A robust state estimation method for power grid voltage sag

Abstract

The embodiment of the present invention discloses a robust state estimation method for grid voltage sag, which relates to the technical field of power quality monitoring of power systems and can improve the robustness and accuracy of grid voltage sag state estimation. The present invention includes: after establishing a voltage sag disturbance signal model for state estimation, collecting voltage sag monitoring waveform data and inputting it into an improved nonlinear filtering system; using the improved nonlinear filtering system, performing dynamic phasor measurement of the voltage sag parameters of each monitoring point; establishing a structural risk-minimized robust state estimation model for voltage sag, using the instantaneous voltage phasor of each monitoring point as the basic quantity measurement for state estimation, and simultaneously using the instantaneous voltage phasor of non-monitoring points as the state quantity; using the structural risk-minimized robust state estimation model for voltage sag, adopting a quasi-real-time incremental support vector machine algorithm driven by a hybrid model and data, to obtain a robust state estimation result for distribution network voltage sag.

Description

Technical Field

[0001] The present invention relates to the technical field of power quality monitoring of power systems, and in particular to a robust state estimation method for grid voltage sag. Background Art

[0002] Voltage sags are a key indicator of power quality in distribution networks, causing a range of serious problems for numerous grid devices and power users, including tripping of computers and electronic equipment, and abnormal operation of motors, wind turbines, speed control systems, and various industrial equipment. Power quality issues cost China over 100 billion yuan annually, with voltage sags accounting for 70%-90% of all power quality issues. Furthermore, the increasing proportion of renewable energy connected to the grid in recent years has significantly impacted power quality in distribution networks, including voltage sags.

[0003] To address the voltage sag problem in new distribution networks, it is necessary to monitor and analyze the voltage sag, conduct state assessment, locate disturbance sources, and manage it throughout the entire distribution network to ensure safe and economical operation of the new distribution network. However, due to technical and financial limitations, it is quite difficult and uneconomical to establish a complete power quality monitoring network similar to WAMS in the short term. In this context, some researchers have proposed applying state estimation to voltage sag analysis and assessment. However, due to the uncertainty of actual distribution models, measurement errors, and model linearization errors, the distribution network voltage sag state estimation is required to have stronger robustness. That is, the voltage sag observation and estimation system must have the characteristics of maintaining certain estimation performance (such as estimation accuracy, robustness, observability / local observability, convergence, etc.) under certain network parameter disturbances. However, existing power quality state estimation, especially voltage sag state estimation, still lacks an effective strategy.

[0004] Therefore, due to factors such as measurement redundancy, parameter uncertainty, time scale, and reliability, distribution network voltage sag state estimation is still far from being truly applicable. Addressing these issues will be a key task facing the power quality field for a considerable period of time. In particular, the high penetration of distributed power sources has made new distribution networks more characterized by multiple sources and highly uncertain time-varying power flows. This has brought new challenges to distribution network voltage sag state estimation, and technological innovation is urgently needed to address these challenges. Summary of the Invention

[0005] An embodiment of the present invention provides a robust state estimation method for power grid voltage sag, which can improve the accuracy of robust state analysis of power grid voltage sag.

[0006] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0007] In an improved nonlinear filtering system, after a voltage sag disturbance signal model for state estimation is established, voltage sag monitoring waveform data is collected and input into the improved nonlinear filtering system; dynamic phasor measurement of the voltage sag parameters of each monitoring point is performed through the improved nonlinear filtering system; a robust state estimation model for voltage sag with structural risk minimization is established, wherein the instantaneous voltage phasor of each monitoring point is used as the basic quantity measurement for state estimation, and the instantaneous voltage phasor of non-monitoring points is used as the state quantity; the robust state estimation model for voltage sag with structural risk minimization is used, and a quasi-real-time incremental support vector machine algorithm driven by a hybrid model and data is adopted to obtain analysis results of the robust state of the distribution network voltage sag.

[0008] The robust state estimation method for grid voltage sag provided by the embodiment of the present invention uses an improved nonlinear filtering method to obtain dynamic phasor measurements of transient response capability, establishes a time-varying instantaneous phasor model, and achieves good dynamic phasor tracking under transient small disturbance conditions, providing good data support for subsequent voltage sag state estimation. A model and data hybrid-driven robust state estimation model for distribution network voltage sag is established, effectively considering the distribution network structural parameters, power supply uncertainty, and system time-varying characteristics during disturbances. A large amount of distribution network simulation data and measurement data are integrated into the voltage sag state estimation model, and a robust state estimation theory for voltage sag based on structural risk minimization with a relatively rigorous theoretical basis is established. By adopting the Log-coth error evaluation function and the structural risk minimization state estimation model, the estimation model still has high accuracy and robustness under the strong noise, parameter time-varying, uncertain load and power supply interference conditions of the high-penetration new distribution network. In addition, SCADA system measurements are added to the estimation model, which greatly improves data redundancy, avoids information islands, and further improves estimation accuracy. The present invention plays an important role in the management and improvement of power quality of modern new distribution networks and in ensuring the safe and economical operation of distribution networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0010] Figure 1 A schematic diagram of a method flow chart of a specific example provided in an embodiment of the present invention;

[0011] Figure 2 A schematic diagram of a method flow chart provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention will be described in detail below, with examples of the embodiments illustrated in the accompanying drawings. Throughout, identical or similar reference numerals represent identical or similar elements or elements having identical or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and intended only to explain the present invention and are not to be construed as limiting the present invention. Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "the" used herein may also include the plural forms. It should be further understood that the term "comprising" as used in the description of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when an element is referred to as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or intervening elements may be present. Furthermore, "connected" or "coupled" as used herein may include wireless connections or couplings. The term "and / or" as used herein includes any and all combinations of one or more associated listed items. It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art, and will not be interpreted in an idealized or overly formal sense unless defined as such herein.

[0013] At present, in order to solve the problem of voltage sag in distribution networks, it is necessary to monitor and analyze the voltage sag, conduct state assessment, locate the disturbance source, and manage the voltage sag of the entire distribution network to ensure the safe and economical operation of the new distribution network. However, due to technical and financial limitations, it is quite difficult and uneconomical to establish a complete power quality monitoring network similar to WAMS in the short term. In this context, some researchers have proposed applying state estimation to voltage sag analysis and assessment. For example: in some existing schemes, the least squares method is used to search for the path where the fault point is located, and a voltage sag state estimation scheme for radial distribution networks is proposed, thereby estimating the voltage sag information of non-monitoring nodes; for another example: the voltage sag state estimation equation is solved by using an electromagnetic simulation algorithm, and the computational efficiency is greatly improved through intelligent optimization, but there are problems such as parameter setting relying on subjective experience, unclear mechanism, and convergence. Other researchers have proposed a statistical analysis method based on field measurements to estimate the frequency of voltage sags at the common connection point of the low-voltage power grid based on the voltage sag characteristics of medium and high-voltage substations. Similarly, some researchers have used transient state estimation methods to analyze and optimize the number and location of voltage sag monitoring points in the transmission network, and calculated the voltage sag index based on the configured monitoring point information. There are also studies on voltage sag state estimation based on generalized inverse algorithms and singular value decomposition algorithms. However, due to the uncertainty of actual distribution models, measurement errors, and model linearization errors, the distribution network voltage sag state estimation is required to have stronger robustness, that is, the voltage sag observation and estimation system must have certain estimation performance characteristics (such as estimation accuracy, robustness, observability / local observability, convergence, etc.) under certain network parameter disturbances. However, existing power quality state estimation, especially voltage sag state estimation, still lacks effective strategies.

[0014] In summary, while distribution network voltage sag state estimation has made some progress in recent years, it remains far from practical application due to factors such as measurement redundancy, parameter uncertainty, time scale, and reliability. Addressing these issues will be a key task facing the power quality field for a considerable period of time. In particular, the high penetration of distributed power sources has led to new distribution networks with multiple sources and highly uncertain time-varying power flows. This has brought new challenges to distribution network voltage sag state estimation, and technological innovation is urgently needed to address these challenges.

[0015] The design purpose of this embodiment is to solve or improve practical problems in the current state of the art. The specific design concept is to adopt a robust state estimation method for voltage sags in complex distribution networks, use an improved nonlinear filtering method to obtain dynamic phasor measurements of transient response capability, establish a time-varying instantaneous phasor model, achieve good dynamic phasor tracking under small transient disturbance conditions of the system, and provide good data support for subsequent voltage sag state estimation. A model and data hybrid-driven robust state estimation model for voltage sags in distribution networks is established, effectively considering distribution network topology and parameter, and power supply uncertainty. A large amount of distribution network simulation data and measurement data are integrated into the voltage sag state estimation model, and a robust state estimation theory for voltage sags based on structural risk minimization with a relatively rigorous theoretical basis is established. By adopting the Log-coth error evaluation function and the structural risk minimization state estimation model, the estimation model maintains high accuracy and robustness under the strong noise, time-varying parameter, uncertain load and power supply interference conditions of high-penetration new distribution networks. In addition, the addition of SCADA system measurements to the estimation model greatly improves data redundancy, avoids information islands, and further improves estimation accuracy. The present invention plays an important role in the management and improvement of power quality of modern new distribution networks and in ensuring the safe and economical operation of distribution networks.

[0016] The embodiment of the present invention provides a robust state estimation method for power grid voltage sag, such as Figure 2 Shown, including:

[0017] S1. After establishing a voltage sag disturbance signal model for state estimation in an improved nonlinear filtering system, voltage sag monitoring waveform data is collected and input into the improved nonlinear filtering system.

[0018] In this embodiment, a voltage sag disturbance signal model for state estimation is established for the non-stationary time-varying grid voltage, current amplitude and phase with frequency disturbance, non-periodic transient and noise interference caused by highly renewable distributed power sources injected into the distribution network.

[0019] S2. Through the improved nonlinear filtering system, dynamic phasor measurement of the voltage sag parameters of each monitoring point is performed.

[0020] In this embodiment, it is necessary to implement detection and processing of voltage sag characteristic parameters of the new distribution network oriented to state estimation for monitoring points in the new distribution network oriented to state estimation.

[0021] S3. Establish a robust state estimation model for voltage sag with minimal structural risk.

[0022] The instantaneous voltage phasor at each monitoring point is used as the basic quantity for state estimation, while the instantaneous voltage phasor at non-monitoring points is used as the state quantity. Specifically, during the pre-preparation process, the distribution network topology and network parameters can be obtained through SCADA (Supervisory Control And Data Acquisition).

[0023] S4. Utilizing the voltage sag robust state estimation model with minimized structural risk, and adopting a model-and-data hybrid-driven incremental support vector machine quasi-real-time algorithm, obtain analysis results of the robust state of the distribution network voltage sag.

[0024] The analysis result at least includes information on voltage sag of the entire distribution network.

[0025] This embodiment can reliably track the time-varying dynamic phasors of voltage and current during voltage sag transients under system frequency disturbances. It features low computational complexity, simple implementation, and strong anti-interference capabilities, effectively meeting the requirements for preprocessing measurement information prior to voltage sag state estimation. It can also detect characteristic parameters in three dimensions: amplitude, duration, and phase jumps of voltage sag disturbances at monitoring points. Considering that during the operation of new distribution networks, voltage and current signals are often composed of sinusoidal components with time-varying amplitude and phase, non-periodic transient components, and various noises, and that the system frequency also fluctuates to a certain extent, it is necessary to establish a reasonable descriptive model for the signal's transient nature, time-varying properties, frequency disturbances, and noise, and to perform effective detection signal processing. This provides a good data source for subsequent transient state estimation, reduces the measurement error and algorithm complexity of the state estimation model, and improves the estimation accuracy and real-time performance of transient state estimation.

[0026] In this embodiment, the establishment of a voltage sag disturbance signal model for state estimation includes:

[0027] Since the power injected into the distribution network by renewable distributed generation is mostly volatile, intermittent, and uncertain, and the power supply is frequently switched on and off, the grid voltage and current are actually non-stationary signals with time-varying amplitude and phase, and may also be subject to frequency disturbances, non-periodic transients, and noise interference. A voltage sag disturbance signal model for state estimation is established based on the transient, time-varying, frequency disturbance, and noise characteristics of the signal:

[0028]

[0029] Among them, t* is the time when the voltage sag occurs, T PQ is the duration of the voltage sag event (for voltage sag, T PQIt may be tens of milliseconds to several seconds. It is the relative change of voltage amplitude after the system reaches a new steady state. g(t) is the harmonic and noise component. D PQ (tt*) is the sag depth, U is the voltage rating, t is the time, f0 is the system frequency, is the fundamental phase before disturbance, is the fundamental phase during the disturbance, is the fundamental phase after the disturbance ends, g(t) is the non-periodic transient and noise, u i (t) is the input voltage. T PQ 、D PQ (tt*), The core feature of voltage sag disturbance is that voltage sag state estimation is to effectively estimate the T of non-monitoring points using the information of monitoring points. PQ 、D PQ (tt*), In addition, due to the relative voltage sag disturbance, the voltage harmonic signal of the actual power grid as a whole is relatively small, and the voltage harmonics are not considered in the present invention.

[0030] Specifically, the improved nonlinear filtering system can be used to implement a new distribution network voltage sag characteristic parameter processing for state estimation. Specifically, in order to effectively obtain the monitoring point voltage sag parameter T for voltage sag state estimation, PQ 、D PQ (tt*), The present invention considers using a nonlinear adaptive filtering system to achieve voltage sag parameter detection in the presence of harmonics, system frequency fluctuations, and measurement noise. The improved nonlinear filtering system performs dynamic phasor measurement of the voltage sag parameters at each monitoring point, including:

[0031] The improved nonlinear filtering system established:

[0032]

[0033] in, is the instantaneous phase of voltage, U(t) is the instantaneous amplitude of voltage, t is time, k1 and k2 are two constants of the filtering system, u i (t) is the input voltage, ω0 is the initial angular frequency constant of the filter system, and i represents the sign of the input. By setting the system parameters k1 to 60 and k2 to 20, dynamic detection of the instantaneous amplitude and phase of the input voltage signal can be achieved with good convergence speed and accuracy. Based on the detected instantaneous amplitude and phase, the instantaneous amplitude, frequency, and phase of the dynamic phasor at each monitoring point are calculated.

[0034] Among them, the characteristic root The system's approximation speed is determined, and the system stably converges to a uniquely determined periodic orbit near the neighborhood of A0 sin(ω0t+δ0). The size of the neighborhood is determined by k1, k2, and g(t). By selecting and setting the system parameters k1 and k2 to optimize the convergence speed and accuracy of the instantaneous amplitude and instantaneous phase, the system can better complete the instantaneous fundamental amplitude and instantaneous phase tracking of the external input voltage and current of the measuring equipment under conditions such as small system frequency disturbances, transient oscillation interference, and white noise, and realize the measurement of voltage, current dynamic phasor and frequency with transient response capability. Then, through simple algorithm processing, the voltage sag parameters can be obtained: sag duration T PQ , sag depth D PQ (tt*), temporary dip phase jump And obtain the instantaneous amplitude, frequency and phase of the dynamic phasor.

[0035] In this embodiment, in the process of establishing a voltage sag robust state estimation model with minimized structural risk, obtaining the voltage vector of the monitoring node includes:

[0036] The instantaneous phasor measurements of voltage and current at the PQMs monitoring point are obtained. and T PQ 、D PQ (tt*), The voltage sag parameters are measured with the instantaneous voltage phasor of each monitoring point as the basic quantity for state estimation, and the instantaneous voltage phasor of non-monitoring points as the state quantity. A robust voltage sag state estimation model with minimized structural risk is established.

[0037] The complex grid node admittance matrix is divided according to the monitoring points and non-monitoring points of PQMs, and the voltage equation of n nodes is obtained: Among them, Um and Im are the voltage vector and injection current vector of the monitoring node, Uu and Iu are the voltage vector and injection current vector of the non-monitoring point, Ymm and Yuu are the self-admittance of the busbar of the monitoring node and the non-monitoring node, and Ymu and Yum are the mutual admittance between the monitoring node and the non-monitoring node.

[0038] Estimate the apparent power of the equivalent load and source at non-monitoring points, where: Yu is the load equivalent admittance, Iu is the current phasor, Uu is the voltage phasor, is the apparent power, and Uu(i) is the node voltage amplitude. The linear measurement equation for instantaneous voltage phasor state estimation is established: Um=H·Uu+ε, where H is the measurement matrix and ε is the measurement error. The voltage vector of the monitoring node is obtained Among them, H is the state estimation measurement matrix, H=[Yum -1 (diag[Yu]-Yuu)], ε is the measurement error.

[0039] Specifically, in the process of establishing a robust state estimation model for voltage sag with minimal structural risk, the following steps are involved: using the Log-coth error evaluation function to preliminarily establish a state estimation model for voltage sag; According to the relationship between state variables and measurement vectors, the optimal state variable solution is obtained, and the basic model for voltage sag state estimation is established:

[0040] Where m is the number of monitoring nodes, i.e., the number of measurements, and coth() is the hyperbolic cotangent function.

[0041] Aiming at the structural error caused by the matrix H itself, a robust state estimation model for voltage sag is established with minimal structural risk: Among them, C emp is the empirical risk penalty coefficient, m is the number of monitoring nodes, i.e., the number of measurements, For structural risk.

[0042] In the process of obtaining the analysis results of the robust state of the distribution network voltage sag, it is necessary to collect measurement data to participate in the grid state estimation, including: using the SCADA (Supervisory Control And Data Acquisition) section scanning time as the time reference, obtaining all PQM (Power Quality Monitor) data sections within a certain time range. Calculating the matching degree between the section of the SCADA measurement data and the section of each PQM measurement data, and selecting the section of the PQM measurement data with the greatest matching degree. Specifically, to improve the synchronization of SCADA data and PQMs data, the present invention uses the SCADA section scanning time as the time reference, obtains all PQM measurement data within a certain time range, calculates the matching degree between the section of the SCADA measurement data and the section of each PQM measurement, selects the PQM measurement with the greatest matching degree as the section that matches the SCADA data, and then uses the PQM measurement with the greatest matching degree and the measurement data in the SCADA measurement data section to jointly participate in the grid state estimation calculation.

[0043] In this embodiment, the state estimation algorithm model can be combined with the existing weighted least squares (WLS) estimation algorithm and the least absolute value estimation (LAV) and weighted least absolute value estimation (WLAV) robust estimation methods. The present invention improves the error evaluation function of the estimation algorithm from the perspective of the loss function and uses the Log-coth error evaluation function to establish a preliminary voltage sag state estimation model, which is in the form of: Compared with the minimum mean square error and weighted mean square error (WLS), the Log-coth error evaluation function is more robust to poor measurement system data. Furthermore, compared with the minimum absolute error estimate (WLAV), its gradient decreases as the error decreases, making the estimation result more accurate. It combines the advantages of minimum mean square error and minimum absolute error while maintaining excellent robustness and estimation accuracy. Furthermore, compared with the Huber error evaluation function, the Log-coth evaluation function is second-order differentiable, providing a better foundation for solving subsequent state estimation models. It uses the simpler Newton method to find the optimal point for subsequent state estimation, effectively improving the algorithm's accuracy and efficiency. Furthermore, it eliminates the need to consider the hyperparameter δ setting and optimization issues, making it more convenient to use.

[0044] After establishing the error evaluation function, the state estimation of voltage sag is carried out. That is, the optimal state variable solution is obtained according to the relationship between the state variable and the measurement vector. The following basic model for voltage sag state estimation is established: Where m is the number of monitoring nodes, i.e., the number of measurements.

[0045] The measurement matrix H is obtained by calculating the power grid SCADA section measurement data with a period of 1 / 30 minute. At the same time, in order to improve the synchronization of SCADA data and PQMs data, the present invention uses the SCADA section scanning moment as the time reference, obtains all PQM data sections within a certain time range (±30 seconds), calculates the section matching degree between the SCADA measurement data section and each PQM measurement, selects the PQM measurement with the greatest matching degree as the section that matches the SCADA data, and then uses the PQM measurement with the greatest matching degree and the measurement data in the SCADA measurement data section together to participate in the power grid state estimation calculation. Although the use of the above-mentioned Log-coth error evaluation function can improve the estimation error caused by poor data in the measurement system and achieve stronger estimation robustness, this error is the error between the value and the measurement estimate value, which is actually still an empirical error, and the above-mentioned process does not effectively deal with the structural error of the measurement matrix H itself. To improve the estimation robustness, based on the previous preliminary estimation model to improve the estimation error caused by bad data of the measurement system, the structural error caused by the measurement matrix H itself is further measured. Drawing on statistical learning theory, this paper proposes a state estimation model based on structural risk minimization to establish a state estimation model for voltage sag:

[0046] in, is the structural risk, is the second-order norm of the estimated model structural parameters, m is the number of monitoring nodes, i.e., the number of measurements, C empThe empirical risk penalty coefficient balances the stability and accuracy of the estimation algorithm, ensuring that the estimation model maintains high accuracy and robustness under the conditions of strong noise, time-varying parameters, and uncertain load and power supply interference in high-penetration new distribution networks. Furthermore, the support vector machine has strong robustness and nonlinear interpolation and extrapolation capabilities, which can effectively describe the nonlinear characteristics of the estimation model under transient voltage disturbances.

[0047] The measurement matrix H is calculated using grid SCADA section measurement data with a 1 / 30 minute period. To improve the synchronization between SCADA data and PQMs data, the present invention uses the SCADA section scan time as the time reference, acquires all PQM data sections within a certain time range (±30 seconds), calculates the degree of matching between the SCADA measurement data section and each PQM measurement section, selects the PQM measurement with the highest matching degree as the section that matches the SCADA data, and then uses the PQM measurement with the highest matching degree and the measurement data in the SCADA measurement data section together in the grid state estimation calculation.

[0048] In this embodiment, a model- and data-driven, hybrid-driven, incremental support vector machine algorithm for quasi-real-time distribution network voltage sag state estimation is proposed for the established robust state estimation model for voltage sags with minimized structural risk. The specific process can be designed as follows: First, using a professional simulation platform for distribution systems, taking into account the uncertainty of distribution network structure and component parameters, the randomness and volatility of distributed power sources and loads, a Monte Carlo method is used to generate massive simulated distribution network sag patterns within a neighborhood based on the voltage sag analysis model, establish a platform for monitoring and analyzing various voltage sag disturbance events and synchronous state estimation data acquisition for the entire network, and construct a massive label database (i.e., a data sample set containing both the quantity measurement and its corresponding state quantity label information) for support vector machine offline supervised learning of voltage sag state estimation measurements, state quantities, and state estimation model parameters. Then, a labeled database for support vector machine learning of voltage sag state estimation is constructed using data from actual monitoring systems such as voltage sag monitoring systems (or PQMs) and SCADA. Afterwards, a voltage sag robust state estimation model is pre-trained using a simulated label dataset of voltage sag state estimation corresponding to SCADA parameters, and an initial model and initial parameters of a support vector machine for voltage sag state estimation in the distribution network are constructed. Then, combining the cooperative training principle in semi-supervised learning, regression learning is performed from the estimated quantity measurement to the estimated state quantity. The existing labeled state quantity values of the SVM are used to obtain the corresponding state quantity values and confidence levels for the measurement data. In order to select appropriate labeled measurement samples to label the state quantity, the present invention adopts the nearest neighbor concept to select learning samples. Furthermore, considering the estimation speed and the dynamic nature of the sag monitoring system's measurements, the present invention also introduces an incremental algorithm to shorten training time and state estimation speed. Specifically, after each machine learning iteration obtains an estimation model, a small set of support vectors and corresponding support network coefficients obtained through training are dynamically saved. After this iterative training, the next round of instantaneous voltage estimation begins, retaining the support vector set retained from the previous round of sag monitoring system measurements and estimation learning. Using a combination of improved KKT conditions and an error-driven strategy, a certain number of samples are selected from the sag monitoring data set to form the current training sample set. This process continues until the incremental support vector machine state estimation model is solved, resulting in the instantaneous voltage phasors for each analysis period during the entire voltage sag disturbance process. This significantly reduces the number of samples involved in training, significantly reducing the training time for the voltage sag state estimation method driven by a hybrid model and data, thereby improving the speed and real-time performance of voltage sag state estimation.

[0049] The embodiment of the present invention can adjust the application mode according to the specific scenario. For example, Figure 1 As shown, the robust state estimation method for voltage sag in a complex distribution network of this embodiment can be implemented as the following process:

[0050] Step 1: Construct a new distribution network voltage sag description model for state estimation. Specifically, since the power injected into the distribution network by renewable distributed power sources is mostly volatile, intermittent, and uncertain, and the power supply is frequently switched on and off, the grid voltage and current are actually non-stationary signals with time-varying amplitude and phase, and may also be subject to frequency disturbances, non-periodic transients, and noise interference. A voltage sag disturbance signal model for state estimation is established based on the transient, time-varying, frequency disturbance, and noise characteristics of the signal:

[0051]

[0052] Where t* is the time when the voltage sag occurs, T PQ is the duration of the voltage sag event, δU is the relative change in voltage amplitude after the system reaches a new steady state, and g(t) is the harmonic and noise component. For voltage sag, T PQ It may be tens of milliseconds to several seconds. PQ (tt*) is the sag depth. Therefore, T PQ 、D PQ (tt*), The core feature of voltage sag disturbance is that voltage sag state estimation is to effectively estimate the T of non-monitoring points using the information of monitoring points. PQ 、D PQ (tt*), In addition, due to the relative voltage sag disturbance, the voltage harmonic signal of the actual power grid as a whole is relatively small, and the voltage harmonics are not considered in the present invention.

[0053] Step 2: Use the improved nonlinear filtering system to implement the new distribution network voltage sag characteristic parameter processing for state estimation. Specifically, in order to effectively obtain the monitoring point voltage sag parameter T for voltage sag state estimation PQ 、D PQ (tt*), The present invention considers using a nonlinear adaptive filtering system to achieve voltage sag parameter detection in the presence of harmonics, system frequency fluctuations, and measurement noise. The improved nonlinear filtering system can be expressed by the following differential equation:

[0054]

[0055] Its characteristic root The system's approximation speed is determined, and it can be seen that the system is stable and converges to a unique periodic orbit near the neighborhood of A0 sin(ω0t+δ0). The size of the neighborhood is determined by k1, k2, and g(t). By selecting and setting the system parameters k1 and k2, the convergence speed and accuracy of the instantaneous amplitude and instantaneous phase can be optimized. The system can better complete the instantaneous fundamental amplitude and instantaneous phase tracking of the external input voltage and current of the measuring equipment under conditions such as small system frequency disturbances, transient oscillation interference, and white noise, and realize the measurement of voltage, current dynamic phasor and frequency with transient response capability. Then, the voltage sag parameter T of the PQ monitoring point can be obtained through simple algorithm processing. PQ 、D PQ (tt*), The instantaneous amplitude, frequency, and phase of the dynamic phasor are obtained. This algorithm can reliably track the time-varying dynamic phasors of voltage and current during voltage sag transients under system frequency disturbances. It has low computational complexity, simple implementation, and strong anti-interference capabilities. It can effectively meet the requirements for preprocessing measurement information before estimating voltage sag states. It can also detect characteristic parameters in three dimensions: voltage sag disturbance amplitude, duration, and phase jump at the monitoring point.

[0056] Considering that during the operation of the new distribution network, the voltage and current signals are actually often composed of sinusoidal components with time-varying amplitude and phase, non-periodic transient components and various noises, and the system frequency also has certain fluctuations, we need to establish a reasonable description model for the transient nature, time-varying nature, frequency disturbance and noise of the signal and perform effective detection signal processing to provide a good data source for subsequent transient state estimation, thereby reducing the measurement error and algorithm complexity of the state estimation model and improving the estimation accuracy and real-time performance of the transient state estimation.

[0057] Step 3: measure the instantaneous phasor of voltage and current at the PQMs monitoring point obtained in step 2. and T PQ 、D PQ (tt*), The voltage sag parameters are measured using the instantaneous voltage phasor of each monitoring point as the basic quantity for state estimation. The instantaneous voltage phasor of the non-monitoring point is used as the state quantity. A robust state estimation model for voltage sag is established to minimize structural risk. The specific process is as follows:

[0058] First, the complex grid node admittance matrix is divided according to the monitoring points and non-monitoring points of PQMs to obtain the n-node voltage equation:

[0059]

[0060] Where Um and Im are the voltage vector and injection current vector of the monitoring node, Uu and Iu are the voltage vector and injection current vector of the non-monitoring node, Ymm and Yuu are the self-admittance of the busbar of the monitoring node and the non-monitoring node respectively, and Ymu and Yum are the mutual admittance between the monitoring node and the non-monitoring node. The equivalent load and power supply of the non-monitoring node are estimated by their apparent power:

[0061]

[0062] Taking the measurement noise into account, the linear measurement equation for instantaneous voltage phasor state estimation is established:

[0063] Um=H·Uu+ε (5)

[0064] Where H is the state estimation measurement matrix, H=[Yum -1 (diag[Yu]-Yuu)], ε is the measurement error.

[0065] In terms of state estimation algorithm models, the existing technologies mainly include weighted least squares (WLS) estimation algorithm, least absolute value estimation (LAV), and weighted least absolute value estimation (WLAV) robust estimation method. The present invention improves the error evaluation function of the estimation algorithm from the perspective of loss function, and adopts the Log-coth error evaluation function to solve the state estimation problem of voltage sag, which is in the following form:

[0066]

[0067] Compared with the minimum mean square error (MSE) and weighted mean square error (WLS), the Log-coth error evaluation function is more robust to poor measurement system data. Furthermore, compared with the minimum absolute error estimates (LAV and WLAV), its gradient decreases as the error decreases, making the estimation more accurate. It combines the advantages of both the MSE and the absolute error while maintaining excellent robustness and estimation accuracy. Furthermore, compared with the Huber error evaluation function, the Log-coth error evaluation function is second-order and everywhere differentiable, providing a better foundation for solving subsequent state estimation models. It uses the simpler Newton method to find the optimal point for subsequent state estimation, effectively improving the algorithm's accuracy and efficiency. Furthermore, it eliminates the need to consider the hyperparameter δ setting and optimization issues, making it more convenient to use.

[0068] After establishing the above error evaluation function, the state estimation of voltage sag, that is, obtaining the optimal state variable solution based on the relationship between the state variable and the measurement vector, is essentially solving the following mathematical problem:

[0069]

[0070] Where m is the number of monitoring nodes, i.e., the number of measurements.

[0071] The measurement matrix H is calculated using grid SCADA section measurement data with a 1 / 30 minute period. To improve the synchronization between SCADA data and PQMs data, the present invention uses the SCADA section scan time as the time reference, acquires all PQM data sections within a certain time range (±30 seconds), calculates the degree of matching between the SCADA measurement data section and each PQM measurement section, selects the PQM measurement with the highest matching degree as the section that matches the SCADA data, and then uses the PQM measurement with the highest matching degree and the measurement data in the SCADA measurement data section together in the grid state estimation calculation.

[0072] Although the Log-Coth error evaluation function can improve the estimation error caused by poor data in the measurement system and achieve strong estimation robustness, this error is the error between the value and the measurement estimate, which is actually still an empirical error. The above process does not effectively deal with the structural error of the measurement matrix H itself. Drawing on statistical learning theory, this paper proposes a state estimation model based on structural risk minimization to further optimize the above model. The state estimation model for structural risk minimization is as follows:

[0073]

[0074] Among them C emp is the empirical risk penalty coefficient, is the structural risk, and is the second-order norm of the estimated model structural parameters. C emp This approach balances the stability and accuracy of the estimation algorithm, ensuring that the estimation model maintains high accuracy and robustness under the conditions of strong noise, time-varying parameters, and uncertain load and power supply interference found in highly penetrated new distribution networks. Furthermore, research has demonstrated that the statistical learning tool support vector machine (SVM) exhibits strong robustness and nonlinear interpolation and extrapolation capabilities, effectively capturing the nonlinear characteristics of the estimation model under transient voltage disturbances.

[0075] Step 4: For the voltage sag robust state estimation model with structural risk minimization established in step 3, a model- and data-driven incremental support vector machine quasi-real-time distribution network voltage sag state estimation solution algorithm is proposed. The specific process is as follows:

[0076] First, utilizing a specialized distribution system simulation platform, taking into account the uncertainties of distribution network structure and component parameters, as well as the randomness and volatility of distributed power sources and loads, a Monte Carlo method was used to generate a large number of simulated distribution network sag patterns within a neighborhood based on a voltage sag analysis model. This approach then established a platform for simultaneous data acquisition for monitoring, analysis, and state estimation of various voltage sag disturbance events across the entire network. This approach then constructed a massive labeled database (i.e., a data sample set containing both the measured quantities and their corresponding state quantity labels) for offline supervised learning of support vector machines (SVMs) for voltage sag state estimation. Furthermore, a labeled database for SVM learning of measured quantities and state estimation model parameters for voltage sag state estimation was constructed using data from actual monitoring systems, such as voltage sag monitoring systems (PQMs) and SCADA systems. Subsequently, a robust voltage sag state estimation model was pre-trained using the simulated labeled dataset of SCADA parameters for state estimation to construct an initial SVM model and parameters for voltage sag state estimation in the distribution network. Then, combining the collaborative training principle in semi-supervised learning, regression learning is performed from the estimated quantity measurement to the estimated state quantity. The existing state quantity values of the SVM are used to obtain corresponding labels and confidence levels for the measurement data. To select appropriate labeled measurement samples to label the state quantity, the present invention adopts the nearest neighbor principle to select learning samples. In addition, considering the estimation speed and the dynamic nature of the sag monitoring system measurement, the present invention also introduces an incremental algorithm to shorten the sample training time and state estimation speed. That is, after each machine learning to obtain the estimation model, a small number of support vector sets and corresponding support network coefficients corresponding to the training results are dynamically saved. After this iterative training, the next round of instantaneous voltage estimation is entered. The support vector set retained from the previous round of sag monitoring system measurement and estimation learning is retained. A certain number of samples are selected from the labeled sag monitoring data set using the improved KKT condition and the error-driven strategy to form the current training sample set. This is done until the incremental support vector machine state estimation model is solved, and the instantaneous voltage phasors for each analysis period during the entire voltage sag disturbance process are obtained. In this way, by reducing the number of samples involved in learning and training, the training time of voltage sag state estimation driven by model and data hybrid can be greatly reduced, thereby improving the speed and real-time performance of voltage sag state estimation.

[0077] The existing voltage sag state estimation methods do not take into account the uncertain parameters of the distribution network and have low estimation robustness. With the access of high-penetration random renewable distributed power sources, the distribution network operates in a highly variable environment. Due to the changes in active power, voltage, current, load, power supply, and grid component parameters as well as the existence of interference, the mathematical model of the distribution system during the voltage sag disturbance has strong uncertainty. Due to the uncertainty of the actual distribution model and the existence of external unmeasurable interference, measurement errors, and linearization errors of nonlinear models. The existing voltage sag state estimation methods cannot guarantee estimation accuracy, robustness, and convergence under certain network structures and component parameter disturbances (bounded disturbances). The present invention adopts the Log-coth error evaluation function and the structural risk minimization state estimation model, so that the estimation model still has high accuracy and robustness under the conditions of strong noise, time-varying parameters, uncertain load and power supply interference in high-penetration new distribution networks. It plays an important role in the management and improvement of power quality of modern new distribution networks and in ensuring the safe and economic operation of distribution networks.

[0078] In this embodiment, a voltage sag disturbance signal model for state estimation is established for the non-stationary time-varying grid voltage, current amplitude, and phase of power injected into the distribution network by highly renewable distributed power sources, accompanied by frequency disturbances, non-periodic transients, and noise interference. An improved nonlinear filtering system is used to implement a novel distribution network voltage sag characteristic parameter detection and processing for state estimation. Under conditions such as small system frequency disturbances, transient oscillation interference, and white noise, the instantaneous fundamental amplitude and instantaneous phase tracking of the external input voltage and current are detected, achieving the dynamic phasor measurement of the voltage sag parameters described in step 1. The instantaneous voltage phasor at each monitoring point is used as the state estimation quantity, and the instantaneous voltage phasor at non-monitoring points is used as the state quantity. The stability and accuracy of a robust state-effective balance estimation algorithm for voltage sag with structural risk minimization is established using a highly robust Log-coth empirical risk error evaluation function and a second-order norm structural risk evaluation function for estimating model structural parameters. The robust state estimation model for voltage sag with structural risk minimization established in step 3 is then used to solve the distribution network voltage sag state estimation using a quasi-real-time incremental support vector machine algorithm driven by a hybrid model and data. By employing the Log-Coth error evaluation function and a structural risk minimization state estimation model, this paper achieves highly accurate and robust estimation under the conditions of strong noise, time-varying parameters, and uncertain load and power supply interference found in high-penetration new distribution networks. This plays a significant role in managing and improving the power quality of these networks and ensuring their safe and economical operation.

[0079] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

[0080] Various aspects of the present invention are described in the present disclosure with reference to the accompanying drawings, in which many illustrative embodiments are shown. Embodiments of the present invention are not necessarily defined to encompass all aspects of the present invention. It should be understood that the various concepts and embodiments described above, as well as those described in more detail below, may be implemented in any of a number of ways, as the concepts and embodiments disclosed herein are not limited to any particular implementation. In addition, some aspects of the present disclosure may be used alone or in any suitable combination with other aspects disclosed herein.

[0081] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A robust state estimation method for power grid voltage sag, characterized in that: include: S1. After establishing a voltage sag disturbance signal model for state estimation in an improved nonlinear filtering system, voltage sag monitoring waveform data is collected and input into the improved nonlinear filtering system; S2. Dynamic phasor measurement of voltage sag parameters at each monitoring point is performed through an improved nonlinear filtering system; S3. Establish a robust state estimation model for voltage sag with minimized structural risk, wherein the instantaneous voltage phasor of each monitoring point is used as the basic quantity measurement for state estimation, and the instantaneous voltage phasor of non-monitoring points is used as the state quantity; S4. Utilizing the voltage sag robust state estimation model with structural risk minimization and employing a model- and data-driven hybrid incremental support vector machine quasi-real-time algorithm, obtain analysis results of the robust state of the distribution network voltage sag; The establishing of a voltage sag disturbance signal model for state estimation includes: Among them, t * is the time when the voltage sag event occurs, T PQ is the duration of the voltage sag event, δU is the relative change in voltage amplitude after the system reaches a new steady state, g(t) is the harmonic and noise component, D PQ (tt * ) is the sag depth, U is the voltage rating, t is the time, f0 is the system frequency, is the fundamental phase before disturbance, is the fundamental phase during the disturbance, is the fundamental phase after the disturbance ends, g(t) is the non-periodic transient and noise, u i (t) is the input voltage, δ represents the hyperparameter, and i represents the sign of the input; The improved nonlinear filtering system is used to perform dynamic phasor measurement of the voltage sag parameters at each monitoring point, including: Establish an improved nonlinear filtering system: in, is the instantaneous phase of voltage, U(t) is the instantaneous amplitude of voltage, t is time, k1 and k2 are two constants of the filtering system, u i (t) is the input voltage, ω0 is the initial angular frequency constant of the filter system, and i is the symbol representing the input; Set the system parameters k1 and k2, and then calculate the instantaneous amplitude, frequency and phase of the dynamic phasor of each monitoring point based on the detected instantaneous amplitude and instantaneous phase; Aiming at the structural error caused by the matrix H itself, the voltage sag robust state estimation model with minimal structural risk is established as follows: Among them, C emp is the empirical risk penalty coefficient, m is the number of monitoring nodes, i.e., the number of measurements, For structural risk, U m is the voltage vector of the monitoring node, U u is the voltage vector of the non-monitoring point, and H is the measurement matrix.

2. The method according to claim 1, characterized in that In the process of establishing a robust state estimation model for voltage sag with structural risk minimization, the voltage vector of the monitoring node is obtained, including: According to the monitoring points and non-monitoring points, the complex power grid node admittance matrix is divided and the n-node voltage equation is obtained: Among them, Im is the injection current vector of the monitoring node, Iu is the injection current vector of the non-monitoring point, Ymm and Yuu are the self-admittance of the busbars of the monitoring node and the non-monitoring node respectively, and Ymu and Yum are the mutual admittance between the monitoring node and the non-monitoring node respectively; Estimate the apparent power of the equivalent load and source at non-monitoring points, where: Yu is the load equivalent admittance, Iu is the current phasor, Uu is the voltage phasor, is the apparent power, Uu(i) is the node voltage amplitude; A linear measurement equation for instantaneous voltage phasor state estimation is established: Um = H·Uu + ε, where ε is the measurement error; Get the voltage vector of the monitoring node 3. The method according to claim 2, characterized in that The process of establishing a robust state estimation model for voltage sag with minimal structural risk includes: The voltage sag state estimation model is preliminarily established using the Log-coth error evaluation function. Where m is the number of monitoring nodes, i.e., the number of measurements, and coth() is the hyperbolic cotangent function; According to the relationship between state variables and measurement vectors, the optimal state variable solution is obtained, and the basic model for voltage sag state estimation is established:

4. The method according to claim 2, characterized in that Also includes: The measurement matrix H is obtained by using the power grid SCADA section measurement data with a period of 1 / 30 minute.

5. The method according to claim 2, characterized in that In the process of obtaining the analysis results of the robust state of the distribution network voltage sag, it is necessary to collect measurement data to participate in the grid state estimation, including: Taking the SCADA section scanning time as the time reference, obtain all PQM data sections within a certain time range; The matching degree between the section of SCADA measurement data and the section of each PQM measurement data is calculated, and the section of PQM measurement data with the greatest matching degree is selected.

Images