-power" function characteristic curve, such as reducing the frequency modulation action dead zone, adjusting the slope parameter, introducing hysteresis compensation, etc., to enhance the sensitivity of the unit to frequency disturbance. The above existing technology still has deficiencies: high-precision sensors are high in cost, complex to install and high in operation and maintenance requirements, making it difficult to popularize in large-scale units; while the control strategy based on function optimization improves the sensitivity, it still cannot balance the stability of power grid frequency and the safety and economy of the unit. Therefore, there is an urgent need for a new frequency modulation method that can balance the reliability of signal measurement and the adaptability of control strategy. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a primary frequency modulation control method based on big data and signal homology to solve the problem of unstable frequency control in high penetration of new energy in power grid.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In the first aspect, the present application provides a primary frequency modulation control method based on big data and signal homology, which comprises:

[0008] Collecting voltage signals, preprocessing terminal voltage and PMU output voltage signals, and correcting voltage signals;

[0009] Fusing and calibrating the corrected voltage signals to obtain calibrated real-time grid frequency;

[0010] Based on the calibrated real-time grid frequency, the working condition is clustered, the characteristic of the working condition clustering result is identified to obtain a frequency modulation control signal, and the primary frequency modulation control is performed according to the frequency modulation control signal;

[0011] In the process of collecting voltage signals, threshold method is used to trigger alarm and remove bad points.

[0012] As a preferred scheme of the primary frequency modulation control method based on big data and signal homology, the preprocessing and correction of terminal voltage or PMU output voltage signals comprise: using a 16-bit analog-to-digital converter to sample the voltage signals, filtering the sampled signals through an 8th-order Butterworth band-pass filter, the passband frequency range of the filter being 45-55Hz, and then using a sliding quartile range method to remove outliers from the filtered signals, the window length of the sliding quartile range method being 10 cycles;

[0013] The correction of voltage signals comprises: calculating the initial frequency based on the preprocessed voltage signals through a fundamental frequency estimation formula, correcting the amplitude using a spectrum leakage compensation formula, and outputting the calculation result; the fundamental frequency estimation formula is:

[0014]

[0015] Wherein, f0 is the estimated fundamental frequency, N is the dynamically adjusted sampling point number, f s is the sampling frequency, φ is the fundamental phase angle, and dφ / dt is the change rate of the fundamental phase angle; the spectrum leakage compensation formula is:

[0016]

[0017] Wherein, A corrected is the corrected amplitude, A k is the original amplitude of the kth frequency point, A k+1 is the original amplitude of the k+1th frequency point, A k-1 is the original amplitude of the k-1th frequency point, and δ is the frequency offset rate.

[0018] The output result is corrected by using the improved dynamic calibration DFT technology, and the improved dynamic calibration DFT technology comprises that, under the adaptive window length adjustment mechanism, the non-synchronous sampling error is phase compensated by using the Taylor expansion formula; the Fourier transform window length is dynamically adjusted according to the frequency change rate df / dt, the window length adjustment range of the adaptive window length adjustment mechanism is 25-100 cycles, and the specific rule of the adaptive window length adjustment mechanism is:

[0019]

[0020] Wherein, N new is the adjusted Fourier transform window length, N prev is the Fourier transform window length before adjustment, and Δf is the frequency change amount, and |Δf| is the absolute value of the frequency change amount.

[0021] As a preferred scheme of the power frequency control method based on big data and signal homology, the PMU data and the RTU data are fused with the synchronous phasor data of the SCADA through a federal Kalman filter, frequency data after fusion is output, and the frequency data after fusion is calibrated in real time, and the calibrated power grid real-time frequency is output.

[0022] The federal Kalman filter comprises a PMU data local filter, an RTU data local filter and a SCADA synchronous vector local filter, each local filter outputs a local frequency state estimation value, the local frequency state estimation value is uploaded to a fusion center for fusion, and the fused state quantity is frequency data; the local state estimation value is fused to obtain the fused state quantity The fusion formula of the federal Kalman filter is:

[0023]

[0024] Wherein, is the fused state quantity at the k moment, K1 is the weight coefficient of the PMU data, K2 is the weight coefficient of the RTU data, and K3 is the weight coefficient of the SCADA data, is the state estimation value of the PMU data, is the state estimation value of the RTU data, is the state estimation value of the SCADA data.

[0025]

[0026] Wherein, K iis the weight coefficient of the i-th type of data source, K1 is the weight coefficient of PMU data, K2 is the weight coefficient of RTU data, K3 is the weight coefficient of SCADA data, σ1 is the confidence index of PMU data, σ2 is the confidence index of RTU data, and σ3 is the confidence index of SCADA data;

[0027] weight coefficient K i When the confidence index σ i of any data source decreases, the weight coefficient K i of the data source is decreased synchronously; when the confidence index σ i of any type of data source increases, the weight coefficient K i of the data source is increased synchronously.

[0028] When the fused data is calibrated in real time, the reference clock is corrected in real time by using the IEEE 1588 precision time protocol, the time error of the protocol is <1 μs; the temperature drift is compensated by using a polynomial regression model, the determination coefficient R 2 of the polynomial regression model is >0.98, the compensation temperature range is -40-85℃, and the output power grid real-time frequency is output after calibration.

[0029] As a preferred scheme of the primary frequency control method based on big data and signal homology, when the voltage signal correction calculation is performed, the anti-harmonic interference enhancement algorithm is realized by reconstructing the fundamental component and setting the hysteresis threshold; first, a harmonic matrix h maxtrix containing 6 harmonics is constructed, and a linear equation group of the harmonic matrix and the input signal is solved based on the least square method to obtain a least square solution ls solution The coefficients of the harmonic components are determined, and the fundamental component is reconstructed based on the first two columns of the harmonic matrix and the first two elements of the least square solution:

[0030] fundamental=h matrix [:,0:2]@ls solution [0][0:2]

[0031] Wherein, fundamental is the reconstructed fundamental component signal; h matrix is the harmonic matrix, the basis functions of each order of harmonics are arranged in columns, h matrix [:,0:2] is the basis function of extracting the first two columns of the matrix, the sine and cosine components of the basis function; ls solution is the least square solution vector, containing the amplitude estimation results of each order of harmonic basis functions; ls solution [0][0:2] is the amplitude coefficient of the first two elements of the solution vector, the sine and cosine components; @ is a matrix multiplication operator, indicating that the basis function matrix is multiplied by the amplitude coefficient to obtain the reconstructed signal.

[0032] The adaptive zero-crossing detection algorithm with a hysteresis value of 0.005 is used to identify the zero-crossing point of the fundamental component crossings , and false zero-crossing points caused by noise or residual harmonics are removed, the time interval of adjacent zero-crossing points is recalculated, and the mean value of multiple time intervals is obtained as the time interval mean value T avg , and the frequency of the voltage signal is calculated based on the time interval mean value:

[0033]

[0034] Wherein, f is the frequency of the voltage signal, represents the grid fundamental frequency obtained after anti-interference processing; T avg is the time interval mean value, which is calculated from the time interval of adjacent zero-crossing points, and the mean value of multiple results is taken to eliminate sampling jitter and local errors;

[0035] Finally, the voltage signal frequency that offsets local sampling errors and jitter interference is obtained, and anti-interference sampling is realized.

[0036] As a preferred scheme of the one-frequency control method based on big data and signal homology, the working condition clustering based on the calibrated real-time frequency of the power grid comprises: performing working condition clustering of an improved K-Medoids mining algorithm: obtaining total valve position values of all actual working conditions through weighted calculation of the control system of the steam turbine generator set from the sampled voltage signal, selecting at least 250 data rows as rough initial centers of clusters, assigning all working condition data rows to the nearest cluster to quickly reduce the difference within the cluster; taking the working condition data row closest to the mean value of all data rows in each cluster as the initial center of the cluster to improve robustness, and again assigning all working condition data rows to the nearest cluster to make the clustering structure more stable;

[0037] The following operations are iteratively performed to gradually optimize the cluster center and reduce the overall error: selecting 50 non-cluster center data rows closest to the current cluster center, calculating the total cost TC ih of each non-cluster center data row replacing the current cluster center, if the minimum total cost is less than 0, replacing the current cluster center with the corresponding non-cluster center data row, and reassigning all working condition data rows to the nearest cluster; when the cluster center no longer changes, output the working condition clustering result to provide an input basis for flow characteristic identification; and judging the feasibility of replacing the cluster center with the non-cluster center data through the total cost formula:

[0038]

[0039] Wherein, TC ih is the total cost of replacing the cluster center O h with the non-cluster center O i , and n is the total number of non-cluster center objects, C jihFor the jth non-cluster center object O j Cost generated by replacement.

[0040] As a preferred scheme of the large data and signal homology based frequency control method, the cost C of the improved K-Medoids mining algorithm is calculated jih The calculation rule of the cost C is improved, only the distance between the new center and the second nearest center is compared to quickly determine whether the sample is replaced and calculate the cost, and the stability of the clustering result is maintained.

[0041] If O j belongs to the cluster represented by the center O i : when d(O j , O h ) ≥ d(O j , O j.2 ), C jih = d(O j , O j.2 )-d(O j , O i );

[0042] When d(O j , O h ) < d(O j , O j.2 ), C jih = d(O j , O h )-d(O j , O i );

[0043] If O j belongs to the cluster represented by the second nearest center O j.2 : when d(O j , O h ) ≥ d(O j , O j.2 ), C jih = 0;

[0044] When d(O j , O h ) < d(O j , O j.2 ), C jih = d(O j , O h )-d(O j , O j.2 ); wherein d is the Euclidean distance, O j.2 is the second nearest cluster center to O j .

[0045] As a preferred scheme of the primary frequency modulation control method based on big data and signal homology provided by the application, the characteristic identification of the working condition clustering result comprises: after the working condition clustering result is converted into quantifiable unit operation inlet characteristics through flow characteristic identification, a frequency modulation control signal is output to the unit control system; the flow characteristic identification is realized based on the improved Friis formula in which the temperature parameter is replaced by the product of the working medium pressure and specific volume and the characteristic flow area method; the characteristic flow area method is based on the main steam flow G B output by the improved Friis formula to calculate the unit inlet flow under different working conditions, generate a main steam valve flow characteristic curve, and output a frequency modulation control signal.

[0046]

[0047] wherein G A is the main steam flow under working condition A; G B is the main steam flow under working condition B; v 1A is the specific volume of the working medium in front of the low-pressure cylinder under working condition A; v 1B is the specific volume of the working medium in front of the low-pressure cylinder under working condition B; p 1A is the pressure of the working medium in front of the low-pressure cylinder under working condition A; p 1B is the pressure of the working medium in front of the low-pressure cylinder under working condition B; π A is the pressure ratio of the low-pressure cylinder under working condition A; π B is the pressure ratio of the low-pressure cylinder under working condition B; when the pressure ratio reaches a critical value, π A or π B is 0.

[0048] In the second aspect, the application provides a primary frequency modulation control system based on big data and signal homology, comprising: a signal homology frequency acquisition module; a signal preprocessing unit for preprocessing the terminal voltage or the voltage signal output by the PMU; a DFT dynamic calculation unit for correcting the preprocessed voltage signal; a multi-source data fusion unit for fusing the dynamic DFT calculation result, the PMU data, the RTU data and the SCADA synchronous phasor data by using a federal Kalman filter and outputting the fused frequency data; a dynamic calibration unit for calibrating the fused frequency data in real time and outputting the calibrated power grid real-time frequency; and a control strategy module based on big data analysis, wherein a working condition clustering unit of an improved K-Medoids mining algorithm is used to cluster and classify the acquired unit operation data, and a flow characteristic identification unit is used to establish the steam flow characteristic curve of the unit under different working conditions to provide an input basis for the primary frequency modulation.

[0049] In the third aspect, the application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any step of the primary frequency modulation control method based on big data and signal homology provided by the first aspect of the application is realized.

[0050] In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements any step of the method for primary frequency modulation control based on big data and signal homology according to the first aspect of the present application.

[0051] The method for primary frequency modulation control based on big data and signal homology provided by the present application avoids mechanical delay of traditional speed signals by using machine terminal voltage or PMU to obtain frequency through the signal homology module; reduces measurement error, makes frequency data more accurate, and thus enables the unit to quickly respond to power grid frequency disturbance and avoid frequency drop risk by combining dynamic DFT calculation, multi-source data fusion and dynamic calibration; the anti-interference unit automatically switches to traditional frequency modulation function when the machine terminal voltage is abnormal through the bad point alarm switching mechanism, avoiding frequency modulation failure caused by signal abnormalities; the multi-cycle frequency tracking technology eliminates abnormal frequency values, can resist random noise and local disturbance, and ensures that frequency modulation only responds to real demand of the power grid, adapting to complex operating conditions; the big data module accurately clusters operating conditions by improving the K-Medoids algorithm, distinguishes between valid and invalid disturbances, and avoids excessive adjustment to invalid disturbances; the flow identification based on the characteristic flow area generates accurate flow characteristic curves, realizes on-demand adjustment, reduces invalid disturbance of the unit, and balances the primary frequency modulation index of the power grid and the safe and economic operation of the unit. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0053] Figure 1 The flowchart of the method for primary frequency modulation control based on big data and signal homology provided by the embodiments of the present application. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail in conjunction with the drawings of the specification.

[0055] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0056] Second, the "one embodiment" or "an embodiment" referred to herein can include a particular feature, structure, or characteristic. The various embodiments are not mutually exclusive, but a single embodiment can be selected from a wider set of possible embodiments.

[0057] Referring to Figure 1 For one embodiment of the present application, the embodiment provides a large data and signal homology-based frequency modulation control method, comprising the following steps:

[0058] S1: Collecting voltage signals, preprocessing the terminal voltage and the voltage signal output by the PMU, and correcting the voltage signal. A 16-bit analog-to-digital converter is used to sample the voltage signal, the model of the ADC is AD7606, and the sampling frequency is ≥4.8 kHz; an 8th-order Butterworth band-pass filter is used to filter the sampled signal, the passband frequency range of the filter is 45-55 Hz; and a sliding interquartile range method is used to remove outliers from the filtered signal, the window length of the sliding interquartile range method is 10 cycles.

[0059] When performing voltage signal correction calculation, the anti-harmonic interference enhancement algorithm is realized by reconstructing the fundamental component and setting the hysteresis threshold to obtain the sampling frequency; first, a harmonic matrix h matrix and based on the least squares method, a system of linear equations of the harmonic matrix and the input signal is solved to obtain the least squares solution ls solution The coefficients of each harmonic component are determined, and the fundamental component is reconstructed based on the first 2 columns of the harmonic matrix and the first 2 elements of the least squares solution:

[0060] fundamental=h matrix [:,0:2]@ls solution [0][0:2]

[0061] Wherein, fundamental is the reconstructed fundamental component signal; h matrix is the harmonic matrix, which arranges the basis functions of each order harmonic by column, h matrix [:,0:2] is the basis function of the extracted first two columns of the matrix, which is the basis function of the fundamental sine and cosine components; ls solution is the least squares solution vector, which contains the amplitude estimation results of each order harmonic basis function; ls solution [0][0:2] is the amplitude coefficient of the first two elements of the solution vector, which is the fundamental sine and cosine components; @ is the matrix multiplication operator, which means multiplying the basis function matrix and the amplitude coefficient to obtain the reconstructed signal.

[0062] An adaptive zero-crossing detection algorithm with a hysteresis value of 0.005 is used to identify the zero-crossing point of the fundamental component zero crossings, eliminating false zero-crossing recognition caused by noise or residual harmonics, recalculating the time interval of adjacent zero-crossing points, and taking the mean of multiple time intervals to obtain the time interval mean T avg , calculating the frequency of the voltage signal based on the time interval mean, obtaining the time interval mean T avg , calculating the frequency of the voltage signal based on the time interval mean:

[0063]

[0064] where f is the frequency of the voltage signal, represents the fundamental frequency of the power grid obtained after anti-interference processing; T avg is the time interval mean, which is calculated from the time interval of adjacent zero-crossing points and is averaged multiple times to eliminate sampling jitter and local errors;

[0065] Finally, the voltage signal frequency that offsets local sampling errors and jitter interference is obtained, realizing anti-interference sampling.

[0066] Based on the pre-processed voltage signal, the initial frequency is calculated through the basic frequency estimation formula, and the amplitude is corrected using the spectrum leakage compensation formula to output the calculation result; the basic frequency estimation formula is:

[0067]

[0068] where f0 is the estimated basic frequency, N is the dynamically adjusted sampling point number, f s is the sampling frequency, φ is the fundamental phase angle, and dφ / dt is the change rate of the fundamental phase angle; the spectrum leakage compensation formula is:

[0069]

[0070] where A corrected is the corrected amplitude, A k is the original amplitude of the kth frequency point, A k+1 is the original amplitude of the k+1th frequency point, A k-1 is the original amplitude of the k-1th frequency point, and δ is the frequency offset rate;

[0071] For the output result, an improved dynamic calibration DFT technique is used to correct the voltage signal; under the adaptive window length adjustment mechanism, the Taylor expansion is used to compensate for the phase of the non-synchronous sampling error; the Fourier transform window length is dynamically adjusted according to the frequency change rate df / dt, the window length adjustment range of the adaptive window length adjustment mechanism is 25-100 cycles, and the specific rules of the adaptive window length adjustment mechanism are:

[0072]

[0073] wherein N new is the adjusted Fourier transform window length, N prev is the unadjusted Fourier transform window length, Δf is the frequency variation, and |Δf| is the absolute value of the frequency variation.

[0074] S2: fusing and calibrating the corrected voltage signal to obtain the calibrated real-time grid frequency.

[0075] The PMU data and the RTU data are fused with the SCADA synchronous phasor data through a federated Kalman filter, and the fused frequency data is output. The fused frequency data is calibrated in real time, and the calibrated real-time grid frequency is output.

[0076] The federated Kalman filter includes a PMU data local filter, an RTU data local filter, and a SCADA synchronous vector local filter. Each local filter outputs a local frequency state estimate value, which is uploaded to a fusion center for fusion. The fused state quantity is frequency data. The fusion formula of the federated Kalman filter is:

[0077]

[0078] wherein, is the fused state quantity at the kth moment, K1 is the weight coefficient of the PMU data, K2 is the weight coefficient of the RTU data, and K3 is the weight coefficient of the SCADA data. is the state estimate value of the PMU data, is the state estimate value of the RTU data, is the state estimate value of the SCADA data.

[0079]

[0080] wherein, K i is the weight coefficient of the ith type of data source, K1 is the weight coefficient of the PMU data, K2 is the weight coefficient of the RTU data, K3 is the weight coefficient of the SCADA data, σ1 is the confidence index of the PMU data, σ2 is the confidence index of the RTU data, and σ3 is the confidence index of the SCADA data.

[0081] The weight coefficient K i is dynamically adjusted according to the following rules: when the confidence index σ i of any data source decreases, the weight coefficient K i of the data source decreases synchronously; and when the confidence index σ i of any type of data source increases, the weight coefficient Ki Increase synchronously.

[0082] When performing real-time calibration on the fused data, the IEEE 1588 precision time protocol is used to correct the reference clock in real time, and the time error of the protocol is <1μs; a polynomial regression model is used to compensate for temperature drift, and the coefficient of determination R of the polynomial regression model is [value missing]. 2 >0.98, compensation temperature range is -40~85℃, output real-time grid frequency after calibration.

[0083] S3: Based on the calibrated real-time frequency of the power grid, perform operating condition clustering, identify the characteristics of the operating condition clustering results, obtain the frequency modulation control signal, and perform a first-order frequency modulation control based on the frequency modulation control signal.

[0084] The operating condition clustering based on the calibrated real-time frequency of the power grid includes performing an improved K-Medoids mining algorithm for operating condition clustering: the sampled voltage signals are weighted and calculated by the control system of the steam turbine generator set to obtain the total valve position value of all actual operating conditions, at least 250 data rows are selected as coarse initial centers of clusters, and all operating condition data rows are assigned to the nearest cluster to quickly reduce intra-cluster differences; the operating condition data row closest to the mean of all data rows in each cluster is used as the initial center of the cluster to improve robustness, and all operating condition data rows are again assigned to the nearest cluster to make the cluster structure more stable.

[0085] Iteratively perform the following operations to gradually optimize cluster centers and reduce overall error: Select the 50 non-cluster center data rows closest to the current cluster center, and calculate the total cost (TC) of replacing the current cluster center with each non-cluster center data row. ih If the minimum total cost is less than 0, the current cluster center is replaced with the corresponding non-cluster center data row, and all operating condition data rows are reassigned to the nearest cluster. When the cluster center no longer changes, the operating condition clustering results are output, providing an input basis for traffic characteristic identification. The feasibility of replacing the cluster center with non-cluster center data is determined by the total cost formula.

[0086]

[0087] Among them, TC ih non-cluster center O h Replace cluster center O i The total cost, where n is the total number of non-cluster center objects, and C jih For the j-th non-cluster center object O j The cost of replacement.

[0088] As a preferred embodiment of the primary frequency modulation control method based on big data and signal homology described in this invention, the improved K-Medoids mining algorithm has a cost C. jihThe computing rule is that only the distance between the new center and the second nearest center is compared to quickly determine whether the sample is replaced and calculate the cost, thereby maintaining the stability of the clustering result.

[0089] If O j belongs to the cluster represented by O i , when d(O j , O h ) ≥ d(O j , O j.2 ), C jih = d(O j , O j.2 )-d(O j , O i ).

[0090] When d(O j , O h ) < d(O j , O j.2 ), C jih = d(O j , O h )-d(O j , O i ).

[0091] If O j belongs to the cluster represented by O j.2 , when d(O j , O h ) ≥ d(O j , O j.2 ), C jih = 0.

[0092] When d(O j , O h ) < d(O j , O j.2 ), C jih = d(O j , O h )-d(O j , O j.2 ); wherein d is the Euclidean distance, O j.2 is the second nearest cluster center to O j .

[0093] The working condition clustering result is output to the unit control system after the working condition category is converted into quantifiable unit inlet characteristics by flow characteristic identification. The flow characteristic identification is realized based on the improved Froude formula in which the temperature parameter is replaced by the product of working fluid pressure and specific volume and the characteristic flow area method. The characteristic flow area method outputs the main steam flow G BThe unit inlet steam flow under different working conditions is calculated, the high-pressure cylinder inlet to the governing stage of the steam turbine is selected as the research stage, the characteristic flow area of the research stage is constant when the unit total valve position value and each valve opening degree are unchanged, the main steam throttle flow characteristic curve is generated, and the frequency modulation control signal is output.

[0094]

[0095] G A is the main steam flow under working condition A; G B is the main steam flow under working condition B; v 1A is the specific volume of the working medium in front of the stage under working condition A; v 1B is the specific volume of the working medium in front of the stage under working condition B; p 1A is the pressure of the working medium in front of the stage under working condition A; p 1B is the pressure of the working medium in front of the stage under working condition B; π A is the pressure ratio of the stage under working condition A; π B is the pressure ratio of the stage under working condition B; when the pressure ratio reaches a critical value, π A or π B is 0.

[0096] S4: in the process of collecting the voltage signal, a threshold method is used to trigger an alarm and cut off the bad point.

[0097] The signal same source frequency acquisition module is configured with an anti-interference unit, when the machine terminal A-phase voltage is lower than 86.6% of the rated secondary value, the anti-interference unit triggers a bad point alarm and sends an alarm signal to the distributed control system (DCS), automatically cuts off the signal same source module, and switches to the primary frequency modulation function of the digital electro-hydraulic control system (DEH) of the steam turbine; at the same time, the anti-interference unit uses multi-cycle frequency tracking technology, calculates the frequency every 20 ms, takes the average value of the continuously 5 configurable frequencies, and if the current frequency value deviates from the average value by more than ±0.05 Hz, the current frequency value is rejected.

[0098] The embodiment also provides a primary frequency modulation control system based on big data and signal homology, comprising: a signal homology frequency acquisition module; a signal preprocessing unit pre-processes a terminal voltage or a voltage signal output by a PMU; a DFT dynamic calculation unit corrects the pre-processed voltage signal; a multi-source data fusion unit fuses dynamic DFT calculation results, PMU data, RTU data and SCADA synchronous phasor data by using a federal Kalman filter, and outputs fused frequency data; a dynamic calibration unit calibrates the fused frequency data in real time, and outputs calibrated power grid real-time frequency; and a control strategy module based on big data analysis, comprising: a working condition clustering unit that improves a K-Medoids mining algorithm to cluster and classify collected unit operation data; and a flow characteristic identification unit that establishes a steam flow characteristic curve of a unit under different working conditions to provide an input basis for primary frequency modulation.

[0099] The embodiment also provides a computer device suitable for the primary frequency modulation control method based on big data and signal homology, comprising: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the primary frequency modulation control method based on big data and signal homology provided by the above embodiment.

[0100] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved by WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0101] The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for realizing primary frequency modulation control based on big data and signal homology proposed in the above embodiment; the storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk or an optical disk.

[0102] In summary, the present application has the following advantages:

[0103] The signal homology module collects frequency through PMU / terminal voltage, avoids mechanical delay of the traditional speed signal, eliminates sampling error, harmonic interference and environmental influence by combining dynamic DFT, multi-source fusion and dynamic calibration, makes the frequency measurement quickly track the real frequency of the power grid, provides effective support for the unit in the initial stage of frequency disturbance, and reduces the risk of frequency drop.

[0104] The bad point alarm switching mechanism of the anti-interference unit and the multi-cycle frequency tracking technology can resist local disturbance of the unit, random noise and voltage anomaly, and ensure that the frequency modulation only responds to the real demand of the power grid; the adaptive window length adjustment, temperature compensation and protocol compatibility design make the method adapt to complex working conditions such as harmonic interference, temperature change, island / network switching, and the stability is greatly improved.

[0105] Reduce invalid disturbance of the unit: the big data module accurately distinguishes valid / invalid disturbance by improving the K-Medoids algorithm, only dynamically increases the frequency modulation amplitude for valid disturbance, avoids excessive adjustment for invalid disturbance, and ensures that the adjustment amplitude matches the actual steam admission capacity of the unit by generating accurate flow characteristic curves through feature through-flow area identification, reduces the shock of the high-pressure regulating valve of the steam turbine, and balances the primary frequency modulation index of the power grid and the safe and economic operation of the unit.

[0106] Improve the Frugal formula and the high-pressure cylinder stage selection, so that the flow characteristic identification is suitable for any type of unit; dynamic weight distribution of federal Kalman filter and iterative optimization of K-Medoids ensure that the data and clustering results of each link are reliable, provide accurate protection for primary frequency modulation, and adapt to the frequency stability control demand of the power grid with high penetration rate of new energy.

[0107] The problem of unstable frequency control of a new energy high-penetration power grid is solved.

[0108] It should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A primary frequency modulation control method based on big data and signal homology, characterized in that: Acquire voltage signals, preprocess the terminal voltage and the voltage signal output by the PMU, and correct the voltage signals; The corrected voltage signals are fused and calibrated to obtain the calibrated real-time power grid frequency. Based on the calibrated real-time frequency of the power grid, operating condition clustering is performed. By identifying the characteristics of the operating condition clustering results, a frequency modulation control signal is obtained, and a first-order frequency modulation control is performed based on the frequency modulation control signal. During the voltage signal acquisition process, a threshold method is used to trigger alarms and remove faulty pixels.

2. The primary frequency modulation control method based on big data and signal homology as described in claim 1, characterized in that: The preprocessing and correction of the terminal voltage or PMU output voltage signal includes sampling the voltage signal using a 16-bit analog-to-digital converter; filtering the sampled signal using an 8th-order Butterworth bandpass filter with a passband frequency range of 45-55Hz; and then using a sliding interquartile range (IIL) method to remove outliers from the filtered signal, with a window length of 10 cycles. The corrected voltage signal includes, based on the preprocessed voltage signal, calculating the initial frequency using the basic frequency estimation formula, correcting the amplitude using the spectral leakage compensation formula, and outputting the calculation result. The basic frequency estimation formula is as follows: Where f0 is the estimated fundamental frequency, N is the number of dynamically adjusted sampling points, and f s The sampling frequency is φ, the fundamental phase angle is dφ / dt, and the rate of change of the fundamental phase angle is dφ / dt. The spectral leakage compensation formula is: Among them, A corrected For the corrected amplitude, A k Let A be the original amplitude at the k-th frequency point. k+1 For the original amplitude at the (k+1)th frequency point, A k-1 Let δ be the original amplitude at the (k-1)th frequency point, and δ be the frequency offset rate. The output is corrected using an improved dynamic calibration DFT technique. This improved DFT technique includes phase compensation for asynchronous sampling errors via Taylor expansion under an adaptive window length adjustment mechanism. The window length adjustment range of the adaptive window length adjustment mechanism is 25-100 cycles, and the specific rules of the adaptive window length adjustment mechanism are as follows: Where, N new N is the adjusted Fourier transform window length. prev Δf is the length of the Fourier transform window before adjustment, Δf is the frequency change, and |Δf| is the absolute value of the frequency change.

3. The primary frequency modulation control method based on big data and signal homology as described in claim 2, characterized in that: The fusion and calibration of the corrected voltage signal includes fusing PMU data and RTU data with SCADA synchronization phasor data through a federated Kalman filter, outputting fused frequency data, calibrating the fused frequency data in real time, and outputting the calibrated real-time grid frequency. The federated Kalman filter includes a PMU data local filter, an RTU data local filter, and a SCADA synchronization vector local filter. Each local filter outputs a local frequency state estimate, which is then uploaded to a fusion center for fusion. The fused state variable is the frequency data. The local state estimates are then weighted and fused to obtain the fused state variable. The fusion formula of the federated Kalman filter is: in, Let K1 be the weighting coefficient of the PMU data, K2 be the weighting coefficient of the RTU data, and K3 be the weighting coefficient of the SCADA data after fusion at time k. The state estimate of the PMU data. The state estimate of the RTU data. This is a state estimate of the SCADA data; Among them, K i K1 represents the weight coefficient of the i-th type of data source, K2 represents the weight coefficient of the PMU data, K3 represents the weight coefficient of the RTU data, K3 represents the weight coefficient of the SCADA data, σ1 represents the confidence index of the PMU data, σ2 represents the confidence index of the RTU data, and σ3 represents the confidence index of the SCADA data. Weighting coefficient K i The dynamic adjustment rule applies when the confidence index σ of any data source... i When the weighting factor K of the data source decreases, i Synchronously decrease; when the confidence index σ of any type of data source decreases i When the weight coefficient K of the data source increases, i Increase synchronously.

4. The primary frequency modulation control method based on big data and signal homology as described in claim 2, characterized in that: The corrected voltage signal includes an anti-harmonic interference enhancement algorithm implemented by reconstructing the fundamental component and setting a hysteresis threshold during voltage signal correction calculation; firstly, a harmonic matrix h containing the 6th harmonic is constructed. matrix Then, based on the least squares method, the linear equations between the harmonic matrix and the input signal are solved to obtain the least squares solution l. ssolution The coefficients of each harmonic component are determined, and the fundamental component is reconstructed based on the first two columns of the harmonic matrix and the first two elements of the least squares solution. An adaptive zero-crossing detection algorithm with a hysteresis value of 0.005 is used to identify the zero-crossing points of the fundamental component. crossings False zero-crossings caused by noise or residual harmonics are eliminated, and then the time interval between adjacent zero-crossing points is calculated. The average of multiple time intervals is taken to obtain the average time interval T. avg The frequency of the voltage signal is calculated based on the average time interval, using the formula f = 1 / (2T). avg The final voltage signal frequency f is calculated to compensate for local sampling errors and jitter interference, thus achieving anti-interference sampling.

5. The primary frequency modulation control method based on big data and signal homology as described in claim 2, characterized in that: The operating condition clustering based on the calibrated real-time grid frequency includes performing an improved K-Medoids mining algorithm for operating condition clustering: the sampled voltage signals are weighted and calculated by the control system of the steam turbine generator set to obtain the total valve position value of all actual operating conditions; at least 250 data rows are selected as coarse initial centers of clusters; all operating condition data rows are assigned to the nearest cluster to quickly reduce intra-cluster differences; the operating condition data row closest to the mean of all data rows in each cluster is used as the initial center of the cluster to improve robustness; all operating condition data rows are again assigned to the nearest cluster to make the cluster structure more stable. Iteratively perform the following operations to gradually optimize cluster centers and reduce overall error: Select the 50 non-cluster center data rows closest to the current cluster center, and calculate the total cost (TC) of replacing the current cluster center with each non-cluster center data row. ih If the minimum total cost is less than 0, the current cluster center is replaced with the corresponding non-cluster center data row, and all operating condition data rows are reassigned to the nearest cluster. When the cluster center no longer changes, the operating condition clustering results are output to provide an input basis for traffic characteristic identification. Determine the feasibility of replacing cluster centers with non-cluster center data using the total cost formula: Among them, TC ih non-cluster center O h Replace cluster center O i The total cost, where n is the total number of non-cluster center objects, and C jih For the j-th non-cluster center object O j The cost of replacement.

6. The primary frequency modulation control method based on big data and signal homology as described in claim 5, characterized in that: The operating condition clustering based on the calibrated real-time power grid frequency includes improving the K-Medoids mining algorithm at a cost C. jih The calculation rule only compares the distance between the new center and the second nearest center to quickly determine whether the sample has changed its affiliation and calculate the cost, thus maintaining the stability of the clustering results; If O j Belongs to center O i The represented cluster: when d(O j O h )≥d(O j O j.2 ), C jih =d(O j O j.2 )-d(O j O i ); This d(O j ,O h )<d(O j ,O j.2 ), C jih = d(O j ,O h )-d(O j ,O i ); If O j Belongs to the sub-center O j.2 The represented cluster: when d(O j O h )≥d(O j O j.2 ), C jih =0; When d(O) j O h )<d(O j O j.2 ), C jih =d(O j O h )-d(O j O j.2 ); where d is the Euclidean distance, O j.2 Distance O j The second nearest cluster center.

7. The primary frequency modulation control method based on big data and signal homology as described in claim 5, characterized in that: The characteristic identification of the operating condition clustering results includes: after the operating condition clustering results are identified by flow characteristics, the operating condition categories are converted into quantifiable unit steam inlet characteristics, and then the frequency modulation control signal is output to the unit control system. The flow characteristic identification is based on the improved Flueger formula, which replaces the temperature parameter with the product of the working fluid pressure and specific volume, and the characteristic flow area method. The characteristic flow area method is based on the main steam flow rate G output by the improved Flueger formula. B Calculate the steam intake of the unit under different operating conditions, generate the main steam regulating valve flow characteristic curve, and output frequency regulation control signal; Among them, G A G represents the main steam flow rate under operating condition A; B v represents the main steam flow rate under operating condition B; 1A v is the specific volume of the working fluid before the next stage group in operating condition A; 1B p is the specific volume of the working fluid before the next stage group in operating condition B; 1A The pressure of the working fluid before the next stage group in operating condition A; p 1B The pressure of the working fluid before the next stage group in condition B; π A The pressure ratio of the lower group under operating condition A; π B The pressure ratio of the lower group under operating condition B; when the pressure ratio reaches the critical value, π A or π B Take 0.

8. A primary frequency modulation control system based on big data and signal homology, based on the primary frequency modulation control method based on big data and signal homology as described in any one of claims 1 to 7, characterized in that, include: Signal source frequency acquisition module: The signal preprocessing unit preprocesses the terminal voltage or the voltage signal output by the PMU; The DFT dynamic calculation unit corrects the preprocessed voltage signal; the multi-source data fusion unit uses a federated Kalman filter to fuse the dynamic DFT calculation results, PMU data, RTU data, and SCADA synchronization phasor data, and outputs the fused frequency data; the dynamic calibration unit performs real-time calibration on the fused frequency data and outputs the calibrated real-time grid frequency; the control strategy module based on big data analysis includes: an improved K-Medoids mining algorithm-based operating condition clustering unit that clusters and classifies the collected unit operating data; and a flow characteristic identification unit that establishes steam flow characteristic curves of the unit under different operating conditions, providing an input basis for primary frequency regulation. The signal source frequency acquisition module is equipped with an anti-interference unit. When the voltage of phase A at the terminal is lower than 86.6% of the rated value, an alarm is triggered and the frequency is switched to the DEH primary frequency modulation. A multi-cycle frequency tracking function is set up, which calculates the frequency every 20ms. If the current value deviates from the average value of five consecutive points by more than ±0.05Hz, the value is discarded.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the primary frequency modulation control method based on big data and signal homology as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the primary frequency modulation control method based on big data and signal homology as described in any one of claims 1 to 7.

Images