Photovoltaic micro-grid grid-connected synchronous control method based on Beidou time service

Abstract

The invention discloses a photovoltaic micro-grid grid-connected synchronous control method based on Beidou time service, and relates to the technical field of power system synchronous control, and the method specifically comprises the steps: building a unified time reference of a whole system, calibrating node time through Beidou time service, setting a multi-mode data collection method for multi-mode type data, and screening effective data. A deviation range and a grid-connected state label are set; influences of voltage, frequency and phase deviation on grid-connected instability are quantified through mutual information entropy; a deviation contribution degree coefficient and a risk correction factor are calculated; a core feature is extracted; the method comprises the following steps: constructing Beidou time sequence fingerprints, calculating the similarity between the current Beidou time sequence fingerprints and a fingerprint sample library in real time, screening high-similarity samples to extract strategy elements, selecting an optimal control combination through multi-objective optimization, and adjusting output through information interaction between nodes to realize collaborative optimization.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of synchronous control of power systems, and particularly relates to a photovoltaic micro-grid grid-connected synchronous control method based on Beidou timing. BACKGROUND

[0002] With the transformation of global energy structure to clean and distributed, photovoltaic power generation, as one of the core forms of new energy, has become a key direction for the development of power systems, with its large-scale access to micro-grids and stable grid connection with large power grids. However, the existing technology still has the following core pain points in the synchronous control of photovoltaic micro-grid grid connection.

[0003] In recent years, the installed capacity of photovoltaic power has rapidly grown, and photovoltaic micro-grids are widely used in industrial parks, remote areas and other scenarios because they can flexibly consume distributed photovoltaic power and improve energy utilization efficiency. However, photovoltaic output is affected by natural conditions such as light and temperature, and has strong intermittency and high volatility, which leads to frequent fluctuations in micro-grid output power, and further causes voltage deviation, frequency fluctuation and phase shift of the grid connection interface. If these parameters cannot be synchronized in real time with the large power grid, it may cause grid connection power impact, or even trigger the action of the large power grid protection device, resulting in micro-grid disconnection and affecting the safe and stable operation of the power system. Therefore, the precision and dynamic response speed of grid connection synchronous control have become the key bottleneck restricting the large-scale application of photovoltaic micro-grids.

[0004] The existing time synchronization scheme has two problems: the security and reliability of timing are insufficient, and the deviation accumulation of local time synchronization causes distortion of the synchronization deviation calculation. Traditional control relies on preset fixed coefficients (such as voltage and frequency regulation coefficients), which cannot adapt to dynamic conditions such as photovoltaic output fluctuation and large power grid parameter jump. For example, when the photovoltaic output suddenly drops, the fixed coefficient control is easy to cause frequency deviation overshoot. Each node (photovoltaic inverter, energy storage converter) uses independent control and centralized instruction receiving mode, lacks self-organizing and collaborative ability, and the power distribution between nodes and the matching of load demand lag, further aggravating the synchronization deviation.

[0005] Therefore, there is an urgent need for a photovoltaic micro-grid grid connection synchronous control method based on Beidou timing to solve the above problems SUMMARY

[0006] The purpose of the present application is to provide a photovoltaic micro-grid grid connection synchronous control method based on Beidou timing, which solves the technical problems of insufficient dynamic condition adaptability, low distributed collaborative efficiency and disconnection between control strategy and dynamic condition in the prior art.

[0007] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0008] The photovoltaic micro-grid grid connection synchronous control method based on Beidou timing comprises:

[0009] Step one, build a unified time reference for the whole system, calibrate node time through Beidou timing, set up multi-mode data collection method for multi-modal type data, screen effective data, associate multi-target data at the same time point, and form an associated set;

[0010] Step two, collect historical deviation sequence based on data association set, set deviation range and grid-connected state label, quantify the influence of voltage, frequency and phase deviation on grid instability through mutual information, calculate deviation contribution coefficient and risk correction factor, and extract core features;

[0011] Step three, generate a unique time sequence identity based on the core feature sequence, roll over and associate abnormal events, build a Beidou time sequence fingerprint, calculate the similarity between the current Beidou time sequence fingerprint and the fingerprint sample library in real time, select high-similarity samples to extract strategy elements, combine abnormal event information, select the optimal control combination through multi-objective optimization, determine the core target and constraint setting, adjust the output through node information interaction, and realize collaborative optimization.

[0012] Further, for multi-modal type data, set up multi-mode data collection method, screen effective data, and the specific method is:

[0013] The voltage, frequency, phase of microgrid and power grid, and real-time data of photovoltaic output are collected at fixed intervals and high frequency, and the collection time points are evenly distributed according to the Beidou satellite timing signal;

[0014] The output power, load rate, energy storage state of charge, and reactive power demand data of each distributed node are collected at medium frequency and period, and the collection time points are strictly aligned with the control period of the corresponding node;

[0015] Set abnormal trigger collection conditions, when the Beidou timing signal strength mutates, the photovoltaic output fluctuation amplitude exceeds the normal range set based on historical effective data distribution, or the voltage, frequency and phase parameters of the power grid jump, automatically trigger supplementary collection, the supplementary collection takes the abnormal trigger time point as the center point, covers the time period of the preset time y before and after, and obtains the complete data in this time period through intensive collection.

[0016] Further, collect historical deviation sequence based on data association set, set deviation range and grid-connected state label, and the specific method is:

[0017] Collect real-time sequence of voltage deviation, frequency deviation and phase deviation in system historical operation data, set voltage deviation, frequency deviation and phase deviation range based on historical data;

[0018] Assign grid connection status tags to data acquisition time points. For each time point, when all deviations are within the normal range, the tag is 0, representing normal operation. If any deviation exceeds the limit or a parameter jump is detected, the tag is 1, representing an instability event.

[0019] Furthermore, the impact of voltage, frequency, and phase deviations on grid instability is quantified using mutual information entropy. The specific method is as follows:

[0020] The deviation data is normalized to eliminate dimensional differences. The time series data is divided into fixed time windows, and the statistical characteristics of the deviation within each window are extracted to form a deviation feature sample set. For each type of deviation and grid-connected status label sequence, the joint probability distribution and marginal probability distribution of deviation features and instability events are constructed respectively. The mutual information entropy formula is applied to calculate the mutual information entropy between the features of different deviations and instability events to obtain the mutual information entropy of voltage deviation. Mutual information entropy of frequency deviation Mutual information entropy of phase deviation .

[0021] Furthermore, the deviation contribution coefficient is calculated using the following method:

[0022] Determine the percentage of impact of a single type of deviation on grid connection stability using the formula. Calculate the deviation contribution coefficient, where x = u, f, o, k is the cumulative correction coefficient, t represents the current time, and T represents the cumulative time window duration. For a moment Real-time voltage deviation, For a moment The real-time frequency deviation For a moment The real-time phase deviation.

[0023] Furthermore, the risk correction factor is calculated using the following method:

[0024] Using formula This represents the voltage risk correction factor, used to reflect the current power grid's tolerance for voltage deviations. Indicates the real-time voltage of the large power grid. denoted by , where 'a' represents the rated voltage of the large power grid, and 'a' represents the coefficient reflecting the sensitivity of the power grid voltage deviation to the impact of risk factors.

[0025] Using formula This represents the frequency risk correction factor, used to reflect the current power grid's tolerance for frequency deviations. Indicates the real-time frequency of the large power grid. denoted by , b represents the rated frequency of the large power grid, and b represents the coefficient reflecting the sensitivity of the power grid frequency deviation to the impact of risk factors.

[0026] wherein, φ represents the phase risk correction factor, which is used to reflect the tolerance of the current power grid to the phase deviation, and wherein, σ 2 represents the variance of the phase fluctuation, and c represents a coefficient reflecting the sensitivity of the influence of the phase deviation of the power grid on the risk factor;

[0027] a, b, and c are constant values, which are set according to historical data in combination with actual requirements of the power grid system.

[0028] Further, core features are extracted, and the specific method is as follows:

[0029] The voltage deviation, the frequency deviation, and the phase deviation are integrated through the deviation contribution coefficient and the risk correction factor, and the formula is as follows: wherein, x = u, f, o, represents the real-time deviation value, represents the risk correction factor of x;

[0030] Based on the photovoltaic output time series data, the output change rate and the fluctuation amplitude are calculated by the sliding window method, the sliding window length is determined, the product of the output difference between the head and tail time points in the sliding window and the fluctuation amplitude is divided by the sliding window length, and the fluctuation trend feature is obtained.

[0031] Based on the output power and the load rate data of each distributed node, the parameter difference degree between nodes is calculated, the average output power of all nodes is determined, the sum of the absolute values of the difference between the output power of each single node and the average output power is calculated, and then the sum is divided by the number of nodes and the average output power, to obtain the collaborative consistency feature .

[0032] Further, based on the core feature sequence, a unique time sequence identity is generated, and the identity is updated and associated with abnormal events to construct a Beidou time sequence fingerprint, and the specific method is as follows:

[0033] Taking the high-frequency collection period of Beidou time service as a time unit, the feature sequence of continuous M periods is extracted, the three-dimensional feature vector of each period is normalized and encoded by associating with the Beidou timestamp, the feature value is mapped to the interval [0, 1], the Beidou timestamp information is embedded in the encoding process, and through multi-dimensional time sequence hash mapping, the encoding vector of M periods is converted into a fixed-length hash value to form the BD-TFP of the current working condition. The seed value of the hash mapping is dynamically generated by the Beidou time service day timestamp to avoid fingerprint collision.

[0034] ​Each new high-frequency acquisition cycle, BD-TFP rolling update, establish the association index of BD-TFP and abnormal events, embed the time stamp of Beidou signal anomaly, photovoltaic output sudden change, large grid parameter jump abnormal event into the extension field of corresponding BD-TFP, form the association chain of fingerprint and event, generate massive BD-TFP samples based on historical operation data, each sample is labeled with the corresponding optimal control effect label, and a fingerprint sample library is constructed.

[0035] Further, the similarity of the current Beidou timing fingerprint and the fingerprint sample library is calculated in real time, and the high-similarity sample extraction strategy elements are screened, and the specific method is:

[0036] Domain formula , which represents the similarity of the current Beidou timing fingerprint and the fingerprint sample library, wherein, , which represents the feature encoding of the i-th period of the current BD-TFP, , which represents the feature encoding of the i-th period of the fingerprint library sample, , which represents the current fingerprint timestamp, , which represents the fingerprint library sample timestamp, is a time decay constant, which is set based on the Beidou time synchronization accuracy;

[0037] Based on historical data and actual situation, set the similarity threshold Si1, screen out the fingerprint library samples with similarity greater than or equal to Si1, and extract the corresponding control strategy core elements

[0038] Based on the strategy elements of the screened fingerprint library samples, combined with the event association information of the current BD-TFP (such as containing photovoltaic output sudden change event, then activate the output), through multi-objective optimization algorithm (NSGA-Ⅲ) Select the optimal combination and execution logic, the optimization objectives are: the control response delay is minimum (based on Beidou timestamp calculation); the control energy consumption is minimum; the control effect is optimal, the constraint conditions are module interface timing compatibility (ensure that the execution time of each module is aligned through Beidou timestamp).

[0039] Design timing consensus algorithm (BD-TCA), with as the core target, and , as the consensus constraint condition, realize node self-organization and cooperation, determine the maximum value of the synchronization deviation in the historical data that does not cause instability , which is taken as the critical threshold F1, determine the maximum fluctuation amplitude of the historical photovoltaic output fluctuation that does not induce >F1, which is taken as the warning threshold F2.

[0040] Only when <F1, <F2, start the complete consensus iteration;

[0041] Based on the requirements for equipment operating accuracy and collaborative stability, historical collaborative imbalance conditions were selected. The largest difference value is used as C1 to filter historical collaborative imbalance conditions. The minimum difference value is taken as C2.

[0042] When nodes exchange consensus variables, if adjacent nodes If the difference is greater than C1, then the interaction cycle should be extended (to avoid exacerbating the coordination imbalance). If the difference is less than or equal to C2, the cycle is shortened, and consensus convergence is accelerated.

[0043] Considering node power allocation efficiency and system robustness requirements, we statistically analyzed historical operations and found that node collaboration did not affect system stability. The minimum value is used as the lower limit threshold F3;

[0044] When <F3, the node is promoted. Efforts were made to make core adjustments and real-time monitoring was conducted during the iteration process. , The change, if Greater than or equal to the critical threshold or If the value is greater than or equal to the warning threshold, the iteration should be paused immediately.

[0045] Based on the allowable deviation range of equipment output power and the requirements for collaborative control accuracy, and through statistical analysis of historical consensus convergence cases, we determine the factors that can guarantee the uniformity of power distribution. The maximum standard deviation is used as the convergence threshold C3.

[0046] The core criterion for consensus convergence is that all nodes... The standard deviation of C3 must be ≤ C3, and the following conditions must also be met: <F1, <F2, ensuring that collaborative optimization does not introduce new risks, and achieving the goal of collaborative improvement.

[0047] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0048] 1. This invention quantifies the impact weights of voltage, frequency, and phase deviations on grid instability by using mutual information entropy. It combines the deviation contribution coefficient and risk correction factor calculated under real-time operating conditions to generate comprehensive characteristics of synchronization deviation, fluctuation trend characteristics, and coordination consistency characteristics. These characteristics can accurately map the core causes of instability risk, the intensity of power output fluctuations, and the coordination status of nodes. Based on these characteristics, a BeiDou time-series fingerprint is constructed. The optimal strategy direction is locked through dynamic similarity matching, which breaks away from the limitations of traditional control that triggers strategies according to fixed rules and deeply binds control priority with risk causes.

[0049] 2. This invention uses feature calculations that combine historical data with real-time operating conditions to enable core features to adapt to dynamic operating conditions such as photovoltaic power output fluctuations and large power grid parameter jumps. On this basis, on the one hand, it uses the event association dynamic adjustment strategy of Beidou time-series fingerprints, such as shortening the consensus iteration cycle and increasing the risk factor weight when the signal is interrupted. On the other hand, it uses feature, strategy and effect closed-loop mining to periodically optimize fingerprint matching thresholds, consensus convergence parameters, etc., to achieve the self-evolution of control strategy.

[0050] 3. The collaborative consistency feature set in this invention provides a quantitative target for distributed collaboration. Based on this, the time-series consensus algorithm uses the BeiDou timestamp as the interaction benchmark, allowing each node to reach a power allocation consensus through fingerprint fragment verification and time-series consistency judgment, without relying on the global data aggregation of a centralized control center; thus solving the problems of delayed response of distributed nodes and unbalanced power allocation. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 The diagram illustrates the steps of the photovoltaic microgrid grid-connected synchronous control method based on BeiDou time synchronization according to the present invention.

[0053] Figure 2 The diagram illustrates the steps of the deviation analysis and feature extraction method of the present invention.

[0054] Figure 3 The diagram illustrates the steps of the timing fingerprint optimization control method of the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] like Figure 1 , Figure 2 , Figure 3 The grid-connected synchronization control method for photovoltaic microgrids based on BeiDou time synchronization shown includes the following steps:

[0057] Step 1: Establish a unified time reference for the entire system. Use BeiDou time synchronization to calibrate node times. For multi-modal data types, set up multi-mode data acquisition methods, filter valid data, and associate multiple target data at the same time point to form an associated set.

[0058] In the centralized control center and all distributed nodes including photovoltaic inverters, energy storage converters, and grid-connected interfaces, BeiDou dual-mode receivers and anti-interference components are installed, along with local redundant timekeeping units. This ensures that each node can stably receive BeiDou satellite timing signals, establishing a unified time reference for the entire system. Through a high-precision time synchronization protocol, using BeiDou satellite time as the reference, time calibration is performed on the control units of each node to eliminate time asynchrony deviations between nodes, forming a globally consistent time reference system. A zero-logic for abnormal switching of timing signals is set, automatically activating the local redundant timekeeping unit when the BeiDou signal is interrupted or attenuated to maintain the continuity of the time reference. This unifies the communication protocols and data transmission formats of each node, establishing a low-latency, highly reliable industrial communication link to ensure efficient transmission of timing signals, data signals, and control commands between nodes.

[0059] The voltage, frequency, phase, and real-time photovoltaic output data of microgrids and large power grids are collected at fixed intervals using high-frequency acquisition, with the acquisition time points evenly distributed according to BeiDou time.

[0060] The output power, load rate, energy storage state of charge, and reactive power demand data of each distributed node are collected using medium-frequency periodic acquisition, with the acquisition time points aligned with the node control cycle.

[0061] Anomaly triggering conditions are set. When the BeiDou signal strength changes abruptly, the photovoltaic power output fluctuation exceeds the normal range, or the parameters of the power grid change abruptly, supplementary data acquisition is automatically triggered. The supplementary data acquisition period is centered on the anomalous trigger time point and covers the time period before and after it by y hours. Data is collected intensively during the supplementary data acquisition period. All data acquisition time points are based on the BeiDou time signal. Before each acquisition, the acquisition trigger time is calibrated by the local time signal unit of the node to ensure that the acquisition time points of different nodes and different types of data are strictly aligned.

[0062] Based on pre-set physical constraints and equipment operating range, abnormal data exceeding the range is eliminated. The BeiDou timestamp is used to verify the data's time sequence integrity and consistency. For fluctuating data (such as photovoltaic power output and time signal strength), a dynamic fluctuation threshold is set based on the distribution of historical valid data to eliminate extreme abnormal data (such as instantaneous jump data caused by sudden electromagnetic interference) that exceeds the reasonable fluctuation range, thus obtaining the filtered valid data.

[0063] For the filtered valid data, a unique BeiDou timestamp (including date and time identifier) ​​is attached, and the data is stored according to the logical classification of target type, data category and timestamp. Different target data at the same point in time (such as voltage, frequency and phase data related to synchronization accuracy, and signal strength pre-synchronization error data related to timing health) are associated through timestamps to form a data association set.

[0064] Step 2: Collect historical deviation sequences based on data association sets, set deviation range and grid connection status labels, quantify the impact of voltage, frequency and phase deviations on grid connection instability through mutual information entropy, calculate deviation contribution coefficient and risk correction factor, and extract core features.

[0065] Collect real-time sequences of voltage deviation, frequency deviation, and phase deviation from the historical operating data of the system (voltage deviation, frequency deviation, and phase deviation are obtained by calculating the difference between the measured value and the rated value or reference value (such as the standard value of the large power grid)). Set the range of voltage deviation, frequency deviation, and phase deviation based on historical data (such as voltage deviation exceeding ±5% and frequency deviation exceeding ±0.2Hz).

[0066] Assign grid connection status tags to data acquisition time points. For each time point, when all deviations are within the normal range, the tag is 0, representing normal operation; if any deviation exceeds the limit or a parameter jump is detected (e.g., a voltage jump exceeding 5% is considered a jump, and a frequency jump exceeding 0.2Hz is considered a jump), the tag is 1, representing an instability event.

[0067] The three types of deviation data are normalized to eliminate dimensional differences. The time-series data are divided into fixed time windows, and the statistical characteristics (mean, fluctuation amplitude, and rate of change) of the deviation within each window are extracted to form a deviation feature sample set. For each type of deviation and grid-connected status label sequence, a joint probability distribution and marginal probability distribution of deviation features and instability events are constructed. The marginal probability distribution describes the probability of occurrence of deviation features and the probability of occurrence of instability events, while the joint probability distribution describes the probability of a certain deviation feature and an instability event occurring simultaneously. Based on the constructed probability distributions, the mutual information entropy of different deviation features and instability events is calculated using the mutual information entropy formula to obtain the mutual information entropy of voltage deviation. Mutual information entropy of frequency deviation Mutual information entropy of phase deviation The impact of different deviations on grid instability is quantified by mutual information entropy.

[0068] By determining the proportion of the impact of a single type of deviation on grid connection stability, the formula is used. Calculate the deviation contribution coefficient, where x = u, f, o. Based on the cumulative time window T, statistically analyze the average cumulative deviation under historical normal operating conditions, and use the corresponding correction intensity as the correction coefficient, denoted by k (range 0.1-0.5), where t represents the current time and T represents the duration of the cumulative time window. For a moment Real-time voltage deviation, For a moment The real-time frequency deviation For a moment The phase deviation is dynamically adjusted in real time by the deviation contribution coefficient to follow the current deviation distribution, which both preserves historical patterns and fits real-time operating conditions.

[0069] Using formula This represents the voltage risk correction factor, used to reflect the current power grid's tolerance for voltage deviations. Indicates the real-time voltage of the large power grid. denoted by , where 'a' represents the rated voltage of the large power grid, and 'a' represents the coefficient reflecting the sensitivity of the power grid voltage deviation to the impact of risk factors.

[0070] Using formula This represents the frequency risk correction factor, used to reflect the current power grid's tolerance for frequency deviations. Indicates the real-time frequency of the large power grid. denoted by , b represents the rated frequency of the large power grid, and b represents the coefficient reflecting the sensitivity of the power grid frequency deviation to the impact of risk factors.

[0071] Using formula This represents the phase risk correction factor, used to reflect the current power grid's tolerance for phase deviations. denoted by , where is the variance of phase fluctuations, and c is the coefficient reflecting the sensitivity of the power grid phase deviation to the impact of risk factors.

[0072] a, b, and c are all constant values, set according to historical data and actual needs of the power grid system. The sum of a, b, and c is fixed. Based on historical instability data, the probability of instability caused by the three types of deviations is statistically analyzed. If the proportion of instability caused by voltage deviation is higher, the value of a is increased, and the values ​​of b and c are adjusted at the same time. Conversely, the value of a is decreased. In this embodiment, a+b+c=2 is set.

[0073] By integrating voltage deviation, frequency deviation, and phase deviation with the deviation contribution coefficient and risk correction factor, the formula is used. This indicates the comprehensive characteristics of synchronization deviation. Used to quantify the degree of overall synchronization deviation, where x = u, f, o. Indicates the real-time deviation value. This represents the risk correction factor for x.

[0074] Based on photovoltaic power output time-series data, the rate of change and fluctuation amplitude of power output are calculated using the sliding window method. The duration of the sliding window is then determined. The product of the power output difference at the beginning and end of the sliding window and the fluctuation amplitude is divided by the duration of the sliding window to obtain the fluctuation trend characteristics. ;

[0075] Based on the output power and load rate data of each distributed node, the parameter difference between nodes is calculated to determine the average output power of all nodes. The sum of the absolute values ​​of the differences between the output power of each individual node and the average output power is then calculated and divided by the number of nodes and the average output power to obtain the collaborative consistency characteristics. ;

[0076] Step 3: Generate a unique temporal identity based on the core feature sequence, continuously update and associate with abnormal events, construct a BeiDou temporal fingerprint, calculate the similarity between the current BeiDou temporal fingerprint and the fingerprint sample library in real time, select high similarity samples to extract strategy elements, combine abnormal event information, select the optimal control combination through multi-objective optimization, determine the core objectives and constraint settings, and adjust the output through information interaction between nodes to achieve collaborative optimization.

[0077] Based on the comprehensive characteristics of synchronization deviation, fluctuation trend, and coordination consistency, a BeiDou time-series fingerprint (BD-TFP) is constructed. This transforms dispersed features into unique and traceable dynamic time-series identifiers. Using the high-frequency acquisition cycle of BeiDou time synchronization as the time unit, feature sequences of M consecutive cycles are extracted. ,in Indicates the start time of the BeiDou timestamp. The data collection period is represented by a normalized encoding method that associates the three-dimensional feature vector of each period with the BeiDou timestamp. The feature values ​​are mapped to the interval [0, 1]. During the encoding process, BeiDou timestamp information is embedded, such as using the last three digits of the timestamp as the encoding offset to ensure that the time fingerprint is unique for different times and different working conditions. Through multi-dimensional time hash mapping, the encoding vector of M periods is converted into a hash value of fixed length (such as 256 bits) to form the BD-TFP of the current working condition. The seed value of the hash mapping is dynamically generated by the daily timestamp of BeiDou time synchronization to avoid fingerprint collisions.

[0078] For each new high-frequency acquisition cycle, the BD-TFP is updated on a rolling basis (removing the feature vector of the earliest cycle and adding the data of the latest cycle). An association index between BD-TFP and abnormal events is established. The timestamps of abnormal events such as Beidou signal anomalies, sudden changes in photovoltaic power output, and abnormal changes in large power grid parameters are embedded into the extended fields of the corresponding BD-TFP to form a fingerprint-event association chain. Based on historical operating data (including normal operating conditions and operating conditions where abnormal events occur), a massive number of BD-TFP samples are generated. Each sample is labeled with the corresponding optimal control effect label (such as deviation correction efficiency, fluctuation suppression rate, and coordination consistency improvement degree) to build a fingerprint sample library.

[0079] The current BD-TFP is collected in real time and matched with fingerprint samples in the fingerprint database using a time-series dynamic similarity algorithm (breaking through the traditional static similarity calculation). The specific calculation formula is as follows:

[0080] ;

[0081] in, This represents the feature encoding of the i-th period of the current BD-TFP. This represents the feature encoding of the i-th period of the fingerprint database sample. Indicates the current fingerprint timestamp. Indicates the timestamp of the fingerprint database sample. The time decay constant is set to 10s based on the BeiDou time synchronization accuracy.

[0082] Based on historical data and actual conditions, a similarity threshold Si1 is set, and fingerprint samples with a similarity greater than or equal to Si1 are selected. The core elements of their corresponding control strategies (such as the combination of control modules and the direction of adjustment of key parameters) are then extracted.

[0083] Based on the policy elements of the screened fingerprint database samples, combined with the event association information of the current BD-TFP (such as prioritizing the activation of output if there is a sudden change in photovoltaic output), the optimal combination and execution logic are selected through a multi-objective optimization algorithm (NSGA-Ⅲ). The optimization objectives are: minimum control response delay (calculated based on BeiDou timestamp); minimum control energy consumption; and optimal control effect. The constraint is the timing compatibility of module interfaces (ensuring that the execution time of each module is aligned with the BeiDou timestamp).

[0084] Design a timing consensus algorithm (BD-TCA) to... With the improvement as the core objective, , As a consensus constraint, it enables self-organizing collaboration among nodes, ensuring that synchronization deviations in historical data do not lead to instability. The maximum value is used as the critical threshold F1 to determine whether historical photovoltaic power output fluctuations have been induced. The maximum fluctuation range of F1 is used as the warning threshold F2.

[0085] Only when <F1 and A full consensus iteration is initiated only when F2 is reached; otherwise, the full time-series consensus iteration is paused to prevent the coordination process from exacerbating instability. ≥F1, focusing on voltage, frequency and phase deviation correction, based on deviation contribution coefficient and risk correction factor, prioritize adjusting parameters with higher impact weight; ≥F2, quickly suppresses photovoltaic output fluctuations (such as energy storage charging and discharging compensation, photovoltaic output limiting);

[0086] Based on the requirements for equipment operating accuracy and collaborative stability, historical collaborative imbalance conditions were selected. The largest difference value is used as C1 to filter historical collaborative imbalance conditions. The minimum difference value is taken as C2.

[0087] When nodes exchange consensus variables, if adjacent nodes If the difference is greater than C1, then extend the interaction period (e.g., from 100ms to 150ms) to avoid exacerbating the coordination imbalance. If the difference is less than or equal to C2, the cycle is shortened, and consensus convergence is accelerated.

[0088] Considering node power allocation efficiency and system robustness requirements, we statistically analyzed historical operations and found that node collaboration did not affect system stability. The minimum value is used as the lower limit threshold F3;

[0089] When <F3, the node is promoted. Efforts were made to make core adjustments and real-time monitoring was conducted during the iteration process. , The change, if Greater than or equal to the critical threshold or If the value is greater than or equal to the warning threshold, the iteration should be paused immediately.

[0090] Based on the allowable deviation range of equipment output power and the requirements for collaborative control accuracy, and through statistical analysis of historical consensus convergence cases, we determine the factors that can guarantee the uniformity of power distribution. The maximum standard deviation is used as the convergence threshold C3.

[0091] The core criterion for consensus convergence is that all nodes... The standard deviation of C3 must be ≤ C3, and the following conditions must also be met: <F1, <F2, ensuring that collaborative optimization does not introduce new risks, and achieving the goal of collaborative improvement.

[0092] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0093] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A photovoltaic micro-grid grid-connected synchronization control method based on Beidou timing, characterized in that, Comprise: Step one, build a unified time reference system, calibrate node time through Beidou timing, set up multi-mode data collection method for multi-modal data, screen effective data, associate multi-target data at the same time point, form an associated set; Step two, collect historical deviation sequence based on data association set, set deviation range and grid-connected state label, quantify the influence of voltage, frequency and phase deviation on grid instability through mutual information entropy, calculate deviation contribution coefficient and risk correction factor, and extract core features; Step three, generate a unique time sequence identity based on the core feature sequence, roll over and associate abnormal events, build a Beidou time sequence fingerprint, calculate the similarity between the current Beidou time sequence fingerprint and the fingerprint sample library in real time, select high-similarity samples to extract strategy elements, combine abnormal event information, select the optimal control combination through multi-objective optimization, determine the core target and constraint setting, adjust the output through node information interaction, and realize collaborative optimization. 2.The method of claim 1, wherein, For multi-modal data, set up multi-mode data collection method, screen effective data, the specific method is: The voltage, frequency, phase of microgrid and power grid, and real-time data of photovoltaic output are collected at fixed intervals and high frequency, and the collection time points are evenly distributed according to the Beidou satellite timing signal; The output power, load rate, energy storage state of charge, and reactive power demand data of each distributed node are collected at medium frequency and period, and the collection time points are strictly aligned with the control period of the corresponding node; Set abnormal trigger collection conditions, when the Beidou timing signal strength mutates, the photovoltaic output fluctuation amplitude exceeds the normal range set based on the historical effective data distribution, or the voltage, frequency and phase parameters of the power grid jump, automatically trigger supplementary collection, the supplementary collection takes the abnormal trigger time point as the center point, covers the time period of the preset time y before and after, and obtains the complete data in this time period through intensive collection. 3.The method of claim 1, wherein, Collect historical deviation sequence based on data association set, set deviation range and grid-connected state label, the specific method is: Collect real-time sequence of voltage deviation, frequency deviation and phase deviation in system historical operation data, set voltage deviation, frequency deviation and phase deviation range based on historical data; Assign grid-connected state label to data collection time points, for each time point, when all deviations are within the normal range, the label is 0, representing normal operation, if any deviation is out of limit or parameter jump is detected, the label is 1, representing instability event. 4.The method of claim 1, wherein, Quantify the influence of voltage, frequency and phase deviation on grid instability through mutual information entropy, the specific method is: The deviation data is normalized to eliminate dimension difference, the time series data is segmented according to a fixed time window, statistical features of the deviation in each window are extracted to form a deviation feature sample set, a joint probability distribution of the deviation features and the instability events and an edge probability distribution are respectively constructed for each type of deviation and the grid-connected state label sequence, mutual information entropy formulas are applied to calculate mutual information entropy of different deviation features and the instability events, and mutual information entropy of voltage deviation , mutual information entropy of frequency deviation , and mutual information entropy of phase deviation . 5.The method of claim 1, wherein, Calculate deviation contribution coefficient, the specific method is: Determine the proportion of single-class deviation affecting grid stability, use formula Calculate the deviation contribution coefficient, where x = u, f, o, k is the cumulative correction coefficient, t represents the current time, T represents the cumulative time window length, The voltage real-time deviation at time , The frequency real-time deviation at time , The phase real-time deviation at time . 6.The method of claim 1, wherein, Calculate risk correction factor, the specific method is: Using the formula represents the voltage risk correction factor, which reflects the tolerance of the current power grid to voltage deviation, wherein, represents the real-time voltage of the large power grid, represents the rated voltage of the large power grid, and a represents a coefficient reflecting the sensitivity of the influence of the voltage deviation of the power grid on the risk factor; Using the formula represents the frequency risk correction factor, which reflects the tolerance of the current power grid to frequency deviation, where, represents the real-time frequency of the large power grid, represents the rated frequency of the large power grid, and b represents a coefficient reflecting the sensitivity of the influence of the frequency deviation of the power grid on the risk factor; Using the formula wherein φr denotes the phase risk correction factor, reflecting the current grid tolerance to phase deviation, and wherein wherein σ2φ denotes the variance of the phase fluctuation and c denotes a coefficient reflecting the sensitivity of the grid phase deviation to the risk factor. a, b, c are constant values, set according to historical data and actual demand of power grid system, and the sum of a, b and c is fixed. 7.The method of claim 1, wherein, Extract core features, the specific method is: The voltage deviation, frequency deviation, phase deviation are integrated by the deviation contribution coefficient and the risk correction factor through formula representing the synchronous deviation comprehensive characteristics, wherein x = u, f, o, representing the real-time deviation value, representing the risk correction factor of x; Based on photovoltaic output time series data, the output change rate and fluctuation amplitude are calculated by the sliding window method, the sliding window length is determined, the product of the output difference between the beginning and end time in the sliding window and the fluctuation amplitude is divided by the sliding window length to obtain the fluctuation trend feature ; Based on the output power and load rate data of each distributed node, the parameter difference degree between nodes is calculated, the average output power of all nodes is determined, the sum of absolute values of the difference between the output power of each single node and the average output power is calculated, and then divided by the number of nodes and the average output power, to obtain the cooperative consistency feature . 8.The method of claim 1, wherein, Generate a unique time sequence identity based on the core feature sequence, roll over and associate abnormal events, build a Beidou time sequence fingerprint, the specific method is: Taking the high-frequency acquisition period of Beidou timing as the time unit, the feature sequence of M continuous periods is extracted, the three-dimensional feature vector of each period is normalized and encoded with Beidou timestamp association, the feature value is mapped to the interval [0, 1], the Beidou timestamp information is embedded in the encoding process, and the M period encoding vectors are converted into fixed length hash values through multi-dimensional time sequence hash mapping to form the Beidou time sequence fingerprint of the current working condition. The seed value of hash mapping is dynamically generated by the daily timestamp of Beidou timing to avoid fingerprint collision; The Beidou time sequence fingerprint is updated rolling with each new high-frequency acquisition period, and the association index of Beidou time sequence fingerprint and abnormal event is established. The timestamps of Beidou signal abnormality, photovoltaic output sudden change, and large power grid parameter jump abnormal event are embedded in the expansion field of the corresponding Beidou time sequence fingerprint to form the association chain of fingerprint and event. Based on the historical operation data, a large number of Beidou time sequence fingerprint samples are generated, each sample is labeled with the corresponding optimal control effect label, and a fingerprint sample library is constructed. 9.The method of claim 1, wherein, The similarity between the current Beidou time sequence fingerprint and the fingerprint sample library is calculated in real time, and the high-similarity sample strategy elements are screened out. The specific method is: Domain formula represents the similarity between the current Beidou timing fingerprint and the fingerprint sample library, wherein, represents the feature code of the i th period of the current Beidou timing fingerprint, represents the feature code of the i th period of the fingerprint library sample, represents the current fingerprint timestamp, represents the fingerprint library sample timestamp, is a time decay constant, which is set based on the Beidou time synchronization accuracy; Based on historical data and actual situation, set the similarity threshold Si1, select the fingerprint library samples with similarity greater than or equal to Si1, and extract the corresponding control strategy core elements.

Images