Self-adaptive control method for improving new energy power generation grid-connected stability of micro-grid

Abstract

The invention provides a self-adaptive control method for improving new energy power generation grid-connected stability of a micro-grid, and the method comprises the steps: carrying out the description according to a corrected phase change trend, fusing the real-time feedback data of voltage deviation, generating a reactive compensation instruction sequence segmentation adjustment scheme, and determining a preliminary compensation capacity dynamic distribution strategy; the microgrid system is regulated and controlled in real time through the optimized control instruction sequence in combination with the power grid stability constraint condition, the actual execution effect of the compensation action is obtained, and whether the system voltage deviation meets the stability requirement or not is judged; according to the stability requirement judgment result of the system voltage deviation, the execution effect is fed back to a micro-grid control layer, a closed-loop mechanism is adopted to carry out iterative updating on the instruction sequence segmentation adjustment scheme, and a final reactive compensation regulation and control scheme is determined.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and particularly relates to a self-adaptive control method for improving the stability of new energy power generation grid connection of micro-grid. BACKGROUND

[0002] In modern power systems, micro-grid, as an important form integrating new energy generation and distributed energy management, its stable operation has an invaluable value for ensuring energy supply and promoting green development.

[0003] Especially in the scenario of new energy power generation grid connection, how to maintain power balance and voltage stability of the grid becomes the core issue of improving energy utilization efficiency and system reliability.

[0004] Due to the small scale, large load fluctuation and intermittent characteristics of new energy generation of micro-grid, higher requirements are put forward for the dynamic regulation of reactive power, and the imbalance of reactive power often leads to voltage fluctuation and even system instability.

[0005] However, in the field of reactive power compensation of micro-grid, many methods are not up to the task when facing complex operating environments.

[0006] Traditional compensation methods often have difficulty in adapting to the dual challenges of random changes in new energy generation output and rapid fluctuations in load demand, especially when the system operating state frequently switches, the response speed and regulation accuracy of compensation equipment often cannot meet the demand.

[0007] This deficiency not only affects the power factor optimization of the grid, but also may exacerbate voltage deviation, and thus threatens the safe operation of the entire micro-grid.

[0008] Focusing on the technical challenges, a key difficulty in reactive power compensation in micro-grid is how to real-time perceive and respond to the phase difference changes between voltage and current.

[0009] As an important indicator reflecting power factor, the dynamic fluctuation of phase difference directly affects the reactive power balance of the system.

[0010] Due to the uncertainty of the output of new energy generation equipment, the change of phase difference often presents nonlinear characteristics, which makes it difficult for compensation equipment to accurately match the actual demand when adjusting.

[0011] Further, this nonlinear fluctuation also causes the hysteresis of compensation action, causing the system to be unable to reach the ideal power factor state in a short time.

[0012] For example, in a microgrid connected to a wind power grid, when a sudden change in wind speed causes a sharp fluctuation in power generation, the compensation equipment may be unable to adjust the phase difference in time, resulting in abnormal voltage rises or falls, which directly affects the power quality at the user end.

[0013] Therefore, how to dynamically capture phase difference changes and achieve accurate reactive power compensation in the complex environment of new energy power generation grid connection has become a key issue in improving the stability of microgrids.

[0014] Solving this problem is not only related to the reliability of system operation, but also directly affects the efficient utilization of new energy sources and the overall performance optimization of microgrids. Summary of the Invention

[0015] This invention provides an adaptive control method for improving the grid connection stability of new energy power generation in microgrids, mainly including: Output fluctuation signals and load change data of new energy power generation are collected from the microgrid system. Frequency domain analysis technology is used to process the collected data, extract phase offset dynamic monitoring features, and obtain basic descriptive information of phase change. Based on the phase offset dynamic monitoring features and combined with historical data time series, a periodic analysis model of phase change fluctuation is constructed, prediction time windows are divided, phase abrupt change points are identified, and the future trend prediction results of phase change are obtained. If the phase abrupt change point exceeds the preset fluctuation range, the trend extrapolation logic is adjusted through data smoothing preprocessing technology, and combined with prediction error correction, a corrected phase change trend description is obtained. Based on the corrected phase change trend description, real-time voltage deviation feedback data is integrated to generate a segmented adjustment scheme for reactive power compensation command sequence, and a preliminary dynamic allocation strategy for compensation capacity is determined. If the strategy finds potential inconsistencies in command execution conflict detection, the command priority sorting is adjusted through control delay optimization processing technology, and combined with equipment response time evaluation data, an optimized control command sequence is obtained. Through the optimized control command sequence, combined with grid stability constraints, the microgrid system is controlled in real time to obtain the actual execution effect of compensation actions and determine whether the system voltage deviation meets the stability requirements. Based on the judgment results, the execution effect is fed back to the microgrid control layer, and a closed-loop mechanism is used to iteratively update the segmented adjustment scheme of the command sequence to determine the final reactive power compensation control scheme.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a reactive power compensation and control method for microgrids based on phase deviation. It addresses the problem of voltage instability caused by phase deviation exceeding the range due to fluctuations in renewable energy generation output and load changes in microgrids. This problem is logically related to real-time monitoring of phase dynamics, prediction of future trends, and dynamic adjustment of compensation strategies to maintain system stability. This invention collects real-time data through a sensor network, extracts phase deviation features using frequency domain analysis, constructs a periodic fluctuation model based on historical sequences, identifies abrupt changes, and predicts trends. If the prediction exceeds the range, the trend description is adjusted through data smoothing and error correction, and a compensation command sequence is generated by integrating voltage feedback. If conflicts exist, the command priority and delay are optimized, and control is performed in conjunction with stability constraints. The scheme is updated through closed-loop iteration, ultimately achieving precise reactive power compensation and control. This invention significantly improves the voltage stability and operating efficiency of microgrids, reduces the impact of fluctuations, and ensures efficient integration of renewable energy. Attached Figure Description

[0017] Fig. 1 The flowchart shows an adaptive control method for improving the grid connection stability of new energy power generation in microgrids according to the present invention. Fig. 2 This is a schematic diagram of an adaptive control method for improving the grid connection stability of new energy power generation in microgrids according to the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0019] like Figs. 1-2 This embodiment of an adaptive control method for improving the grid connection stability of new energy power generation in microgrids may specifically include: Step S101: Collect power output fluctuation signals and load change data of new energy power generation from the microgrid system, process the collected data using frequency domain analysis technology, extract phase offset dynamic monitoring features, and obtain basic descriptive information of phase change.

[0020] The system collects power output fluctuation signals and load change data from the microgrid system using a sensor network. A dedicated signal acquisition module performs initial filtering on the raw data to obtain a processed initial signal dataset. Based on this dataset, frequency domain analysis is used to transform the signal, decomposing it into different frequency components and obtaining its frequency domain distribution information. Specific frequency components related to phase shift are extracted from this frequency domain distribution information to determine the dynamic characteristics of phase shift changes. If the dynamic characteristics of phase shift changes exceed a preset threshold, further time-domain reconstruction is performed on these frequency components to obtain detailed time-series data of phase changes. By segmenting and analyzing the detailed time-series data of phase changes, key time points of drastic changes are identified, and abnormal intervals of phase change are determined. Based on the data characteristics of these abnormal intervals and the correspondence between power output fluctuations and load changes, a support vector machine algorithm is used to classify the causes of these abnormal intervals and determine the specific sources of abnormal phases. Corresponding data tags are generated for each specific source of abnormal phases and stored in the system for direct retrieval during subsequent analysis, providing a reference for anomaly handling.

[0021] In one possible implementation, current transformers (CTs) and voltage transformers (PTs) deployed in the microgrid system collect power output fluctuation signals and load change data from the microgrid. A dedicated signal acquisition module performs denoising (removing power line noise, high-frequency switching noise, etc., improving the signal-to-noise ratio, and ensuring the accuracy of frequency domain analysis), standardization / format conversion (converting the acquired digital quantities into unified physical dimensions (such as voltage and current values) and time alignment for unified mathematical processing), and outlier removal (identifying and eliminating obviously erroneous data points caused by sensor failures or transient interference) on the raw data to obtain a processed initial signal dataset. This process provides a clean data foundation for subsequent effective frequency domain analysis. The signal acquisition module is an integrated hardware unit used to receive analog signals from the sensor network and convert them into digital data. This module typically includes the following key functions:

[0022] a. Isolation and amplification circuits: ensure safety and improve the accuracy of small signals.

[0023] b. Anti-aliasing filter: Filters out signals with frequencies higher than the Nyquist frequency before A / D conversion to ensure sampling quality.

[0024] c. High-speed A / D converter: Converts continuous analog signals into discrete digital signals to meet the requirements of real-time performance and high precision.

[0025] In one possible implementation, the signal is transformed using frequency domain analysis techniques based on the initial signal dataset to decompose it into different frequency components and obtain the frequency domain distribution information of the signal. The frequency domain analysis techniques include Fourier transform, short-time Fourier transform, wavelet transform, discrete cosine transform, filter banks, etc. For example, a fast Fourier transform can be used to transform the signal, converting the time-domain signal into a frequency-domain signal, thereby decomposing the different frequency components in the signal.

[0026] In one possible implementation, after obtaining the frequency domain distribution information of the signal, it is necessary to identify and extract the components directly related to the phase shift (i.e., changes in power factor or voltage / current phase difference) from numerous frequency components, determine the dynamic characteristics of the phase shift, and analyze the time-varying patterns of these extracted specific frequency components. Specifically,

[0027] Step 1: Determine the target frequency range: Phase shift is mainly caused by fluctuations in renewable energy output and load changes. These fluctuations typically occur in specific low- or mid-frequency ranges (e.g., power variation frequencies caused by wind speed or solar radiation changes). A target frequency band needs to be pre-defined to exclude meaningless extremely low frequencies (DC components) and extremely high frequencies (noise).

[0028] Step 2: Threshold Screening and Comparison: Within the target frequency range, analyze the amplitude (or power spectral density) of each frequency component. Frequency components with a high correlation to phase shift typically have amplitudes significantly higher than the background noise level. By setting a dynamic or static threshold, filter out frequency components whose amplitudes exceed that threshold.

[0029] Step 3: Correlation Analysis (Optional): By comparing with historical data or system models, confirm whether these selected frequency components truly validate known power output fluctuations or load variation patterns. For example, if a certain load periodic variation is known to be 0.1Hz, then the high-amplitude frequency components near 0.1Hz are the target components.

[0030] In one possible implementation, the dynamism of these features is determined by analyzing (quantifying) whether they exceed a preset threshold range. Specifically,

[0031] Step A (Threshold Determination): Calculation / Judgment: The extracted dynamic change features (such as amplitude and frequency drift) are compared with the system's preset dynamic values.

[0032] Quantification: If the dynamic characteristics of phase shift (such as harmonic content or amplitude at a specific frequency) exceed the preset threshold range, it is considered to be drastic and requires further detailed analysis.

[0033] Step B (Time Domain Restoration): Calculation: If the threshold is exceeded, the system performs time-domain restoration processing on the frequency component. This is typically achieved through inverse Fourier transform (IFFT) or other time-frequency analysis methods, converting the frequency domain information back into detailed time-series data showing phase changes.

[0034] Step C (Segment Recognition):

[0035] Calculation: Analyze this detailed time series data by calculating the rate of change (such as the first derivative). The degree of change can be quantified by the second derivative or the first derivative.

[0036] Quantification: By segmenting and comparing the rate of change with thresholds, key time points of dramatic change are identified, and abnormal intervals of phase change are determined. This is a direct quantification of the dynamics of phase change.

[0037] Step D (Cause Classification): Calculation: Based on the data characteristics of the abnormal intervals and the correspondence between power output fluctuations and load changes, machine learning methods such as the support vector machine (SVM) algorithm are used to classify the causes of the abnormal intervals.

[0038] Quantization: The SVM model quantitatively determines whether the abnormal phase originates from a specific source such as "sudden change in wind power output" or "large load switching" by calculating the hyperplane distance between feature vectors.

[0039] The determination of the dynamic change characteristics is first made by threshold comparison in the frequency domain, then the frequency domain characteristics are converted into a time series by time domain restoration, and finally quantitative and qualitative analysis is performed by rate of change calculation and support vector machine classification.

[0040] Furthermore, the preset threshold range is not a fixed value, but a dynamic or empirical critical value. Its purpose is to define whether the dynamic change in phase shift has reached a level requiring urgent intervention. The dynamic change characteristic refers to the amplitude and / or frequency drift of specific frequency components extracted from frequency domain analysis. The threshold can be quantified in the following form:

[0041] 1. Amplitude threshold for specific frequency components: Set the amplitude of specific frequency components related to phase shift (such as power fluctuation at 0.1 Hz) to not exceed Z% of the system reference amplitude. If If the value exceeds the threshold range, it is determined that the value exceeds the threshold range.

[0042] 2. Phase change rate (derivative) threshold: Although the feature is in the frequency domain, its corresponding time domain change rate ( Phase angle is a core indicator for measuring dynamism. It can be set that the rate of change of phase angle should not exceed X degrees / second. If... .

[0043] 3. Harmonic Distortion Rate Threshold: If the dynamic characteristics of phase shift are related to a sudden increase in the system's harmonic content, the total harmonic distortion (THD) can be set to not exceed Y%. If THD > Y%, it is determined to be outside the threshold range.

[0044] Conclusion: The preset threshold range is a set of engineering parameters determined by the microgrid designer based on system stability requirements, equipment performance limitations, and historical operating data. Its purpose is to immediately trigger a more detailed and real-time time-domain reconstruction analysis when the frequency domain characteristics indicate severe or potentially unstable phase fluctuations.

[0045] In one possible implementation, when classifying the causes of abnormal intervals and determining the specific source of abnormal phases using a Support Vector Machine (SVM) algorithm based on the data characteristics of abnormal intervals (e.g., fluctuation amplitude, frequency, and temporal relationship with load changes), combined with the correspondence between output fluctuations and load changes, the correspondence between output fluctuations and load changes serves as a core input feature to distinguish different types of system disturbances. This correspondence refers to the dynamic change pattern between renewable energy output (e.g., wind power, photovoltaic) and microgrid load (user demand) within the same time period when phase anomalies occur. Renewable energy output fluctuations include the amplitude, rate of change (slope), and direction of change (increasing or decreasing). Load changes include the amplitude, rate of change, and direction of change (connection or disconnection). Correspondence refers to whether output and load changes are in the same or opposite direction, and their time difference (delay) and proportional relationship.

[0046] The specific implementation method for classifying the causes of anomalous intervals and determining the specific sources of anomalous phases using the Support Vector Machine (SVM) algorithm is as follows: The SVM algorithm classifies anomalous phases into predefined causal categories by identifying patterns in these input feature vectors (i.e., correspondences) in a multidimensional space. Specifically, Preset cause classification (input) 1: The output of new energy sources has increased dramatically; the corresponding "output and load" relationship characteristics (input) are: the output of new energy sources has increased rapidly and significantly; at the same time, the load remains unchanged or decreases slightly.

[0047] Preset cause classification (input) 2: Sudden cut-off of large load; corresponding "output and load" relationship characteristics (input): rapid and significant load drop; at the same time, the output of new energy remains unchanged or fluctuates slightly.

[0048] Preset cause classification (input) 3: Sudden input of large load; corresponding "output and load" relationship characteristics (input): load rises rapidly and significantly; at the same time, the output of new energy remains unchanged or fluctuates slightly.

[0049] Preset cause classification (input) 4: Output / load coordinated change; corresponding "output and load" relationship characteristics (input): new energy output and load change rapidly at the same time and with the same magnitude (e.g., both decrease rapidly in a short period of time).

[0050] Real-time logic of SVM: 1. Feature extraction: The output change rate and load change rate within the abnormal interval are used as the feature vector x of SVM.

[0051] 2. Classification: SVM separates different causal categories by finding an optimal hyperplane. For example, if vector x falls in the region corresponding to "output surge", the specific source of the anomalous phase is determined to be "output surge".

[0052] 3. Determine the source: Finally, for the specific source of the abnormal phase (e.g., determined to be "sudden shedding of heavy load"), generate corresponding data tags and store them so that they can be directly called upon in subsequent analysis to obtain reference for anomaly handling.

[0053] Step S102: Based on the dynamic monitoring characteristics of phase shift and combined with historical data time series, construct a periodic analysis model of phase change fluctuation, divide the prediction time window, identify phase change abrupt points, and obtain the prediction result of the future trend of phase change.

[0054] Dynamic monitoring data of phase shift is obtained from the storage system, along with corresponding historical data and time series information. Through data cleaning, outliers and missing values ​​are removed to obtain a standardized phase monitoring dataset. For this standardized dataset, a time series decomposition method is used to separate periodic and non-periodic components, determining the periodic fluctuation pattern in phase changes. Based on this pattern, a phase change model is constructed, and a Long Short-Term Memory (LSTM) network algorithm is used to simulate the future trend of phase changes, obtaining preliminary trend prediction results. Based on these preliminary predictions, multiple prediction time windows are defined, and the phase change amplitude within each window is analyzed. If the amplitude exceeds a preset threshold, a potential abrupt change is identified within that window. For these potential abrupt change locations, the abrupt change points are confirmed by combining them with historical phase change records. If the abrupt change point matches the historical pattern, the accuracy of the abrupt change location is confirmed. Based on the identified abrupt change locations and the preliminary trend prediction results, the prediction direction of the future trend is adjusted, generating the final phase change trend prediction data. Using the final phase change trend prediction data, combined with the dynamic monitoring results, a future trend report of phase changes is output for subsequent analysis.

[0055] In one possible implementation, for the standardized phase monitoring dataset, a time series decomposition method is used to separate the periodic fluctuation components and non-periodic components, and to determine the periodic fluctuation pattern in the phase change. The time series decomposition method can be classical time series decomposition and STL decomposition, as follows:

[0056] 1. Classical time series decomposition decomposes time series... It can be broken down into three main components: trend term Periodic terms (Seasonal / Periodic) and Residual Term .

[0057] Specific feasible implementation methods:

[0058] Model selection: Based on the "periodic fluctuation law" of the phase change mentioned in this application, a multiplicative model or an additive model can be used.

[0059] • Additive model: (Applicable to situations where the amplitude of periodic fluctuations does not change over time).

[0060] Multiplication model: (Applicable to situations where the amplitude of periodic fluctuations varies over time, which is common in power systems).

[0061] Implementation steps:

[0062] Trend item ( Determined: The standardized phase monitoring dataset was smoothed using the moving average method to calculate the trend term.

[0063] • Periodic term ( Separation: Subtract (or divide by, depending on the model) the trend term from the original data to obtain a combination of periodicity and residuals. Then, perform a periodic average on these residuals to estimate the periodic term. .

[0064] • Non-periodic components (residuals) Determining: Subtract (or divide) the trend term and periodic term from the original data; the remaining term is the non-periodic residual. .

[0065] 2. STL decomposition is a more robust decomposition method, particularly suitable for outliers that may exist in power system data, and can handle periodic components of any form.

[0066] Specific feasible implementation methods:

[0067] Core algorithm: Using the Loess (locally weighted regression scatter smoothing) method.

[0068] Implementation steps:

[0069] • Periodic term ( Estimation: Perform Loess locally weighted regression on the detrended sequence to estimate the periodic term.

[0070] • Deperiodicization: Subtracting periodic terms from the original sequence.

[0071] Trend item ( Estimation: Apply Loess locally weighted regression again to the deperiodic series to estimate the trend term.

[0072] • Non-periodic components (residuals) Confirmed: Through The residual term is calculated using the (additive model).

[0073] The subsequent steps of this application explicitly state that after determining the periodic fluctuation pattern, a Long Short-Term Memory (LSTM) algorithm will be used to construct a phase change model and simulate future trends. Therefore, regardless of which time series decomposition method is used, the isolated periodic fluctuation pattern will serve as input or reference to guide the construction of the subsequent LSTM prediction model.

[0074] In one possible implementation, the phase change model is constructed based on the Long Short-Term Memory (LATM) network algorithm, which uses periodic fluctuation patterns to guide the LSTM model to simulate and predict future trends in phase changes.

[0075] In one possible implementation, the preliminary trend prediction results are used to divide the time into multiple prediction windows. The phase change amplitude within each window is analyzed. If the amplitude exceeds a preset amplitude threshold, it is determined that there is a potential abrupt change within that window. Common methods for dividing the prediction time windows include:

[0076] 1. Based on power grid cycle: Divide the window according to the inherent cycle of the power system (such as several seconds, tens of seconds or minutes) to capture the dynamic changes in system operation.

[0077] 2. Based on prediction accuracy: dynamically divide the window, that is, use a longer window when the prediction trend is stable, and use a shorter window when the prediction trend changes drastically.

[0078] 3. Fixed time interval: Set a fixed and uniform prediction step size as the time window (e.g., every 5 seconds as a prediction window).

[0079] The preset amplitude threshold Typically, it is a percentage or angle value representing the system's tolerance to phase deviation. Specific feasible implementation methods are as follows:

[0080] 1. Implementation method based on historical data statistics

[0081] This method uses historical data from microgrid systems under normal and stable operating conditions to determine a reasonable benchmark.

[0082] • Data Acquisition and Processing: Collect the phase change amplitude of the microgrid system under fault-free and voltage-stable conditions. Time series data.

[0083] • Statistical analysis: Calculating the statistical characteristics of phase variation amplitude in historical data. Commonly used methods include:

[0084] • Standard Deviation (σ): Calculates the mean μ and standard deviation σ of historical data.

[0085] • Threshold determination: Set the threshold to the historical average change plus a safety factor (e.g., k times the standard deviation).

[0086]

[0087] Typically, the safety factor k can be between 2 and 3. This means that changes exceeding this threshold have an extremely low probability of occurring under normal circumstances (approximately 5% or 0.3%), and are therefore considered potential mutations or anomalies.

[0088] 2. Implementation methods based on power grid standards and equipment limitations

[0089] This method directly uses engineering specifications and equipment performance as constraints for setting thresholds.

[0090] • Grid connection specifications: Refer to national or industry standards (such as power quality standards) to determine the maximum allowable voltage deviation at the microgrid connection point. .

[0091] • Calculate correlation: Establish phase changes through power grid models (e.g., through power flow calculations or sensitivity analysis). With voltage deviation The mathematical relationship between them.

[0092]

[0093] Where P and Q represent the active power and reactive power of the current system.

[0094] • Reactive power compensation device limitations: Refer to the maximum response speed and compensation capacity of the reactive power compensation devices (such as STATCOM, SVC) equipped in the system. The threshold should be set at the boundary of the range that these devices can correct in a timely and effective manner. If the phase change amplitude exceeds this limit, the compensation device may lag or fail, thus being identified as a potential abrupt change.

[0095] If the threshold is exceeded, subsequent correction logic will be triggered:

[0096] like

[0097] The threshold is set to ensure that when the phase change amplitude exceeds the normal fluctuation range of the system, the adjustment and correction of the predicted trend can be initiated in a timely manner, thereby improving the accuracy of subsequent reactive power compensation allocation (S104).

[0098] In one possible implementation, when the trend predicted by the LSTM prediction model is confirmed to change suddenly and significantly at a specific location (the abrupt change point), the model's output after the abrupt change point needs to be intervened and corrected to better fit the actual physical constraints. Specific adjustments to the prediction direction of the future trend based on the determined abrupt change location and preliminary trend prediction results to generate the final phase change trend prediction data may include:

[0099] 1. Data weight adjustment before and after the mutation point:

[0100] Maintain preliminary trend predictions until a clear breakout point is identified.

[0101] After a mutation occurs, the weight of real-time monitoring data (or confirmed mutation history patterns) near the mutation point is temporarily increased in subsequent predictions, forcing the LSTM model to "turn" or "converge" in a new direction in the short term.

[0102] 2. Correction of trend extrapolation logic (related to S103):

[0103] If the initial prediction continues to extrapolate along the old trend direction after the mutation point, the adjustment logic will force the introduction of a slope change (i.e., a new rate of change) at the mutation point.

[0104] This new slope may be based on historical patterns of abrupt changes (e.g., a certain load jump always causes a phase change within the next 50ms). horn).

[0105] 3. Resetting the prediction window:

[0106] Once the mutation point is identified, the prediction results before that point should be truncated immediately.

[0107] The LSTM model is reinitialized or retrained using the actual observation at the mutation point as the new initial input state, and a new short-term prediction is made starting from that point.

[0108] For example, suppose the system identifies a potential mutation site at time t.

[0109] time :

[0110] Preliminary trend prediction (LSTM output): Phase rises slowly;

[0111] Actual observations / mutation point confirmation: None;

[0112] Adjusted final forecast trend: phase rises slowly.

[0113] time :

[0114] Preliminary trend prediction (LSTM output): The phase continues to rise slowly (breakdown point);

[0115] Actual observation / mutation point confirmation: Mutation confirmed: The phase suddenly and sharply dropped by 10° was actually observed;

[0116] Adjusted final forecast trend: The phase continues to rise slowly.

[0117] time :

[0118] Preliminary trend forecast (LSTM output): Slow upward trend continues to be extrapolated;

[0119] Actual observations / mutation point confirmation: None;

[0120] Adjusted final forecast trend: Forced adjustment: from Initially, the trend is forced to decline rapidly, and then slowly recover or remain stable based on the new stable state (such as the trend description revised in S103).

[0121] In this example, "adjusting the prediction direction of future trends" means changing the LSTM prediction from... The subsequent slow upward trend was forcibly corrected to a trend of rapid decline followed by stabilization, in order to reflect the true impact of the confirmed abrupt change events on the system phase.

[0122] Step S103: If the phase abrupt change point exceeds the preset fluctuation range, the trend extrapolation logic is adjusted by data smoothing preprocessing technology, and combined with prediction error correction, a corrected phase change trend description is obtained.

[0123] By collecting raw phase change data, initial predictive trend data is obtained. A pre-established analytical model is used to obtain a preliminary distribution of abrupt change points. If the initially obtained abrupt change points exceed a preset fluctuation range, data smoothing techniques, such as exponential smoothing, are applied to the excess phase change data to determine the smoothed data sequence. Based on the smoothed data sequence, combined with a trend extrapolation logic adjustment method, an adjusted trend prediction path is obtained. It is then determined whether the adjusted trend prediction path still contains points exceeding the fluctuation range. If the adjusted trend prediction path still contains points exceeding the fluctuation range, a secondary processing method is used to correct the prediction path, resulting in a corrected phase change trend. The stability of the corrected phase change trend is evaluated using preset standards to determine if the final correction result meets the expected range. Analysis of the final correction result generates detailed phase change descriptions, outputting a complete trend change record. Based on the complete trend change record and the combined processing methods, the parameters of the subsequent prediction model are optimized to obtain a more accurate prediction basis.

[0124] In one possible implementation, initial predicted trend data is obtained from the future trend prediction results of phase change by calling the corresponding data interface. The prediction module in step S102 passes the results to the correction module in S103. The raw data collected in S103 is used for verification, while the obtained initial predicted trend data is used for comparison and analysis of abrupt change points.

[0125] In practical microgrid projects, the preset fluctuation range is usually an angle value (e.g., ±5°) or a percentage value (e.g., ±2%), and its determination needs to be based on the following engineering factors:

[0126] 1. Stability constraints of the power grid: Refer to the maximum allowable power factor or voltage phase angle deviation standard of the microgrid system at the grid connection point.

[0127] 2. Historical data statistics: Based on the historical statistical results of the phase change amplitude during the stable operation of the system (e.g., 2 or 3 times the standard deviation).

[0128] 3. Equipment tolerance: The maximum phase deviation that the compensation equipment (such as a reactive power compensator) can quickly and effectively respond to and correct.

[0129] The logical adjustment method for trend extrapolation is based on the physical constraint slope limitation method. At the end of the smoothed sequence, its tangent slope (i.e., the rate of phase change) is calculated. During extrapolation, a physical constraint is applied to this slope.

[0130] Method: Set a maximum allowable slope change. This value is determined based on the ramp rate of renewable energy generation in the microgrid, the load change rate, or the maximum response speed of reactive power compensation equipment.

[0131] Objective: To ensure that the adjusted prediction path converges to a more stable slope in the short term, rather than continuing to extrapolate along extreme slopes that could lead to abrupt changes.

[0132]

[0133] This logical adjustment is the key to correcting the predicted trend in step S103. It transforms the smoothed "clean" data into a trend prediction path that conforms to the system's physical constraints and has convergence characteristics.

[0134] In one possible implementation, the stability of the corrected phase change trend is evaluated using a preset standard to determine whether the final correction result meets the expected range. The preset standard is a criterion used to measure whether the corrected phase change trend is sufficiently stable, controllable, and reliable. Feasible engineering standards may include:

[0135] 1. Preset standard for fluctuation amplitude: Maximum permissible deviation rate: The deviation between the phase value of the corrected trend curve and the reference phase value for stable system operation at any prediction point in time shall not exceed a very small percentage. The goal is to ensure that the corrected trend is convergent and stable in magnitude.

[0136] 2. Preset standard for rate of change (slope): Maximum instantaneous rate of change limit: The slope (first derivative) of the corrected trend curve at any point in time must not exceed the maximum rate of change that the system can physically withstand. The goal is to ensure that the corrected trend change is not too drastic, and that the compensation equipment can keep up in time.

[0137] 3. Convergence Characteristics: Convergence Time Constant: The corrected trend curve must converge within a predetermined time (e.g., within a certain time frame). The system converges to a steady state or the target phase value within a specified error range (e.g., ±0.1°) within milliseconds. The aim is to ensure that the system can quickly recover to stability after a sudden change.

[0138] The expected range refers to the corrected phase change trend, which must remain within a safe and acceptable range throughout the entire prediction time window.

[0139] Based on the safe operating range, the expected range is the system stability margin range: the boundary of this range is usually determined by the stability margin in the system design, that is, the phase angle limit that ensures the microgrid will not disconnect from the grid or cause a cascading reaction when operating in grid-connected mode. arrive The goal is to ensure that the corrected result is physically safe.

[0140] Based on the effective control range, the expected range is the effective range of reactive power compensation: this range is directly related to the control capability of reactive power compensation equipment in the microgrid. The corrected trend must lie within the range where the compensation equipment can effectively provide compensation and regulation. The aim is to ensure that the system has sufficient control capability to cope with this corrected trend.

[0141] In summary, this step ensures that, after complex prediction corrections (data smoothing, trend extrapolation, error correction), the final prediction is not only mathematically smooth but also engineering-stable and safe, and will not lead to system voltage instability.

[0142] In one possible implementation, the step of optimizing the parameters of the subsequent prediction model based on complete trend change records and a combined processing method to obtain more accurate prediction basis refers to the combined application of a series of correction and adjustment methods in step S103, including: data smoothing technology for processing phase change data exceeding the fluctuation range; a trend extrapolation logic adjustment method for obtaining the adjusted trend prediction path; a prediction error correction method for secondary processing of the prediction path; and stability evaluation, using preset standards to evaluate trend stability. The parameters of the prediction model refer to the internal configuration and learning parameters of the LSTM model. Specific optimizable parameters include: model structure parameters: the number of hidden layer units, which determines the model's complexity; learning process parameters: the learning rate, which determines the speed at which the model updates weights in each iteration; training data parameters: the window size for inputting historical data, which determines how much historical data the model uses to predict the future; and the number of iterations and batch size, which affect the model's training efficiency and generalization ability.

[0143] The optimization of parameters for the subsequent prediction model can be achieved through the following feasible parameter optimization methods:

[0144] 1. Error Feedback Adjustment: If S103 finds a persistent and systematic deviation between the initial prediction result (LSTM output) and the corrected result, it adjusts the learning rate or weights of the LSTM to make it tend to output a more conservative trend that better fits the physical constraints.

[0145] 2. Hyperparameter search: Using methods such as grid search or Bayesian optimization, within the possible values ​​of model structure and learning parameters (such as the number of hidden layer units and learning rate), the model is retrained and tested to select the parameter combination that minimizes the correction error.

[0146] 3. Online learning / transfer learning: The complete trend change records generated by S103 are used as new, corrected training samples to incrementally train or fine-tune the LSTM model, enabling the model to adapt to the latest system operating state and mutation patterns more quickly.

[0147] The purpose of the optimization is to adjust the parameters of the LSTM model based on the complete trend change record obtained in step S103 (i.e., the corrected trend that is closer to reality), so that it is more accurate in future predictions and reduces the frequency of needing to perform the complex correction process of S103.

[0148] Step S104: Based on the corrected phase change trend description, integrate the real-time feedback data of voltage deviation, generate a segmented adjustment scheme for the reactive power compensation command sequence, and determine the preliminary dynamic allocation strategy for compensation capacity.

[0149] By collecting phase change data during system operation, trend description information of phase change is obtained to determine its initial impact on system stability. Based on the trend description information of phase change, combined with real-time feedback voltage deviation data, a data combination method is used to determine the correlation between voltage deviation and phase change. If the correlation between voltage deviation and phase change exceeds a preset correlation threshold, a reactive power compensation command sequence is triggered, and the priority order of the command sequence is determined. Through the priority order of the command sequence, specific parameters for segmented adjustment are obtained, and adjustment schemes for different time periods are obtained using a segmented adjustment method. Based on the segmented adjustment schemes, the allocation requirements for compensation capacity are obtained, and a preliminary allocation strategy for compensation capacity is determined using dynamic allocation logic. For the preliminary allocation strategy of compensation capacity, the allocation strategy is optimized using a support vector machine algorithm based on real-time feedback data updates, resulting in the final reactive power compensation allocation result. Based on the final reactive power compensation allocation result, system operation stability data is obtained to determine the evaluation of the adjustment effect of reactive power compensation on system operation.

[0150] In one possible implementation, after collecting phase change data and obtaining trend descriptions of phase changes during system operation, the next step is: "Based on the trend descriptions of phase changes, a Support Vector Machine (SVM) algorithm is used to classify the causes of abnormal phases and obtain a preliminary impact assessment." The SVM algorithm acts as a classifier here. It categorizes the current system operating state or phase change trend into different classes based on the input phase change trend descriptions (feature vectors). These classes constitute the preliminary impact assessment on system stability. For example, the SVM model might classify phase changes as:

[0151] Category 1: Stable (no effect or negligible effect);

[0152] Category 2: Slightly unstable (requires minor compensation);

[0153] Category 3: Severe instability (requiring substantial compensation or rapid response);

[0154] Category 4: Instability caused by sudden changes (e.g., sharp fluctuations in the output of new energy power generation, sudden changes in load, etc.).

[0155] The specific feasible implementation methods for determining the initial impact on system stability are as follows:

[0156] 1. Feature vector construction:

[0157] The collected trend description information of phase change (e.g., maximum change amplitude, instantaneous change rate, change duration, etc.) is used to construct a feature vector X.

[0158] 2. SVM model training (pre-completed):

[0159] The SVM model is trained using historical operating data, including normal phase changes and phase changes caused by various faults / mutations.

[0160] Each training sample has a label, which is the category of its actual impact on system stability (e.g., load mutation, failure, normal fluctuation).

[0161] 3. Real-time classification and judgment (preliminary impact determination):

[0162] The current feature vector X is input into the trained SVM model. The SVM model outputs a classification result, which is the most likely cause of the phase change and its corresponding preliminary impact judgment.

[0163] Therefore, the core mechanism for determining the initial impact is to use the SVM algorithm to classify and map the real-time phase change trend characteristics to a preset cause category related to stability.

[0164] In one possible implementation, if the correlation judgment result between voltage deviation and phase change exceeds a preset correlation threshold, a reactive power compensation command sequence is triggered. In determining the priority order of the command sequence, the correlation refers to the degree of matching between the phase change trend and the system stability impact category (preliminary impact judgment). The correlation threshold... This is an indicator used to measure the reliability or importance of SVM classification results. The method for setting the preset correlation threshold is as follows:

[0165] 1. SVM-based confidence score: The threshold can be set as the confidence score of the SVM model for a certain classification result. For example:

[0166]

[0167] (When the SVM algorithm determines that the current phase change has a greater than 80% probability of belonging to the "severely unstable" category, a compensation command is triggered.)

[0168] 2. Based on urgency: The threshold can be based on the urgency level classified by SVM.

[0169] The correlation is considered to have "exceeded the threshold" only when the SVM classification result belongs to a category requiring compensation, such as "slightly unstable" or "severely unstable". If it is in the "normal fluctuation" category, no instruction generation is triggered.

[0170] The purpose of generating the reactive power compensation instruction sequence is to transform the initial assessment of the stability impact into a series of specific compensation actions that can be executed by the equipment. A specific feasible implementation method for generating the instruction sequence is as follows:

[0171] 1. Rule-based:

[0172] Based on the initial impact assessment obtained from SVM (e.g., "load mutation leads to severe system instability"), the system searches for the corresponding standard compensation scheme from the pre-established control instruction library.

[0173] Instruction content: The compensation scheme is broken down into a series of instruction sequences, which explicitly specify: the target device (such as STATCOM, SVC), the compensation capacity (such as how many Mvars to allocate), and the action time (such as within 5ms).

[0174] 2. Based on dynamic strategy:

[0175] The initial impact judgment is input into the corrected phase change trend description in step S102, combined with the real-time voltage deviation feedback data of the system.

[0176] Referring to the final goal described in S104: "Determine the initial dynamic allocation strategy for compensation capacity", calculate the required compensation capacity based on the dynamic allocation strategy, and then generate the corresponding equipment action instructions.

[0177] The prioritization is to ensure that the most urgent and impactful compensation actions on the system are executed first. Feasible prioritization criteria include:

[0178] 1. The urgency of the system impact:

[0179] Highest priority: Indicators classified as "severely unstable" or involving "risk of disconnection" based on stability and SVM classification.

[0180] Second highest priority: Instructions whose stability is judged as "slightly unstable".

[0181] 2. Equipment response speed:

[0182] Prioritize scheduling reactive power compensation devices with the fastest response speed (e.g., STATCOM is typically faster than SVC, and SVC is faster than mechanically switched capacitors).

[0183] 3. Forecast time window:

[0184] Priority is negatively correlated with the time distance of the predicted mutation point. The closer the predicted mutation point (e.g., within the next 50 milliseconds), the higher the priority of the corresponding compensation instruction.

[0185] 4. The importance of compensation capacity:

[0186] Prioritize scheduling instructions involving critical nodes and maximum compensation capacity to ensure the fastest possible suppression of phase changes.

[0187] In one possible implementation, the specific parameters for segmented adjustment are obtained by prioritizing the instruction sequence. Segmented adjustment is then used to obtain adjustment schemes for different time periods. Based on these schemes, the allocation requirements for compensation capacity are determined. Finally, a preliminary allocation strategy for compensation capacity is determined using dynamic allocation logic. In this segmented adjustment, the entire prediction time window (or instruction execution cycle) is divided into several sub-intervals, where the reactive power compensation capacity, equipment, or target differs within each sub-interval. This segmented strategy is key to achieving "dynamic allocation." The specific parameters for segmented adjustment constitute a complete control scheme for time, capacity, and equipment. It transforms the phase trend (when it changes) and voltage deviation (how much compensation is needed) corrected by S103 into precisely executable, phased reactive power compensation instructions, specifically:

[0188] Time segmentation parameters: Segment start point and the end point Defines the execution time interval of the instruction sequence. For example, in the sequence... arrive Execute the first instruction between them. arrive The second instruction is executed between these two points.

[0189] Compensation capacity parameters: allocated capacity , which defines the reactive power compensation capacity that the system needs to add or remove within the i-th time segment (e.g., adding 1.5 Mvar). This is the core result of dynamic allocation.

[0190] Equipment selection parameters: Target equipment number / type The term is defined as which reactive power compensation device (such as STATCOM-1, SVC-2, or capacitor bank) is responsible for providing compensation within the i-th time segment.

[0191] Control target parameter: target phase angle / Target voltage Defined as the target value that the control system attempts to stabilize the phase angle or voltage to within the i-th time segment.

[0192] Adjust the rate parameter: slope / rate of change Defined within the first segment, the rate at which compensation capacity is introduced or the rate at which the phase angle recovers to the target, to match the system's response requirements to sudden changes.

[0193] The segmented adjustment scheme is the segmented adjustment scheme of the reactive power compensation instruction sequence generated in the previous step. As mentioned earlier, this scheme consists of a set of parameters to guide the compensation actions. These parameters define the compensation actions the system should take in different time segments. The scheme content (i.e., the parameter set) mainly includes:

[0194] Time Segmentation: The entire compensation period is divided into segments. arrive , arrive Equal time intervals.

[0195] Target state: The target phase angle or target voltage deviation that the system needs to achieve within each time period.

[0196] Compensation requirements: Within each segment, a preliminary estimate of the total reactive power compensation required is made (this is the initial value of the "allocation requirement of compensation capacity").

[0197] Equipment constraints: The current status and maximum capacity of available reactive power compensation equipment.

[0198] The allocation requirement for compensation capacity refers to the total reactive power (Q) required to pull the system back from its current unstable state (determined by the corrected phase trend and voltage deviation) to the target stable state within each time segment. This allocation requirement is obtained through calculations using a power grid model. This process involves adjusting the target state (Q) in the "segmented adjustment scheme content" to reflect the desired state. , ) and current state ( , The results are compared and solved in conjunction with system parameters.

[0199]

[0200] Specific implementation methods include:

[0201] 1. Sensitivity Matrix Method: This method uses the Jacobian matrix or sensitivity matrix of the microgrid to calculate the voltage deviation. and phase deviation change Required reactive power.

[0202] 2. Target Deviation Method: Based on the corrected phase change trend (S103 result) and voltage deviation (S104 input), the total amount of reactive power missing in each segment of the system is directly calculated.

[0203] The dynamic allocation logic refers to optimizing the allocation of total compensation demand among multiple compensation devices based on the real-time status of the system (voltage deviation, phase deviation, degree of abrupt change) and the real-time capabilities of the equipment. The process. The dynamic allocation logic is based on priority sorting, and its relationship with priority sorting is as follows:

[0204] 1.S104 The previous step (sub-step two) has generated an instruction sequence based on SVM classification and correlation judgment, and determined the priority order of the instruction sequence (for example, the fastest STATCOM compensation instruction has the highest priority).

[0205] 2. The dynamic allocation logic utilizes this priority sorting, combined with the current status of the devices (whether they are saturated, whether they are faulty, and their remaining capacity), to allocate the total... Assign it to devices that have high priority and the ability to execute it.

[0206] In short, priority ordering (instruction sequence) tells the system "which device should act first," while dynamic allocation logic tells the system "how much capacity each device should allocate." "to meet the overall allocation requirements" And ensure overall optimization. For example, in dynamic allocation, although STATCOM-1 has the highest priority, if it has reached its capacity limit, the dynamic allocation logic will allocate the remaining compensation requirements to the device with the second highest priority that still has capacity. Step S105, if the strategy finds potential inconsistencies in instruction execution conflict detection, the instruction priority sorting is adjusted through control latency optimization processing technology, and combined with device response time evaluation data, an optimized control instruction sequence is obtained.

[0207] A conflict detection mechanism is used to identify potentially inconsistent flags during instruction execution, thus determining a set of abnormal instructions. Based on this set, a control latency analysis tool is employed to determine the latency impact range of each instruction, resulting in a latency priority list. For this priority list, device response data is used to assess the real-time requirements of instruction execution. If the response time of an instruction exceeds a preset response time threshold, its priority is increased, resulting in an adjusted ranking. From this adjusted ranking, a subset of high-priority instructions is extracted, and combined with time evaluation data, the execution interval between instructions is determined, generating a preliminary control instruction sequence. This preliminary sequence is then optimized to determine if execution conflicts exist. If conflicts are detected, the timing of the relevant instructions is fine-tuned to obtain a conflict-free instruction sequence. Based on this conflict-free sequence and response data analysis, feedback information on the device's execution status is obtained, determining the final control instruction sequence. Finally, using this final control instruction sequence, an automated distribution tool pushes the instructions sequentially to the target device, completing the execution adjustment process.

[0208] In one possible implementation, based on the set of abnormal instructions, a control delay analysis tool is used to obtain the delay impact range of each instruction and derive a delay priority list. The control delay analysis tool is a simulation or calculation module specifically designed to evaluate the time required for control instructions throughout the entire execution path and the impact of this time variation on the system. It needs to consider the following key delay aspects:

[0209] 1. Calculation latency: The time required from the generation of an instruction to the instruction being ready to be sent.

[0210] 2. Communication delay: The time required for instructions to be transmitted to the target device through a communication network (such as fiber optic or Ethernet).

[0211] 3. Equipment response delay: The time required from the time the equipment receives the instruction to the time it actually begins to execute it (e.g., switching action, IGBT triggering).

[0212] The delay impact range refers to the deviation between the actual execution time of the instruction and the predetermined execution time in the instruction sequence, as well as the additional fluctuations in the system phase angle or voltage caused by this deviation. The specific implementation method for obtaining the delay impact range of each instruction is as follows:

[0213] a. Acquisition based on simulation models (main method)

[0214] 1. Construct a latency model:

[0215] • Establish detailed time delay models for each reactive power compensation device (such as STATCOM, SVC) and its communication path.

[0216] The latency model includes: network topology, timing characteristics of communication protocols, and minimum and maximum response times of the devices (determined according to the device specification manual).

[0217] 2. Monte Carlo / Worst-Case Analysis:

[0218] Input the instruction sequence into the simulation model.

[0219] Worst-case analysis: Calculate the instruction execution time under maximum latency conditions (e.g., communication congestion, slow device response). .

[0220] • Scope of impact of delay :calculate Compared to ideal execution time The time difference.

[0221] • Scope of system impact: The modified trend description obtained from S103 is input into the input, and the result is calculated by extrapolation. The resulting additional phase deviation amplitude .

[0222]

[0223] b. Statistical analysis based on historical data

[0224] 1. Data Acquisition: Collect and record historical latency data for each instruction from its transmission to the actual action of the device in real time.

[0225] 2. Statistical Analysis: Perform statistical analysis on historical data to calculate the average latency of each instruction. and standard deviation of delay .

[0226] 3. Define the scope: Set the scope of the delay's impact as a statistical interval, for example... If the delay of the instruction exceeds this range, it is considered high risk.

[0227] Summarize:

[0228] The range of delay impact is a dual indicator: the time deviation range (e.g., 5–15 milliseconds) and the resulting system instability magnitude (e.g., ±0.5° phase deviation). Analysis tools use simulation or historical data to calculate this range, thus providing a basis for subsequent delay priority lists.

[0229] In one possible implementation, the delay priority list is used to determine the real-time requirements of instruction execution based on device response data. If the response time of an instruction exceeds a preset response time threshold, its priority is increased, resulting in an adjusted ranking. The real-time requirement for instruction execution refers to the maximum time limit allowed by the system from the issuance of a compensation instruction to its actual completion. This depends on the current instability of the system and the need for rapid compensation. The compensation instruction must be executed within a specified time (e.g., ...). The action must be completed within milliseconds to ensure effective suppression of phase abrupt changes or voltage deviations. Furthermore, higher real-time requirements indicate a more urgent instability situation corresponding to the instruction, and a greater contribution to system stability.

[0230] The real-time requirement for determining instruction execution needs to combine the importance of the instruction itself with the current instability state of the system. A feasible determination method is as follows:

[0231] 1. SVM-based classification results:

[0232] • High requirements: If the SVM classification in step S104 is "severely unstable" or "high-risk mutation", the corresponding compensation instruction must be set with extremely high real-time requirements (i.e., extremely short required time).

[0233] • General requirements: If the classification is "slightly unstable", the real-time requirement can be appropriately relaxed.

[0234] 2. Based on the phase change rate:

[0235] • Based on the revised trend description in S103, calculate the phase change rate corresponding to instruction execution. The faster the phase change rate, the higher the real-time requirement for instruction execution.

[0236] 3. Scope of impact based on device latency:

[0237] • If the delay effect range calculated in the previous step of S105 ( The data shows that if the delay of the instruction causes an additional deviation to the system exceeding the safety threshold (e.g., ±0.3°), then the real-time requirements of the instruction must be increased.

[0238] The preset response time threshold It is a time value, a baseline for judging whether instruction execution is "too slow." This is based on the actual response time of the device. (Using device response data) exceeded If the real-time performance of the instruction is deemed to be at risk, its priority needs to be increased. The feasible methods for setting the preset response time threshold are as follows:

[0239] 1. Based on equipment specifications:

[0240] The threshold is set to the typical or fastest response time of the target reactive power compensation device (such as STATCOM) (usually provided by the manufacturer, such as 1-5 milliseconds).

[0241] 2. Based on system control cycle:

[0242] The threshold is set to one or half a control cycle of the microgrid control system. For example, if the control system makes a decision every 10 milliseconds, the threshold may be set to 5 milliseconds or 10 milliseconds.

[0243] 3. Based on statistical analysis:

[0244] The threshold is set as a safe multiple of the historical average response time for this type of instruction, such as... .

[0245] The logic for prioritizing instructions: If the latency risk of an instruction is high (i.e. > To ensure it can complete its action before the system becomes unstable, the system must increase its priority in the queue, allowing it to execute earlier. This is a delay compensation mechanism.

[0246] In one possible implementation, the initial control instruction sequence is processed using optimization techniques to determine if execution conflicts exist. If a conflict is detected, the relevant instructions are fine-tuned in timing to obtain a conflict-free instruction sequence. The optimization techniques typically employ constraint satisfaction problem (CSP) solving or heuristic algorithms. Under the premise of satisfying all constraints (device response time, communication bandwidth), the total instruction delay or the total timing fine-tuning magnitude is minimized. Algorithm examples might use genetic algorithms, particle swarm optimization (PSO), or specialized scheduling optimization algorithms to find the optimal fine-tuning time point. An execution conflict refers to two or more instructions attempting to operate on the same resource (such as a device or network link) at the same time (or within a very short time interval), or canceling each other out, resulting in an uncertain system state or invalid instructions.

[0247] The feasible implementation methods for the conflict detection mechanism are as follows:

[0248] The conflict type is equipment resource conflict, and its detection method is as follows: check whether there are multiple instructions (such as switching instructions, capacity adjustment instructions) assigned to the same reactive power compensation device (such as STATCOM-1) in the instruction sequence, and whether their ideal execution time interval is less than the device's minimum response time. .

[0249] The conflict type is network communication conflict, and its detection method is to check whether too many instructions are requested to be transmitted in the same millisecond on the same communication bus (e.g., control LAN), causing communication bandwidth saturation or packet collision.

[0250] The conflict type is a control objective conflict, and its detection method is to check whether there are two sets of commands that, although assigned to different devices, have opposite or canceling control objectives. For example, command A requires STATCOM to supply +5 Mvar reactive power, and command B requires SVC to cut off -3 Mvar reactive power, but their overall effect is to cancel each other out and cannot effectively suppress phase deviation.

[0251] The determination of whether an execution conflict exists specifically involves:

[0252] If the instruction and A conflict is considered to exist if one of the following conditions is met:

[0253] 1.

[0254] 2.

[0255] If a conflict is detected, the relevant instructions are fine-tuned in timing to obtain a conflict-free instruction sequence. This timing fine-tuning involves making minimal modifications to the instruction execution time after a conflict is detected to eliminate the conflict and ensure the validity of the instruction sequence. The specific adjustment method is as follows:

[0256] 1. Eliminate device conflicts: Delay execution: The execution time of one or more lower-priority instructions in a conflict is slightly adjusted backward by a very short time interval. (e.g., 1-5 milliseconds). . It must be greater than the device's minimum response time. The goal is to ensure that the device has enough time to complete the previous action and be ready before executing the next instruction.

[0257] 2. Eliminate network collisions: Batch transmission: Divide the conflicting instruction sequence into multiple subsets and send them within different time windows. For example, distribute the 20 instructions at t=10ms into three time points: t=10ms, t=10.5ms, and t=11ms. The purpose is to reduce the instantaneous bandwidth pressure on the communication bus and avoid packet loss.

[0258] 3. Optimize Control Performance: Realignment: While ensuring no conflicts, fine-tune the execution time of instructions based on the delay priority list determined in step S105, bringing it closer to the ideal, delay-free execution time to maximize control performance. The aim is to minimize the impact of timing fine-tuning on the overall control strategy.

[0259] Step S106: By optimizing the control command sequence and combining it with the grid stability constraints, the microgrid system is controlled in real time to obtain the actual execution effect of the compensation action and determine whether the system voltage deviation meets the stability requirements.

[0260] Based on the business content and extracted relevant attributes, the following business solution is generated, outlining the technical process around attributes such as control commands, optimization sequences, grid stability, constraints, microgrid systems, real-time regulation, compensation actions, execution effects, voltage deviation, and stability standards: Using a pre-established control command library, load and power output data for the current period are acquired based on the microgrid system's operating status to determine the initial control command set. An optimization sequence generation method is used to select command combinations that meet grid stability requirements from the initial control command set, resulting in an optimized control command sequence. Based on the optimized control command sequence and constraints, real-time regulation of the microgrid system is implemented, acquiring compensation action data for each node. Through real-time acquisition of compensation action data, the execution effect of each node is analyzed to determine if any abnormal responses exist. If the execution effect of a node deviates from a preset threshold, the abnormal information for that node is recorded. Based on the abnormal information and voltage deviation monitoring data, the voltage fluctuation of each node is analyzed to determine the specific range of voltage deviation. Based on the specific range of voltage deviation, the stability standards are compared to determine whether the system meets the stability requirements, yielding the final regulation result. Based on the final control results, the data content in the control instruction library is updated to form a closed-loop feedback mechanism, optimizing the subsequent real-time control process.

[0261] Specifically, the pre-established control command library is a structured database or table that stores pre-designed, directly executable reactive power compensation equipment operation commands for various operating states and compensation needs in the microgrid. The aim is to achieve rapid decision-making and response. In the initial stage (S106), the system can quickly match the current load and power supply status by querying the command library to determine the initial, optimizable set of control commands. The establishment of the control command library typically involves offline simulation, system modeling, and expert knowledge; specific feasible methods for its establishment are as follows:

[0262] a. Modeling based on system topology and equipment specifications

[0263] 1. Equipment Information Entry: Record detailed technical parameters of all reactive power compensation equipment (such as STATCOM, SVC, and capacitor banks) in the microgrid.

[0264] • Maximum / minimum compensation capacity (e.g., STATCOM maximum input ±5 Mvar).

[0265] • Minimum response time (e.g., STATCOM response time 5ms).

[0266] • Switching step size (e.g., minimum switching step size for capacitor banks is 0.5 Mvar).

[0267] 2. Command / Action Definition: Define standard action command templates based on device functions, for example:

[0268] •CMD-001:STATCOM-A: Input capacity: +1.0 Mvar

[0269] •CMD-002:CAP-Bank-B:Resected volume: -0.5 Mvar

[0270] • CMD-003: SVC-C: Set target voltage: 10kV

[0271] b. Scene matching based on simulation and expert knowledge (core)

[0272] 1. Operational Scenario Classification: Identify the key operational states of the microgrid, for example:

[0273] • High output of new energy sources / low load (system voltage is too high, requiring the removal of reactive power or the activation of inductive reactive power).

[0274] • Low output / high load of new energy sources (system voltage is low, requiring the use of capacitive reactive power).

[0275] c. Specific faults or sudden changes (e.g., a disconnection of a tie line, a sudden change in a specific load).

[0276] 2. Offline simulation: For each defined operating scenario, power flow calculation and transient simulation are performed using offline simulation models of the microgrid (such as PSS / E, DIgSILENTPowerFactory, MATLAB / Simulink).

[0277] 3. Determine the optimal control: In the simulation, find the optimal combination of compensation actions required to restore the system voltage deviation and phase angle deviation to the stable range.

[0278] 4. Command and Status Association: This optimal compensation action combination is used as a command and associated with the current operating status (load data, power output), and stored in the command library.

[0279] The operating status (input) is: load: 10MW, wind power: 0.5MW; the optimal control (output / instruction) is instruction set 1: STATCOM-A: +1.5Mvar + SVC-B: +0.8Mvar.

[0280] The operating status (input) is: load: 5MW, photovoltaic: 4MW; the optimal control (output / instruction) is instruction set 2: CapBank-C: -1.0Mvar.

[0281] In this way, the set of instructions stored in the instruction library becomes the initial compensation scheme that the system can quickly retrieve and apply when encountering specific runtime conditions.

[0282] In one possible implementation, an optimized sequence generation method is employed to select command combinations that meet the grid stability requirements from the initial control command set, such as selecting command combinations that meet the system voltage stability requirements, thus obtaining an optimized control command sequence. Furthermore, the voltage stability threshold is defined as the maximum allowable voltage deviation. It is usually set according to national or industry standards and the grid connection requirements of the microgrid, such as:

[0283] 1. National Standard Deviation (Example): Refer to standards such as "Power Quality Supply Voltage Deviation", for example:

[0284] • Point of common coupling (PCC): Voltage deviation is typically required to be within ±5% of the rated voltage (e.g., for a PCC with a rated voltage of 10kV, the voltage must be maintained between 9.5kV and 10.5kV).

[0285] 2. Microgrid internal control objectives: In order to improve internal power quality, the voltage deviation requirements of internal nodes in a microgrid may be more stringent, for example, within ±2% of the rated voltage.

[0286] The optimized sequence generation method uses these specific values ​​(such as ±5%) as hard constraints to select instruction combinations that ensure all node voltages remain within the threshold.

[0287] In one possible implementation, the real-time control of the microgrid system based on the optimized control command sequence and in combination with constraints, and the acquisition of compensation action data for each node, wherein the constraints are divided into the following three types:

[0288] 1. Power grid stability constraints (hard constraints): These are core constraints that must be met, otherwise the system may become unstable or disconnect from the grid.

[0289] • Voltage stability constraint (core): This is the final evaluation metric explicitly mentioned in S106.

[0290] Constraints:

[0291] • Definition: Voltage of all critical nodes i in a microgrid The system must be kept at the minimum voltage allowed by system design or industry standards. and maximum voltage Between (typically ±5% of the rated voltage).

[0292] • Phase stability constraint:

[0293] Constraints:

[0294] Definition: Voltage phase angle difference between nodes in a system The power angle stability limit of the system must not be exceeded. This is to prevent system oscillation or loss of synchronization.

[0295] • Frequency stability constraints:

[0296] Constraints:

[0297] • Definition: The system frequency f must be kept within the specified allowable range (usually ±0.5Hz of the rated frequency).

[0298] 2. Equipment operation constraints (executability constraints): These constraints ensure that the generated optimization instructions are actually executable by the equipment in the microgrid.

[0299] • Capacity constraints:

[0300] Constraints:

[0301] • Definition: Reactive power compensation capacity that is put into operation or removed The minimum and maximum rated compensation capacity of any compensation device (such as STATCOM, SVC) must not be exceeded.

[0302] • Action rate constraint:

[0303] Constraints:

[0304] • Definition: The rate of change of compensation capacity must not exceed the maximum response speed of the equipment. (e.g., STATCOM ramp rate) to prevent equipment overload or damage.

[0305] • Switching step size constraint: For stepped compensation equipment (such as capacitor banks), the switched capacity must be an integer multiple of its minimum switching step size.

[0306] 3. Economic and operational constraints

[0307] • Network constraints:

[0308] • Branch power flow constraints: Ensure that the current and power flow of all lines in the microgrid do not exceed the thermal stability limit of the line.

[0309] • Constraints on new energy output:

[0310] • Maximum power output constraint: The actual output of the new energy generator set cannot exceed its rated maximum output.

[0311] The aforementioned combination of constraints enables real-time control of the microgrid system. This means that when the system executes the instruction sequence optimized by S106, it continuously monitors all the aforementioned constraints. If any constraint is touched or may be violated, the system will immediately adjust or correct it to ensure that the system always operates within a safe range.

[0312] In one possible implementation, the execution effect of each node is analyzed by real-time acquisition of compensation action data to determine whether there are any abnormal responses. If the execution effect of a certain node deviates from a preset threshold, the abnormal information of that node is recorded. The preset threshold is the maximum allowable deviation between the actual value of the voltage or phase angle of the target node (such as the grid connection point or critical load point) and the stable target value after the system has executed the optimized control command sequence. Specifically,

[0313] Voltage deviation threshold ( Node voltage With rated voltage The maximum percentage deviation allowed between these values. A feasible value is ±2% to ±5% of the rated voltage. This value is typically more stringent than the maximum deviation specified by national grid standards to ensure high power quality within the microgrid.

[0314] Phase angle deviation threshold ( ): Nodal phase angle Angle with the target The maximum permissible angle difference between the two. Possible values ​​are ±0.1° to ±0.5°. This is used to ensure system power angle stability and power transmission stability.

[0315] Since S106 explicitly uses voltage deviation as the final stability requirement for judgment, the threshold here is the tolerance for voltage deviation.

[0316] The logic for determining if the execution effect of a certain node deviates from the preset threshold is as follows:

[0317] like or

[0318] If the execution effect deviates from the threshold, the anomaly information for that node is recorded. This anomaly information is crucial feedback for the closed-loop mechanism and adaptive control.

[0319] Feedback Purpose: The recorded information (e.g., which instruction, at which node, and by how much deviation) will be used to iteratively update the "instruction sequence segmentation adjustment scheme" generated in S104.

[0320] Adaptive Adjustment: The system utilizes this anomaly information to adjust model parameters in the next control cycle. For example, it may increase the compensation capacity allocation weight of the node or adjust the timing of the command sequence to ensure that the next execution effect meets stability requirements. Step S107: Based on the judgment result, the execution effect is fed back to the microgrid control layer, and a closed-loop mechanism is used to iteratively update the segmented adjustment scheme of the command sequence to determine the final reactive power compensation control scheme.

[0321] By acquiring real-time system voltage data from monitoring equipment and analyzing its fluctuations, preliminary deviation assessment results are obtained. Based on these results, a preset deviation threshold is used for comparison. If the deviation exceeds this threshold, an analysis of the control effect is triggered to determine the voltage stability parameters that need adjustment. For these parameters, operating status data from the microgrid layer is acquired, and a corresponding control command sequence is generated through a closed-loop mechanism. The priority distribution of the command sequence is then determined. Based on this priority distribution, the sequence is segmented, decomposing the commands into multiple execution units to obtain segmented adjustment schemes. For these segmented adjustment schemes, an iterative optimization method is used, updating the scheme details through multiple simulation calculations to determine the final reactive power compensation strategy. Key control parameters are extracted from the reactive power compensation strategy and fed back to the microgrid layer for real-time execution. Voltage data after execution is acquired to assess the system voltage stability. Through continuous monitoring of the system voltage stability, the deviation assessment results are updated, forming a dynamic control process under a closed-loop mechanism, resulting in a long-term voltage optimization scheme.

[0322] In one possible implementation, real-time system voltage data is collected from a monitoring device, and the difference between the real-time data and a preset rated voltage value, such as 35kV, is calculated. The preliminary deviation assessment result is the absolute value of the difference.

[0323] In one possible implementation, the step of comparing the deviation assessment result with a preset deviation threshold, and triggering an analysis process for the control effect if the deviation exceeds the threshold range, determines the voltage stability parameters that need adjustment. The preset deviation threshold range is a criterion used to measure whether the system voltage deviation meets the stability requirements after the compensation command is executed in step S106. It is the same indicator as the voltage stability requirement determined at the end of S106, representing a maximum allowable voltage deviation value. This threshold range is typically set to the high power quality standards required for microgrid operation, such as ±2% to ±3% of the rated voltage. This value is more stringent than the ±5% of a typical power grid. If the actual measured node voltage... Its rated target deviation Exceeded this This will trigger the subsequent parameter adjustment process.

[0324] The specific implementation method for determining the voltage stability parameters that need to be adjusted, based on the analysis process of the trigger control effect, is as follows:

[0325] 1. Diagnostic Analysis (Inferring the Source of Error):

[0326] • Analytical objects: the final control instruction sequence executed in S106, the dynamic allocation strategy of compensation capacity in S104 / S106, and the phase trend prediction after correction in S103.

[0327] Method: Error source analysis. The analysis tool will perform a reverse check:

[0328] • Is the current deviation caused by insufficient compensation capacity? (Check) ).

[0329] • Is the current deviation caused by instruction execution delay or conflict (check the S105 result).

[0330] • Is the current deviation caused by an error in the initial trend prediction? (Check the validity of the S102 / S103 model.)

[0331] 2. Define the adjustment goals:

[0332] Based on the source tracing results, determine the root cause of the voltage deviation exceeding the limit. For example, if it is found that the compensation capacity is consistently insufficient, then the voltage stability parameters affecting the compensation capacity allocation need to be adjusted.

[0333] 3. Parameter adjustment (adaptive feedback):

[0334] The analysis results are used as feedback to iteratively update the instruction sequence segmentation adjustment scheme in S104. The system employs a closed-loop mechanism to iteratively update the instruction sequence segmentation adjustment scheme, adaptively adjusting the relevant voltage stability parameters.

[0335] The voltage stability parameters refer to a series of key variables used in microgrid control models to define and evaluate system voltage stability. Adjusting these parameters aims to change the sensitivity and responsiveness of the control strategy. Feasible voltage stability parameters include:

[0336] 1. Voltage stability margin The safety factor used to constrain voltage deviation may be used in the optimization objective or constraint of S106.

[0337] If the voltage exceeds the limit: reduce (Relax constraints to allow for greater capacity compensation), or increase (Tightening constraints and requiring more precise control).

[0338] 2. Voltage phase angle coupling sensitivity A parameter that describes the degree to which reactive power affects voltage.

[0339] If the voltage exceeds the limit: Adjust this parameter so that the system calculates the compensation capacity requirement. At that time, the response to voltage deviation is either more sensitive or less sensitive.

[0340] 3. Target voltage setpoint The control target voltage for each time period in the S104 segmented adjustment scheme.

[0341] If the voltage exceeds the limit: Dynamic fine-tuning This makes it closer to the current actual operating conditions and reduces the difficulty of control.

[0342] In summary, the core of S107 lies in using closed-loop feedback to adaptively modify key parameters related to voltage stability in the control strategy by utilizing deviations in actual execution results, thereby improving the accuracy and effectiveness of the next regulation.

[0343] In one possible implementation, for voltage stability parameters, the operating status data in the microgrid layer is obtained, and a corresponding control command sequence is generated through a closed-loop mechanism. The priority distribution of the command sequence is determined. The generation of the command sequence is based on the reverse correction of the execution deviation of S106. Its logic mainly relies on iteratively updating the segmented adjustment scheme in S104 / S106.

[0344] Specific feasible implementation methods (closed-loop iterative generation):

[0345] 1. Determine the correction requirements (based on the S107 diagnostic results)

[0346] Input data: S107 The voltage stability parameters that need to be adjusted as determined in the previous step, and the node anomaly information fed back in S106 (e.g., the voltage deviation of node A exceeds the ±2% threshold).

[0347] • Calculate the correction amount: Utilize the adjusted voltage stability parameters (e.g., higher voltage sensitivity or a tighter target voltage). ), recalculate the additional compensation capacity required for abnormal nodes. .

[0348]

[0349] 2. Iteratively update the segmented adjustment scheme (core)

[0350] • Update S104 solution: [The following text appears to be a separate, unrelated section:] ... It is superimposed onto the corresponding time period and corresponding node in the instruction sequence segmentation adjustment scheme generated by S104.

[0351] For example, if node A is diagnosed as having a persistent lack of reactive power, then the allocated capacity parameters for that node in future time segments will be adjusted. Increase .

[0352] • Generate a new initial command sequence: Based on the updated segmented adjustment scheme, the system regenerates a new set of initial control command sequences. This new sequence reflects the adaptive control's correction of previous failures.

[0353] 3. Apply the optimization logic of S105 / S106

[0354] To ensure that the newly generated instruction sequence is executable and optimal, the system will again apply the optimization logic from S105 and S106 (although the complete flow of S107 is not directly described in the document, it is necessary in closed-loop control):

[0355] • Priority sorting: Perform S105 priority sorting on the newly generated instruction sequence (based on latency impact and real-time requirements).

[0356] • Optimized screening: The S106 optimized sequence generation method is applied to screen out instruction combinations that meet the power grid stability constraints from the new sequence.

[0357] In summary, the generation of the corresponding control command sequence is not a process that starts from scratch. Instead, it uses the diagnostic results of S107 as feedback to perform closed-loop selection and correction on the original segmented adjustment scheme. Then, the command optimization and sorting process is run again to finally obtain a new command sequence that is expected to successfully correct the voltage deviation.

[0358] In one possible implementation, the key control parameters extracted from the reactive power compensation strategy and fed back to the microgrid layer for real-time execution, along with the acquisition of voltage data after execution and the determination of the system voltage stability, refer to the key control variables that are finally determined after closed-loop iterative correction in S107 (including parameter adaptive adjustment, instruction sequence rearrangement, etc.) and need to be immediately pushed to the microgrid execution layer (such as STATCOM, SVC controller). These parameters mainly include:

[0359] 1. Final Compensation Instruction Sequence: A conflict-free, time-sequence-guaranteed final execution instruction set. It includes the device ID, switching capacity, and precise execution time. The purpose is to ensure that all reactive power compensation actions are executed on time, in the correct quantity, and without conflict.

[0360] 2. Target capacity of equipment ( ): The final reactive power capacity that each reactive power compensation device needs to activate or deactivate within the control cycle. The purpose is to directly control the device to perform reactive power compensation.

[0361] 3. Target running point ( , ): The target phase angle and target voltage values ​​that the system expects to achieve within the control cycle. The purpose is to ensure that the compensation action drives the system to a new stable operating state.

[0362] 4. Dynamic control gain: Adjusted PI / PID controller gain (if included in the control scheme). The purpose is to ensure that the response speed and control accuracy of the compensation equipment meet the latest adaptive requirements.

[0363] The process of extracting key control parameters involves converting the "final reactive power compensation control scheme" iteratively updated in step S107 into machine instructions and control signals that the microgrid control system can recognize. A specific feasible extraction implementation method is as follows:

[0364] 1. Data formatting and packaging:

[0365] The "final reactive power compensation control scheme" (usually structured data or internal model parameters) obtained after the S107 iteration update is formatted and converted into a data packet that conforms to the microgrid control layer communication protocol (such as IEC 61850, Modbus).

[0366] 2. Instruction sequence extraction:

[0367] The conflict-free instruction sequence (including execution time, device ID, and compensation capacity) is parsed from the final solution, which is the most direct and key control parameter.

[0368] 3. Parameter mapping and conversion:

[0369] The higher-level control parameters obtained from the computation layer (such as the optimized target voltage) This maps and converts setpoints or control signals that the lower-level device controller can understand. For example, it translates Mvar-level compensation requirements into current or pulse-width modulation (PWM) instructions for the device controller.

[0370] 4. Automated tool push:

[0371] According to the process mentioned at the end of S105, an automated distribution tool is used to push the extracted key control parameters (i.e., the final instruction sequence) to the target device for execution in sequence and on time.

[0372] The key to extracting critical control parameters from the reactive power compensation strategy lies in accuracy and real-time performance, ensuring that the correction results obtained through complex closed-loop optimization can be quickly and accurately applied to the actual control execution of the microgrid. It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this invention will not describe the various possible combinations separately. Furthermore, different embodiments of this invention can also be arbitrarily combined, as long as they do not violate the spirit of this invention, and should also be considered as part of the disclosure of this invention.

Claims

1. An adaptive control method for improving the grid stability of new energy power generation in a microgrid, characterized in that, The method comprises: Collecting output fluctuation signals of new energy power generation and load change data from a micro-grid system, processing the collected data using frequency domain analysis technology, extracting phase shift dynamic monitoring features, and obtaining basic description information of phase change; According to the phase shift dynamic monitoring features, combining historical data time series, constructing a fluctuation periodicity analysis model of phase change, dividing a prediction time window, identifying phase mutation points, and obtaining future trend prediction results of phase change; If the phase mutation point exceeds the preset fluctuation range, adjust the trend extrapolation logic through data smoothing preprocessing technology, combine the prediction error correction, and obtain the corrected phase change trend description; According to the corrected phase change trend description, fuse voltage deviation real-time feedback data, generate instruction sequence segmented adjustment scheme of reactive power compensation, and determine the preliminary compensation capacity dynamic allocation strategy; If the strategy finds potential inconsistency in instruction execution conflict detection, adjust the instruction priority through control time delay optimization processing technology, combine the equipment response time evaluation data, and obtain the optimized control instruction sequence; Through the optimized control instruction sequence, combining the power grid stability constraint condition, real-time regulation and control of the micro-grid system, obtaining the actual execution effect of compensation action, judging whether the system voltage deviation meets the stability requirement; According to the judgment result, the execution effect is fed back to the micro-grid control layer, and the instruction sequence segmented adjustment scheme is iteratively updated using a closed-loop mechanism to determine the final reactive power compensation regulation scheme. 2.The adaptive control method for improving the grid-connected stability of new energy power generation of a microgrid according to claim 1, characterized in that, The method comprises: Collecting output fluctuation signals of new energy power generation and load change data from a micro-grid system, processing the collected data using frequency domain analysis technology, extracting phase shift dynamic monitoring features, and obtaining basic description information of phase change; According to the phase shift dynamic monitoring features, combining historical data time series, constructing a fluctuation periodicity analysis model of phase change, dividing a prediction time window, identifying phase mutation points, and obtaining future trend prediction results of phase change; If the phase mutation point exceeds the preset fluctuation range, adjust the trend extrapolation logic through data smoothing preprocessing technology, combine the prediction error correction, and obtain the corrected phase change trend description; According to the corrected phase change trend description, fuse voltage deviation real-time feedback data, generate instruction sequence segmented adjustment scheme of reactive power compensation, and determine the preliminary compensation capacity dynamic allocation strategy; If the strategy finds potential inconsistency in instruction execution conflict detection, adjust the instruction priority through control time delay optimization processing technology, combine the equipment response time evaluation data, and obtain the optimized control instruction sequence; Through the optimized control instruction sequence, combining the power grid stability constraint condition, real-time regulation and control of the micro-grid system, obtaining the actual execution effect of compensation action, judging whether the system voltage deviation meets the stability requirement; According to the judgment result, the execution effect is fed back to the micro-grid control layer, and the instruction sequence segmented adjustment scheme is iteratively updated using a closed-loop mechanism to determine the final reactive power compensation regulation scheme. The method comprises: Collecting output fluctuation signals of new energy power generation and load change data from a micro-grid system, processing the collected data using frequency domain analysis technology, extracting phase shift dynamic monitoring features, and obtaining basic description information of phase change; According to the phase shift dynamic monitoring features, combining historical data time series, constructing a fluctuation periodicity analysis model of phase change, dividing a prediction time window, identifying phase mutation points, and obtaining future trend prediction results of phase change; If the phase mutation point exceeds the preset fluctuation range, adjust the trend extrapolation logic through data smoothing preprocessing technology, combine the prediction error correction, and obtain the corrected phase change trend description; According to the corrected phase change trend description, fuse voltage deviation real-time feedback data, generate instruction sequence segmented adjustment scheme of reactive power compensation, and determine the preliminary compensation capacity dynamic allocation strategy; If the strategy finds potential inconsistency in instruction execution conflict detection, adjust the instruction priority through control time delay optimization processing technology, combine the equipment response time evaluation data, and obtain the optimized control instruction sequence; Through the optimized control instruction sequence, combining the power grid stability constraint condition, real-time regulation and control of the micro-grid system, obtaining the actual execution effect of compensation action, judging whether the system voltage deviation meets the stability requirement; According to the judgment result, the execution effect is fed back to the micro-grid control layer, and the instruction sequence segmented adjustment scheme is iteratively updated using a closed-loop mechanism to determine the final reactive power compensation regulation scheme. ​ ​ ​ ​ 3.The adaptive control method for improving the grid-connected stability of new energy power generation of a microgrid according to claim 1, characterized in that, ​ The dynamic monitoring data of the phase shift is acquired, and corresponding historical data and time sequence information are extracted, data cleaning processing is performed, and a standardized phase monitoring data set is obtained; For the standardized phase monitoring data set, a time sequence decomposition method is used to separate the periodic fluctuation component and the aperiodic component, and the periodic fluctuation law in the phase change is determined; According to the periodic fluctuation law, a phase change model is constructed, and the long short-term memory network algorithm is used to simulate the future trend of the phase change, and a preliminary trend prediction result is obtained; Through the preliminary trend prediction result, a plurality of prediction time windows are divided and the phase change amplitude is analyzed, and if the amplitude exceeds a preset amplitude threshold, it is judged that there is a potential mutation position in the window; For the potential mutation position, the phase change record in the historical data is combined to confirm the mutation point, and if the mutation point is consistent with the historical mode, the accuracy of the mutation position is determined; According to the determined mutation position and the preliminary trend prediction result, the prediction direction of the future trend is adjusted, and the final phase change trend prediction data is generated; Through the final phase change trend prediction data, combined with the dynamic monitoring result, the future trend report data of the phase change is output. 4.The adaptive control method for improving the grid-connected stability of new energy power generation of a microgrid according to claim 1, characterized in that, If the prediction result shows that the phase mutation point exceeds the preset fluctuation range, the trend extrapolation logic is adjusted by data smoothing preprocessing technology, combined with prediction error correction, to obtain the corrected phase change trend description, including: Collecting the original data of the phase change, obtaining the initial prediction trend data therefrom, using a pre-established analysis model to obtain the preliminary mutation point distribution; If the preliminary mutation point obtained exceeds the preset fluctuation range, data smoothing technology is used for processing to determine the smoothed data sequence; According to the smoothed data sequence, combined with the logic adjustment method of trend extrapolation, the adjusted trend prediction path is obtained, and it is judged whether there is still a point position exceeding the fluctuation range; If there is still, the prediction path is processed twice by error correction method to obtain the corrected phase change trend; For the corrected phase change trend, the stability of the trend is evaluated using a preset standard to determine whether the final correction result meets the expected range; Analyzing the final correction result, generating detailed phase change description content, and obtaining complete trend change record; According to the complete trend change record, the parameters of the subsequent prediction model are optimized by combining the processing method, and more accurate prediction basis is obtained.

5. The adaptive control method for improving the grid stability of new energy power generation of a microgrid according to claim 1, characterized in that, According to the corrected phase change trend description, the real-time feedback data of the voltage deviation are fused to generate a segmented adjustment scheme of the instruction sequence of reactive power compensation, and a preliminary compensation capacity dynamic allocation strategy is determined, including: By collecting the phase change data in the system operation, the trend description information of the phase change is obtained, and the preliminary influence of the phase change on the system stability is determined; According to the trend description information of the phase change, combined with the real-time feedback voltage deviation data, the correlation between the voltage deviation and the phase change is determined by data combination; If the correlation exceeds a preset correlation threshold, the instruction sequence generation of reactive power compensation is triggered, and the priority of the instruction sequence is determined; By priority sorting of the instruction sequence, specific parameters of the segmented adjustment are obtained, and an adjustment scheme content for different time periods is obtained in a segmented adjustment manner; According to the segmented adjustment scheme content, an allocation requirement of compensation capacity is obtained, and a preliminary allocation strategy of the compensation capacity is determined by using a dynamic allocation logic; For the preliminary allocation strategy of the compensation capacity, the allocation strategy is optimized by using a support vector machine algorithm through real-time feedback data updating, and a final reactive power compensation allocation result is obtained; Through the final reactive power compensation allocation result, stability data of system operation is obtained, and an adjustment effect evaluation of the reactive power compensation on the system operation is determined. 6.The adaptive control method for improving the grid-connected stability of new energy power generation in a microgrid according to claim 1, wherein, If the strategy finds potential inconsistency in instruction execution conflict detection, the instruction priority sorting is adjusted by a control time delay optimization processing technology, and an optimized control instruction sequence is obtained in combination with device response time evaluation data, including: By a conflict detection mechanism, identification information of potential inconsistency is obtained from the instruction execution process, and an abnormal instruction set is determined; According to the abnormal instruction set, a delay influence range of each instruction is obtained by using a control time delay analysis tool, and a delay priority list is obtained; For the delay priority list, the real-time requirement of instruction execution is judged by using device response data, if the response time of a certain instruction exceeds a preset response time threshold, the priority is improved, and an adjusted sorting result is obtained; From the adjusted sorting result, a high-priority instruction subset is obtained, the execution interval between instructions is determined in combination with time evaluation data, and a preliminary control instruction sequence is generated; For the preliminary control instruction sequence, whether there is an execution conflict is judged by using an optimization technology processing, if a conflict is detected, the time sequence of the related instructions is fine-tuned, and a conflict-free instruction sequence is obtained; According to the conflict-free instruction sequence, feedback information of the device execution state is obtained in combination with response data analysis, and a final control instruction sequence is determined; The final control instruction is pushed to the target device in sequence by using an automatic distribution tool, and the execution adjustment process is completed. 7.The adaptive control method for improving the grid-connected stability of new energy power generation of a microgrid according to claim 1, characterized in that, By the optimized control instruction sequence, the actual execution effect of compensation action is obtained by real-time regulation and control of the micro-grid system in combination with the grid stability constraint condition, whether the system voltage deviation meets the stability requirement is judged, including: By a pre-established control instruction library, load data and power output data of the current period are obtained according to the operation state of the micro-grid system, and an initial control instruction set is determined; From the initial control instruction set, an instruction combination meeting the grid stability requirement is selected, and an optimized control instruction sequence is obtained; According to the optimized control instruction sequence, compensation action data of each node is obtained by implementing real-time regulation and control of the micro-grid system in combination with the constraint condition; The compensation action data is collected in real time, the execution effect of each node is analyzed, and whether there is an abnormal response condition is judged, if the execution effect of a certain node deviates from a preset threshold, abnormal information of the node is recorded; According to the abnormal information, the voltage fluctuation of each node is analyzed in combination with the monitoring data of the voltage deviation, and the specific range of the voltage deviation is determined. According to the specific range of voltage deviation, the system is judged whether to meet the stability requirements by comparing with the stability standard, and the final regulation result is obtained; Through the final regulation result, the data content in the control instruction library is updated, and a closed-loop feedback mechanism is formed. 8.The adaptive control method for improving the grid-connected stability of new energy power generation of a microgrid according to claim 1, characterized in that, According to the judgment result, the execution effect is fed back to the micro-grid control layer, and the instruction sequence segmentation adjustment scheme is iteratively updated by adopting a closed-loop mechanism to determine the final reactive power compensation regulation scheme, including: Real-time data of system voltage is obtained from the monitoring equipment, the fluctuation of the system voltage is analyzed, and a preliminary deviation evaluation result is obtained; If the deviation exceeds the preset deviation threshold range, the regulation effect analysis process is triggered, and the voltage stability parameter that needs to be adjusted is determined; For the voltage stability parameter, the operation state data in the micro-grid layer is obtained, the corresponding control instruction sequence is generated through the closed-loop mechanism, and the priority distribution is judged; According to the priority distribution, the sequence segmentation processing is implemented, the instruction is decomposed into multiple execution units, and the segmented adjustment scheme is obtained; For the segmented adjustment scheme, an iterative optimization method is adopted to update the details of the scheme through multiple simulation calculations, and the final reactive power compensation strategy is determined; Key regulation parameters are extracted from the reactive power compensation strategy and fed back to the micro-grid layer for real-time execution, voltage data after execution is obtained, and the stability state of the system voltage is judged; Through continuous monitoring of the stability state of the system voltage, the deviation evaluation result is updated, a dynamic regulation process under the closed-loop mechanism is formed, and a long-term voltage optimization scheme is obtained.

Images