Broadband oscillation identification and early warning method and system based on composite spectrum difference measurement

Abstract

According to the broadband oscillation identification and early warning method and system based on the composite spectrum difference measurement, real-time operation data of all nodes of a power grid are obtained, and a time sequence spectrum is obtained; when the oscillation alarm is relieved and steady-state operation in a set time period is passed, taking a frequency spectrum of a statistical average value of historical operation data in a sliding window in the set time period as a reference frequency spectrum; a frequency spectrum mask is introduced to represent a concerned frequency band; calculating the distance between the time sequence frequency spectrum and the reference frequency spectrum in the concerned frequency band, determining the KL divergence of the time sequence frequency spectrum and the reference frequency spectrum in the concerned frequency band according to the energy of the reference frequency spectrum in the concerned frequency band, and constructing a composite frequency spectrum difference index by using the distance and the KL divergence; taking energy and energy variation of the time sequence spectrum in a concerned frequency band as an energy cumulative evolution index; based on the composite spectrum difference index and the energy cumulative evolution index, early warning or warning is carried out after broadband oscillation is identified, and the monitoring sensitivity, the early warning accuracy, the early symptom capturing capability and the adaptability to the dynamic characteristics of the spectrum are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power grid broadband oscillation identification, and particularly relates to a broadband oscillation identification and early warning method and system based on composite frequency spectrum difference measurement. BACKGROUND

[0002] With the rapid development of new energy power generation technology and the increasing penetration rate of power electronic equipment in the power grid, the operation characteristics and dynamic behavior of modern power systems show a significant trend of complexity and diversification. Under this background, broadband oscillation phenomenon has become one of the key challenges to the safe and stable operation of the power grid.

[0003] Existing broadband oscillation monitoring and early warning technologies mostly rely on static threshold comparison of the amplitude of specific frequency points or frequency bands. Such methods usually preset fixed amplitude alarm and duration thresholds, and trigger an alarm when the amplitude of the monitored signal exceeds the threshold. This kind of method is relatively simple to implement, but it lacks the ability to perceive the dynamic evolution of the overall structure of the frequency spectrum, and it is difficult to effectively capture subtle but critical signs such as oscillation mode migration, gradual drift of dominant frequency, and redistribution of energy among different frequency bands in the early development stage of broadband oscillation. Therefore, this kind of method has inherent limitations in the timeliness and accuracy of early warning, missing the best intervention opportunity and thus developing into an oscillation risk.

[0004] In practical applications, the existing broadband oscillation monitoring method mainly based on fixed frequency point amplitude static threshold has a slow perception of overall structural changes in the frequency spectrum: simply relying on whether the amplitude of the fixed frequency point exceeds the limit as the basis for judgment cannot effectively identify the drift of the oscillation dominant frequency with the change of the operating condition, the redistribution of oscillation energy in the frequency domain, and the overall distortion of the frequency spectrum pattern. This leads to insufficient sensitivity to early evolution characteristics of oscillation and limited early warning capability; a dilemma between false positives and false negatives: in the early stage of oscillation or when the disturbance is weak, to avoid frequent false positives caused by normal operating fluctuations, the system is often forced to use a higher alarm threshold. However, this directly sacrifices the early identification capability of some broadband oscillations with slow amplitude growth or hidden manifestations, resulting in a risk of false negatives. Conversely, if the threshold is lowered to pursue sensitivity, the false positive rate will be significantly increased, causing unnecessary interference to the dispatching and operating personnel and affecting the overall operational availability; lack of dynamic focusing ability on key risk frequency bands: the risk of some broadband oscillations often concentrates in specific frequency ranges that dynamically change with the system state. The existing full-band indiscriminate monitoring or fixed-band monitoring methods cannot achieve sensitive identification and accurate tracking of trends in these high-risk areas. SUMMARY

[0005] In order to solve the problems in the prior art, the application provides a wide-frequency oscillation identification and early warning method and system based on composite spectrum difference measurement, which overcomes the deficiencies of the prior wide-frequency oscillation monitoring and early warning method in monitoring sensitivity, early warning accuracy, early symptom capturing capability and adaptability to spectrum dynamic characteristics.

[0006] The application adopts the following technical scheme.

[0007] The application provides a wide-frequency oscillation identification and early warning method based on composite spectrum difference measurement, which comprises the following steps:

[0008] Real-time operation data of each node of the power grid are acquired to obtain time series spectrum;

[0009] During a historical oscillation event of the power grid, when the oscillation alarm is removed and steady-state operation is performed for a set period, the spectrum of the statistical average value of the historical operation data in the sliding window in the set period is taken as a reference spectrum;

[0010] A spectrum mask is introduced to represent a frequency band of interest; the distance between the time series spectrum and the reference spectrum in the frequency band of interest is calculated, the KL divergence between the time series spectrum and the reference spectrum in the frequency band of interest is determined according to the energy of the reference spectrum in the frequency band of interest, and the composite spectrum difference index is constructed by using the distance and the KL divergence;

[0011] The energy of the time series spectrum in the frequency band of interest and the energy change amount are taken as the energy accumulation evolution index;

[0012] Based on the composite spectrum difference index and the energy accumulation evolution index, wide-frequency oscillation identification, early warning or alarm is performed.

[0013] A spectrum mask M(f) is set, wherein M(f) = 1 represents that the frequency is in the frequency band of interest, and M(f) = 0 represents that the frequency is not in the frequency band of interest;

[0014] Based on the spectrum mask, the distance between the time series spectrum and the reference spectrum in the frequency band of interest within the period t is calculated;

[0015] Based on the spectrum mask, the probability density function of the time series spectrum and the reference spectrum in the frequency band of interest within the period t is determined;

[0016] According to the energy of the reference spectrum in the frequency band of interest, the KL divergence between the time series spectrum and the reference spectrum in the frequency band of interest within the period t is calculated by using the probability density function of the time series spectrum and the reference spectrum in the frequency band of interest within the period t;

[0017] The distance and the KL divergence between the time series spectrum and the reference spectrum in the frequency band of interest within the period t are normalized;

[0018] The composite spectrum difference index within the period t is constructed by using the normalized distance and the KL divergence.

[0019]

[0020] where DE M (t) is the distance between the time series spectrum and the reference spectrum in the concerned frequency band at time t, S(t,f) is the time series spectrum in the time period t, S ref (f) is the reference spectrum, and f is the frequency band.

[0021]

[0022] where P'(t,f) and P(t,f) are the probability density function and the probability distribution function of the time series spectrum in the time period t in the concerned frequency band, respectively, P'(t,f) = P(t,f) = S(t,f) / N ref (t,f) and P ref (t,f) are the probability density function and the probability distribution function of the reference spectrum in the time period t in the concerned frequency band, respectively, and ε is a zero probability term for smoothing, and N M is the number of frequency points in the concerned frequency band.

[0023] The energy of the reference spectrum in the concerned frequency band is calculated as If the energy is 0, the KL divergence between the time series spectrum and the reference spectrum in the concerned frequency band in the time period t is set to 0, otherwise it is calculated as

[0024]

[0025] where DKL M (t) is the KL divergence between the time series spectrum and the reference spectrum in the concerned frequency band in the time period t.

[0026] According to the normalized distance and the KL divergence , the variance σ 2 is used to dynamically adjust the weights of and , as follows:

[0027] w KL (t) = 1 - w E (t)

[0028] where w E (t) and w KL (t) are the weights of and in the time period t, respectively.

[0029]

[0030] where CSDI(t) is the composite spectrum difference index in the time period t.

[0031]

[0032] E(t) = ∑S(t,f)df M (t) is the energy of time series spectrum in the concerned frequency band in time period t, S(t,f) is the time series spectrum in time period t, and f is the frequency band;

[0033] E(t) is the energy of time series spectrum in the concerned frequency band in time period t M (t), and ΔE M (t) is the energy accumulation evolution index.

[0034] Based on the composite spectrum difference index and the energy accumulation evolution index, the wideband oscillation is identified to give a pre-warning or an alarm, including:

[0035] When the following two conditions are met simultaneously, the wideband oscillation pre-warning mode is activated:

[0036] 1) Condition 1: when ΔCSDI(t) > θ CSDI,alert , ΔCSDI(t) is the mutation of the composite spectrum difference index in time period t, θ CSDI,alert is the set spectrum difference threshold;

[0037] 2) Condition 2: when E M (t) > θ E,alert or ΔE M (t) > θ ΔE,alert , θ E,alert and θ ΔE,alert are respectively the set energy abnormal threshold and the set energy change abnormal threshold;

[0038] The distribution characteristics of CSDI in the steady operation are counted, the wideband difference allowance of the time series spectrum and the reference spectrum in the concerned frequency band is determined by simulation according to the power grid operation state and control requirements, and the spectrum difference threshold is determined according to the distribution characteristics of CSDI in the steady operation and the wideband difference allowance.

[0039] The distribution characteristics of ECEI in the steady operation are counted, the energy upper limit and the energy fluctuation allowance of the time series spectrum in the concerned frequency band are determined by simulation according to the power grid operation state and control requirements, and the energy abnormal threshold and the energy change abnormal threshold are determined according to the distribution characteristics of ECEI in the steady operation and the energy upper limit and the energy fluctuation allowance.

[0040] If the wideband oscillation pre-warning mode is activated, when A M (t) > θ A,warning and lasts for ΔT warning or when CSDI(t) > θ CSDI,warning and lasts for ΔT warningA M (t) is the amplitude of the time series spectrum in the concerned frequency band in time period t, θ A,warning is the set amplitude warning threshold, θ CSDI,warning is the set broadband oscillation warning threshold, ΔT warning is the set warning duration;

[0041] The amplitude warning threshold θ A,warning is 1.1 times the average amplitude of the reference frequency in the concerned frequency band, and the broadband oscillation warning threshold θ CSDI,warning is 1+θ CSDI,alert ; the warning duration is determined according to the fault recording time length configured by the broadband measurement device.

[0042] Regardless of whether the broadband oscillation warning mode is activated, when A M (t) > θ A,alarm and lasts for ΔT alarm or when CSDI(t) > θ CSDI,alarm and lasts for ΔT alarm , a broadband oscillation warning signal is sent out; wherein θ A,alarm is the set amplitude warning threshold, θ CSDI,alarm is the set broadband oscillation warning threshold, ΔT alarm is the set warning duration;

[0043] The amplitude warning threshold θ A,alarm is 1.2 times the average amplitude of the reference frequency in the concerned frequency band, and the broadband oscillation warning threshold θ CSDI,alarm is 1.1 times the broadband oscillation warning threshold θ CSDI,warning ; the warning duration is determined according to the fault recording time length configured by the broadband measurement device.

[0044] The application also proposes a broadband oscillation identification and warning system based on composite spectrum difference measurement, comprising:

[0045] a spectrum acquisition module, configured to acquire real-time operation data of each node of the power grid to obtain a time series spectrum; during a historical oscillation event of the power grid, after oscillation warning is removed and steady-state operation for a set time period is passed, a spectrum of a statistical average value of historical operation data in a sliding window in the set time period is taken as a reference spectrum;

[0046] an index establishment module, configured to introduce a spectrum mask to represent a concerned frequency band; to calculate a distance between the time series spectrum and the reference spectrum in the concerned frequency band, to determine a KL divergence between the time series spectrum and the reference spectrum in the concerned frequency band according to energy of the reference spectrum in the concerned frequency band, and to construct a composite spectrum difference index by using the distance and the KL divergence; to take energy of the time series spectrum in the concerned frequency band and an energy change amount as an energy accumulation evolution index;

[0047] oscillation early warning module, for identifying and early warning or alarming based on the composite frequency spectrum difference index and the energy accumulation evolution index after wide frequency oscillation.

[0048] The application also relates to a terminal, which comprises a processor and a storage medium; the storage medium is used for storing instructions; and the processor is used for operating according to the instructions to execute the steps of the method.

[0049] The application also relates to a computer readable storage medium, which stores a computer program; the program is executed by a processor to realize the steps of the method.

[0050] Compared with the prior art, the method provided by the application at least has the following beneficial effects: the method constructs a composite frequency spectrum difference index representing the amplitude of a frequency spectrum and the distribution of energy forms, an energy accumulation evolution index representing the instantaneous response of a spectrum mutation and the cumulative effect of energy fluctuation, and introduces a spectrum mask to focus on a key frequency band, and establishes a dynamically adjusted early warning and alarming criterion, thereby significantly improving the early warning capability of wide frequency oscillation, realizing accurate and real-time monitoring of possible phenomena in the development process of wide frequency oscillation, such as dominant frequency drift, spectrum energy distribution mode mutation and abnormal activity in a specific high-risk frequency band, effectively reducing the risk of false positives and false negatives, and providing stronger technical support for the safe and stable operation of a power system. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 is a flowchart of a wide frequency oscillation identification and early warning method based on a composite frequency spectrum difference measurement proposed by the application;

[0052] Figure 2 is a principle diagram for realizing risk frequency band focusing based on a spectrum mask in an embodiment of the application;

[0053] Figure 3 is a structural block diagram of a wide frequency oscillation grading early warning system based on a spectrum mask in an embodiment of the application. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical scheme and advantages of the application clearer, the technical scheme of the application will be described clearly and completely below in combination with the drawings in the embodiments of the application. The embodiments described in the application are only a part of the embodiments of the application, rather than all the embodiments. Based on the spirit of the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0055] The application proposes a wide frequency oscillation identification and early warning method based on a composite frequency spectrum difference measurement, as shown in Figure 1 The method comprises the following steps:

[0056] Step 1, obtain the real-time operation data of each node of the power grid to obtain the time series spectrum.

[0057] In the embodiment, the wideband measurement device, the synchronized phasor measurement unit (PMU) or the fault recording device is used to synchronously and real-timely collect the multi-channel electrical signals of the key nodes of the power system. The wideband measurement device can directly output the spectrum data and can be directly applied. The other devices need to extract the recording data, and the continuous short-time Fourier transform (STFT) or other time-frequency analysis methods are performed on the collected time domain signals to construct the time series spectrum matrix S(t, f) which can reflect the system dynamics, wherein t is the time and f is the frequency. The sequence constitutes the basic data for all subsequent analyses.

[0058] Step 2, during the process of the historical oscillation event of the power grid, when the oscillation alarm is removed and the steady-state operation after the set period is passed, the spectrum of the statistical average value of the historical operation data in the sliding window in the set period is taken as the reference spectrum.

[0059] During the process of the historical oscillation event of the power grid, when the oscillation alarm is removed and the steady-state operation after the set period is passed, the spectrum of the statistical average value of the historical operation data in the sliding window in the set period is taken as the reference spectrum (Reference Spectrum).

[0060] In the embodiment, the length of the sliding window is 300s; when the oscillation alarm is removed and the steady-state operation after the set period is passed, the spectrum of the statistical average value of the real-time operation data in the sliding window in the set period is taken as the reference spectrum S ref (f), the reference spectrum is automatically updated according to the change of the operation state of the system restored to the steady state; wherein the set period is 10 minutes continuously; the obtained reference spectrum can reflect the slow change of the system background noise, so that the spectral mask can be generated based on the neural network, and the concept of the spectral mask is introduced to focus on the specific frequency interval which is highly related to the wideband oscillation risk.

[0061] Step 3, the distance and the KL divergence of the time series spectrum and the reference spectrum in the concerned frequency band are determined by introducing the spectral mask to construct the composite spectrum difference index.

[0062] Specifically, step 3 includes:

[0063] Step 3.1, set the spectral mask M(f), M(f) = 1 represents that the frequency is in the concerned frequency band, and M(f) = 0 represents that the frequency is not in the concerned frequency band;

[0064] The spectrum mask M(f) is a weight vector or a Boolean vector corresponding to the frequency axis, which is used to highlight or select specific frequency ranges when calculating the spectral difference. The spectrum mask will be applied to the subsequent Euclidean distance and KL divergence calculation, so that the difference measure is more focused on the changes in these concerned frequency bands. As shown in Figure 2 The masked spectrum S(t, f)M(f) is obtained by applying the spectrum mask to the original time series spectrum S(t, f), focusing on the risk frequency band, while multiple concerned frequency bands do not overlap with each other.

[0065] Step 3.2, based on the spectrum mask, calculate the distance between the time series spectrum and the reference spectrum in the concerned frequency band within the time period t, which satisfies the following relationship:

[0066]

[0067] In the formula, DE M (t) is the distance between the time series spectrum and the reference spectrum at time t, S(t, f) is the time series spectrum within the time period t, S ref (f) is the reference spectrum, and f is the frequency band.

[0068] Before introducing the spectrum mask, it is necessary to traverse all discrete frequency points for summation, which has a large amount of calculation and cannot exclude the influence of the frequency points not concerned on the calculation result. After introducing the spectrum mask, only the frequency points in the concerned frequency band need to be considered, reducing the amount of calculation and highlighting the difference of specific frequency points.

[0069] Step 3.3, based on the spectrum mask, determine the probability density function of the time series spectrum and the reference spectrum in the concerned frequency band within the time period t, as follows:

[0070]

[0071] In the formula, P'(t, f) and P(t, f) are the probability density function and the probability distribution function of the time series spectrum in the concerned frequency band, respectively, P'(t, f) is the probability density function of the time series spectrum in the concerned frequency band, and P(t, f) is the probability distribution function of the time series spectrum in the concerned frequency band. ref (t, f) and P ref (t, f) are the probability density function and the probability distribution function of the reference spectrum in the concerned frequency band, respectively, ε is a zero probability term for smoothing, which is a fixed minimum positive number in the embodiment, and N M is the number of frequency points in the concerned frequency band.

[0072] Step 3.4, according to the energy of the reference spectrum in the concerned frequency band, the probability density function of the time series spectrum and the reference spectrum in the concerned frequency band is used to calculate the KL divergence of the time series spectrum and the reference spectrum in the concerned frequency band.

[0073] Wherein, where DKL (t) is the KL divergence between the time series spectrum and the reference spectrum in the concerned frequency band at time t, and is calculated as follows:

[0074]

[0075] where DKL M

[0076] The KL divergence between the time series spectrum and the reference spectrum in the concerned frequency band is used to reflect the difference in the probability distribution patterns of the two spectra in the high-risk frequency band.

[0077] Step 3.5, normalize the distance and the KL divergence between the time series spectrum and the reference spectrum in the concerned frequency band.

[0078] The calculated DE M (t) and DKL M (t) are statistically analyzed and normalized to eliminate the dimension effect and improve the comparability.

[0079] In the embodiment, the percentile (e.g., P95) is used for normalization, as follows:

[0080]

[0081] where DE and DKL are the normalized values of the distance and the KL divergence between the time series spectrum and the reference spectrum at time t, respectively, min(DE M_W (t)) and min(DKL M_W (t)) are the minimum values of the distance sequence DE M_W and the KL divergence sequence DKL M_W in the sliding window, respectively, and P95() is a function for calculating the 95th percentile.

[0082] Step 3.6, construct the composite spectral difference index (CSDI) in the time period t using the normalized distance DE and the KL divergence DKL , including:

[0083] According to the fluctuation of the normalized distance DE and the KL divergence DKL in the sliding window, the variance σ 2 is used to measure the fluctuation in the embodiment, and σ ​the weight of the energy change amount ΔE

[0084] w KL (t) = 1 - w E (t)

[0085] wherein w E (t), w KL (t) are the weights of the energy and the energy change amount ΔE and in the time period t, respectively;

[0086] The CSDI in the time period t is constructed in the following relationship:

[0087]

[0088] The CSDI constructed by the present application can comprehensively reflect the complex differences of the spectrum in amplitude and morphology.

[0089] Step 4, obtaining the energy and the energy change amount of the reference spectrum in the frequency band of interest as the energy cumulative evolution indicator.

[0090] The energy cumulative evolution indicator (ECEI) comprises: the energy of the time series spectrum in the time period t in the frequency band of interest the energy change amount ΔE M (t);

[0091] The ECEI indicator not only reflects the current energy level, but also emphasizes the cumulative positive growth trend of the energy in a period of time. The ECEI indicator is the sliding integral of the energy change rate, or the cumulative amount of energy exceeding a certain dynamic baseline. This helps to capture those oscillations that develop relatively slowly but have sustained growth in energy.

[0092] Step 5, identifying wideband oscillations based on the complex spectrum difference indicator and the energy cumulative evolution indicator, and then performing early warning or alarm.

[0093] Specifically, step 5 comprises:

[0094] Step 5.1, when the following two conditions are met simultaneously, activate the wideband oscillation early warning mode:

[0095] Condition 1, real-time monitoring of the mutation amount of the complex spectrum difference indicator, when ΔCSDI(t) > θ CSDI,alert , it is determined that the time series spectrum and the reference spectrum have wideband abnormalities in the frequency band of interest, wherein ΔCSDI(t) is the mutation amount of the complex spectrum difference indicator in the time period t, θ CSDI,alert is the set spectrum difference threshold;

[0096] In the embodiment, the mutation quantity of the CSDI is monitored, and the rate of change of the CSDI monitored in real time or the deviation of the CSDI monitored in real time from the mean value is taken as the mutation quantity; when the mutation quantity is greater than a set spectrum difference threshold θ CSDI,alert , that is, the CSDI has a significant and rapid jump, it is determined that the time series spectrum has a mutation; the CSDI integrates the comprehensive changes of the spectrum amplitude and the distribution form, and a large increase in the value thereof represents that there is a significant wideband anomaly in the concerned frequency band;

[0097] In the embodiment, for a target power grid, long-term data under actual operation of a wideband measurement device / PMU are collected, and the distribution characteristics of the CSDI in a stable operation are counted; according to the operation state and control requirements of the power grid, a simulation method is used to determine the wideband difference allowable value of the time series spectrum in the concerned frequency band from the reference spectrum; and according to the distribution characteristics of the CSDI in the stable operation and the wideband difference allowable value, the spectrum difference threshold is determined.

[0098] Condition 2, the energy accumulation evolution index is monitored in real time; when E M (t)>θ E,alert or ΔE M (t)>θ ΔE,alert , it is determined that the time series spectrum has an energy anomaly in the concerned frequency band; wherein θ E,alert , θ ΔE,alert are respectively an energy anomaly threshold and an energy change anomaly threshold;

[0099] The ECEI not only reflects the current energy level, but also emphasizes the cumulative positive growth trend of the energy in a period of time, and a large increase in the value or change thereof represents that there is a significant energy anomaly in the time series spectrum in the concerned frequency band;

[0100] In the embodiment, for a target power grid, long-term data under actual operation of a wideband measurement device / PMU are collected, and the distribution characteristics of the ECEI in a stable operation are counted; according to the operation state and control requirements of the power grid, a simulation method is used to determine the energy upper limit and energy fluctuation allowable value of the time series spectrum in the concerned frequency band; and according to the distribution characteristics of the ECEI in the stable operation and the energy upper limit and energy fluctuation allowable value, the energy anomaly threshold and the energy change anomaly threshold are determined.

[0101] Step 5.2, if the wideband oscillation pre-warning mode is activated, when A M (t)>θ A,warning and lasts for ΔT warning or when CSDI(t)>θ CSDI,warning and lasts for ΔT warning , a wideband oscillation pre-warning signal is sent out;

[0102] wherein A M(t) is the amplitude of the time series spectrum in the concerned frequency band in time period t, θ A,warning is the set amplitude warning threshold, θ CSDI,warning is the set broadband oscillation warning threshold, ΔT warning is the set warning duration;

[0103] In the embodiment, the preferred value of the amplitude warning threshold θ A,warning is 1.1 times the average amplitude of the reference frequency in the concerned frequency band, and the preferred value of the broadband oscillation warning threshold θ CSDI,warning is 1+θ CSDI,alert ; the warning duration is determined according to the fault recording time length configured by the broadband measurement device, and the preferred value is 1 min.

[0104] Step 5.3, whether the broadband oscillation warning mode is activated or not, when A M (t) > θ A,alarm and lasts for ΔT alarm or when CSDI(t) > θ CSDI,alarm and lasts for ΔT alarm , a broadband oscillation warning signal is sent out;

[0105] wherein, θ A,alarm is the set amplitude warning threshold, θ CSDI,alarm is the set broadband oscillation warning threshold, ΔT alarm is the set warning duration;

[0106] In the embodiment, the preferred value of the amplitude warning threshold θ A,alarm is 1.2 times the average amplitude of the reference frequency in the concerned frequency band, and the preferred value of the broadband oscillation warning threshold θ CSDI,alarm is 1.1 times θ CSDI,warning ; the warning duration is determined according to the fault recording time length configured by the broadband measurement device, and the preferred value is 30 s.

[0107] The present application adopts an enhanced joint mechanism of dynamic warning and alarm considering energy collaborative change, firstly judges whether to activate the broadband oscillation warning mode, monitors the amplitude or CSDI of the time series spectrum in the concerned frequency band in the broadband oscillation warning mode, and when the amplitude or CSDI exceeds the set threshold and lasts for a recording time length, a warning signal is sent out; at the same time, when the amplitude or CSDI of the time series spectrum in the concerned frequency band is monitored to grow dramatically and exceeds the set threshold and lasts for a period of time, the set threshold used by the alarm monitoring is significantly greater than the device threshold used by the warning monitoring, it can be determined that the current broadband oscillation is serious, and the broadband oscillation is so serious that it lasts for no more than a fault recording time length, and immediate alarm is needed whether in the broadband oscillation warning mode or not.

[0108] This invention also proposes a broadband oscillation identification and early warning system based on composite spectral difference measurement, comprising:

[0109] The spectrum acquisition module is used to acquire real-time operating data of each node in the power grid to obtain the time series spectrum. During the historical oscillation event of the power grid, after the oscillation alarm is cleared and the grid has been in steady-state operation for a set period, the spectrum of the statistical average value of the historical operating data within the sliding window of the set period is used as the reference spectrum.

[0110] The index establishment module is used to introduce a spectral mask to represent the frequency band of interest; calculate the distance between the time series spectrum and the reference spectrum in the frequency band of interest; determine the KL divergence between the time series spectrum and the reference spectrum in the frequency band of interest based on the energy of the reference spectrum in the frequency band of interest; construct a composite spectral difference index using the distance and KL divergence; and use the energy of the time series spectrum in the frequency band of interest and the amount of energy change as an energy accumulation evolution index.

[0111] The oscillation early warning module is used to identify broadband oscillations and issue early warnings or alarms based on composite spectral difference indicators and energy accumulation evolution indicators.

[0112] In the embodiments, such as Figure 3 As shown, the broadband oscillation hierarchical early warning system based on spectrum mask is also equipped with a risk frequency band library. It can not only be associated with the spectrum mask setting, but also dynamically set and adjust various thresholds and durations based on the risk frequency bands stored in the library, combined with system operating conditions, historical data analysis, and expected sensitivity and reliability requirements.

[0113] The method and system proposed in this invention effectively overcome the inherent shortcomings of existing static threshold-based monitoring methods in responding to dynamic changes in the spectrum, such as insensitivity to overall spectrum changes and insufficient early warning sensitivity. By introducing a composite spectrum difference metric based on spectrum masking, an adaptive weighting mechanism, and an early warning strategy that coordinates spectrum mutations and energy, the invention significantly enhances the ability to identify and capture early signs of broadband oscillations (such as slow drift of dominant frequencies, abrupt changes in oscillation modes, and abnormal energy accumulation in specific frequency bands). In particular, the focused monitoring and dynamic tracking of high-risk frequency bands, combined with hierarchical early warning and alarm logic, can significantly improve the timeliness and accuracy of early warnings of broadband oscillations while ensuring a low false alarm rate. This provides more reliable decision support for power system dispatchers and operators, and has significant theoretical value and broad engineering application prospects for preventing oscillation accidents and ensuring the safe and stable operation of the power grid.

[0114] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0115] Computer readable storage media can be tangible storage media which can retain and store instructions for use by an instruction execution device. Computer readable storage media can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer readable storage media include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0116] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0117] Computer readable program instructions for carrying out operations of the present disclosure can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0118] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, but not to limit it. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced, and any modification or replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A method for wideband oscillation identification and early warning based on composite spectral difference metric, characterized in that, The method comprises: obtaining real-time operation data of each node of the power grid to obtain a time series spectrum; during a historical oscillation event of the power grid, when an oscillation alarm is removed and after a set period of steady-state operation, taking a spectrum of a statistical average value of historical operation data in a sliding window in the set period as a reference spectrum; introducing a spectrum mask to represent a frequency band of interest; calculating a distance between the time series spectrum and the reference spectrum in the frequency band of interest; determining a KL divergence between the time series spectrum and the reference spectrum in the frequency band of interest according to an energy of the reference spectrum in the frequency band of interest; and constructing a composite spectrum difference index by using the distance and the KL divergence; taking an energy of the time series spectrum in the frequency band of interest and an energy change amount as an energy accumulation evolution index; based on the composite spectrum difference index and the energy accumulation evolution index, identifying a wide-frequency oscillation to perform early warning or alarm.

2. The wide-frequency oscillation identification and early warning method based on the composite spectrum difference index according to claim 1, wherein a spectrum mask M(f) is set, M(f) = 1 indicating that a frequency is in the frequency band of interest, and M(f) = 0 indicating that a frequency is not in the frequency band of interest; based on the spectrum mask, a distance between the time series spectrum and the reference spectrum in the frequency band of interest in a period t is calculated; based on the spectrum mask, a probability density function of the time series spectrum and the reference spectrum in the frequency band of interest in the period t is determined; according to the energy of the reference spectrum in the frequency band of interest, the KL divergence between the time series spectrum and the reference spectrum in the frequency band of interest in the period t is calculated by using the probability density function of the time series spectrum and the reference spectrum in the frequency band of interest in the period t; the distance and the KL divergence between the time series spectrum and the reference spectrum in the frequency band of interest in the period t are normalized; the composite spectrum difference index in the period t is constructed by using the normalized distance and the KL divergence.

3. The wide-frequency oscillation identification and early warning method based on the composite spectrum difference index according to claim 2, wherein 4. The wide-frequency oscillation identification and early warning method based on the composite spectrum difference index according to claim 3, wherein In the formula, DE M (t) is the distance between the time series spectrum at time t and the reference spectrum in the frequency band of interest, S(t,f) is the time series spectrum in the time period t, S ref (f) is the reference spectrum, and f is the frequency band.

5. The wide-frequency oscillation identification and early warning method based on the composite spectrum difference index according to claim 4, wherein where P'(t,f) and P(t,f) are the probability density function and the probability distribution function of the time series spectrum in the frequency band of interest in the time period t, respectively, and P ref '(t,f) and P ref (t,f) are the probability density function and the probability distribution function of the reference spectrum in the frequency band of interest in the time period t, respectively, ε is a zero probability term for smoothing, and N M is the number of frequency points in the frequency band of interest.

6. The wide-frequency oscillation identification and early warning method based on the composite spectrum difference index according to claim 5, wherein calculating the energy of the reference spectrum in the band of interest If the energy is 0, then the KL divergence of the time series spectrum and the reference spectrum in the band of interest over the period t is set to 0, otherwise it is calculated as follows: where DKL M (t) is the KL divergence of the time series spectrum and the reference spectrum in the frequency band of interest for time period t. wherein CSDI(t) is the composite spectrum difference index in the period t. According to the normalized distance and the KL divergence The variance σ 2 is used to dynamically adjust and the weights of as follows: where w E (t) is the weight of the time period t KL (t) is the weight of the time period t and the weight of the time period t 7. The wide-frequency oscillation identification and early warning method based on the composite spectrum difference index according to claim 2, wherein 8. The wide-frequency oscillation identification and early warning method based on the composite spectrum difference index according to claim 1, wherein In the formula, E M (t) is the energy of the time series spectrum in the frequency band of interest in time period t, S(t,f) is the time series spectrum in time period t, and f is the frequency band. Energy E of the time series spectrum in the time period t in the frequency band of interest M (t), the amount of energy change ΔE M (t), as an energy accumulation evolution indicator. based on the composite spectrum difference index and the energy accumulation evolution index, identifying a wide-frequency oscillation to perform early warning or alarm, comprising: when the following two conditions are met simultaneously, a wide-frequency oscillation early warning mode is activated: statistically determining a distribution characteristic of CSDI in a steady operation; according to a power grid operation state and a control requirement, a wide-frequency difference allowable value of the time series spectrum and the reference spectrum in the frequency band of interest is determined by using a simulation method; and according to the distribution characteristic of CSDI in the steady operation and the wide-frequency difference allowable value, a spectrum difference threshold is determined. 1) Condition 1, when ΔCSDI(t) > θ CSDI,alert ΔCSDI(t) is the mutation of the composite spectrum difference index in the period t, θ CSDI,alert is the set spectrum difference threshold; 2), condition 2, when E M (t) > θ E,alert or ΔE M (t) > θ ΔE,alert , θ E,alert , θ ΔE,alert are respectively set energy anomaly threshold, energy change anomaly threshold; ​ The distribution characteristics of the ECEI in the steady operation are counted; the upper limit of the energy spectrum and the allowable value of the energy fluctuation in the concerned frequency band are determined by simulation according to the grid operation state and the control requirements; and the energy abnormal threshold and the energy change abnormal threshold are determined according to the distribution characteristics of the ECEI in the steady operation and the upper limit of the energy spectrum and the allowable value of the energy fluctuation.

9. The method of claim 8, wherein the warning duration is determined according to the fault recording duration configured by the wideband measurement device. If the wideband oscillation early warning mode is activated, when A M (t)>θ A,warning And continue ΔT warning When or when CSDI(t) > θ CSDI,warning And continue ΔT warning When this happens, a wideband oscillation warning signal will be issued; where A M (t) is the amplitude of the time series spectrum in the frequency band of interest at time t, θ A,warning is the set amplitude warning threshold, θ CSDI,warning is the set wideband oscillation warning threshold, ΔT warning is the set warning duration; Amplitude warning threshold θ A,warning 1.1 times the average value of the amplitude of the reference frequency in the frequency band of interest, wide frequency oscillation warning threshold θ CSDI,warning 1 + θ CSDI,alert ; 10. The method of claim 8, wherein the method comprises: a spectrum acquisition module configured to acquire real-time operation data of each node of the power grid to obtain a time series spectrum; Whether the wideband oscillation early warning mode is activated or not, when A M (t) > θ A,alarm and lasts for ΔT alarm or when CSDI(t) > θ CSDI,alarm and lasts for ΔT alarm , a wideband oscillation warning signal is sent out; wherein θ A,alarm is a set amplitude warning threshold, θ CSDI,alarm is a set wideband oscillation warning threshold, and ΔT alarm is a set warning duration. Amplitude alarm threshold θ A,alarm 1.2 times the average value of the amplitude of the reference frequency in the frequency band of interest, the wide frequency oscillation alarm threshold θ CSDI,alarm 1.1 times the wide frequency oscillation warning threshold θ CSDI,warning ; the alarm duration is determined according to the fault recording time length configured by the wide frequency measurement device.

11. A composite frequency spectrum difference measure based wideband oscillation identification and warning system for implementing the composite frequency spectrum difference measure based wideband oscillation identification and warning method according to any one of claims 1 to 10, characterized in that, during a historical oscillation event of the power grid, a frequency spectrum of a statistical average of historical operation data in a sliding window in a set period after the oscillation alarm is removed and steady operation for a set period is performed, is taken as a reference frequency spectrum; an index establishment module configured to introduce a frequency spectrum mask to represent a concerned frequency band; to calculate a distance between the time series spectrum and the reference frequency spectrum in the concerned frequency band; to determine a KL divergence between the time series spectrum and the reference frequency spectrum in the concerned frequency band according to the energy of the reference frequency spectrum in the concerned frequency band; and to construct a composite frequency spectrum difference index by using the distance and the KL divergence; and to take the energy of the time series spectrum in the concerned frequency band and an energy change amount as an energy accumulation evolution index; an oscillation warning module configured to identify a wideband oscillation and perform a warning or an alarm based on the composite frequency spectrum difference index and the energy accumulation evolution index.

12. A terminal comprising a processor and a storage medium, wherein: the storage medium is configured to store instructions; the processor is configured to operate according to the instructions to perform the steps of the method of any one of claims 1-10. The program is executed by the processor to implement the steps of the method of any one of claims 1-10. ​ 13. A computer readable storage medium having stored thereon a computer program, characterized in that ​

Images