Microgrid load distribution control method, device and equipment based on entropy increase disturbance

Abstract

The invention discloses a micro-grid load distribution control method, device and equipment based on entropy increase disturbance, and the method comprises the steps: analyzing the operation state of a micro-grid through chaos measurement, and constructing an entropy increase disturbance excitation mechanism to actively excite the vitality of a system; after random disturbance is applied, inherent frequency characteristics of each power supply are extracted, and a frequency synchronization coordination mechanism is established to realize multi-source coordination; adopting counter-example search to actively identify distribution scheme vulnerabilities, and generating an enhanced distribution strategy to improve robustness; evaluating the load bearing capacity of each power supply, constructing a load conduction topology and analyzing pressure distribution; each power supply bearing limit parameter is obtained through boundary active detection, and the equipment potential is fully excavated; and a final control instruction is generated based on a multi-layer time sequence synchronization mechanism, and multi-power-supply coordinated output is realized. A chaos theory is introduced into micro-grid control, local optimum is avoided through active disturbance, a multi-layer coordination mechanism is adopted to adapt to new energy volatility, and a self-adaptive load distribution control scheme is provided for a micro-grid with high-proportion renewable energy access.

Description

Technical Field

[0001] This invention relates to the field of microgrid control technology, and in particular to a microgrid load distribution control method, apparatus and equipment based on entropy increase disturbance. Background Technology

[0002] Microgrids, as an effective form of distributed energy access to the grid, play a vital role in improving energy efficiency and enhancing power supply reliability. However, with the increasing penetration of renewable energy sources such as photovoltaics and wind power, power balance and load distribution within microgrids face new challenges.

[0003] Current microgrid load sharing technologies primarily rely on pre-defined optimization algorithms and fixed allocation strategies. This approach proves inadequate when dealing with the randomness and volatility of renewable energy output. Distributed power sources often operate independently, lacking effective coordination mechanisms, leading to frequent instances of some sources being overloaded while others remain idle. Furthermore, existing methods lack sufficient understanding of system operating boundaries, and conservative operating strategies limit the full potential of equipment. The load sharing process lacks proactive analysis of potential problems, often only responding passively after a fault occurs. In addition, the dynamic coupling characteristics between multiple power sources are not fully utilized, and the inherent oscillation characteristics and interaction mechanisms of each power source have not been thoroughly explored.

[0004] Given the complexity and uncertainty of the microgrid operating environment, traditional deterministic control methods are no longer sufficient to meet the needs of optimized system operation. There is an urgent need to explore new control concepts and methods to achieve adaptive optimization and coordinated control of microgrid load distribution. Summary of the Invention

[0005] This invention provides a microgrid load distribution control method, device, and equipment based on entropy increase disturbance. It comprehensively utilizes technologies such as chaos theory to stimulate system vitality, frequency coupling to achieve multi-source coordination, counterexample search to discover potential defects, proactive detection to tap equipment potential, and time synchronization to ensure coordinated operation. It constructs a complete technical system from system state perception, establishment of coordination mechanism, strategy optimization and improvement to coordinated control execution, enabling the microgrid to have the intelligent ability to autonomously explore the optimal operating point, dynamically adapt to environmental changes, and coordinate the complementary operation of multiple power sources.

[0006] The first aspect of this invention proposes a microgrid load distribution control method based on entropy increase disturbance, comprising the following steps: Real-time power output data and load demand data of power sources in the microgrid are acquired, and chaotic measurement analysis is performed on the real-time power output data to generate system entropy parameters. Based on the system entropy parameters, an entropy increase disturbance excitation mechanism is constructed. Based on the entropy increase perturbation excitation mechanism, randomized power output perturbation is performed on each power source to generate chaotic state operation data. The inherent frequency characteristics of each power source are extracted from the chaotic state operation data to obtain the resonance coupling matrix. The resonance coupling matrix is ​​used to generate a frequency synchronization coordination mechanism between power sources. A preliminary load allocation scheme is generated based on the frequency synchronization coordination mechanism. Counterexample search is performed on the preliminary load allocation scheme to generate allocation vulnerability identification data. An enhanced allocation strategy is obtained based on the allocation vulnerability identification data. Based on the enhanced allocation strategy, load-bearing capacity assessment data of each power source is obtained. The load demand data and the load-bearing capacity assessment data are matched and analyzed to generate a load transmission topology. The load transmission topology is then used to obtain pressure distribution data among the power sources. Based on the pressure distribution data, a boundary active detection instruction set is generated. The boundary active detection instruction set is used to drive each power supply to gradually approach the operating boundary to obtain boundary response data. Extreme value analysis is performed on the boundary response data to generate the load-bearing limit parameters of each power supply. Based on the load limit parameters, a unified timing adjustment benchmark is set, and the timing adjustment benchmark is used to synchronize the power output adjustment process of each power source to obtain coordinated timing data. Timing synchronization control parameters are generated based on the coordinated timing data. Based on the timing synchronization control parameters, a final load distribution control instruction set is generated. Each power source executes the control instruction set to coordinate output and obtain actual operating status data, thereby completing the microgrid load distribution control.

[0007] A second aspect of this invention provides a microgrid load distribution control device based on entropy increase disturbance, comprising: The data acquisition module is used to acquire real-time power output data and load demand data of the power sources in the microgrid, perform chaotic measurement analysis on the real-time power output data to generate system entropy parameters, and construct an entropy increase disturbance excitation mechanism based on the system entropy parameters. The chaotic coordination module is used to generate chaotic state operating data by randomly perturbing the output of each power source based on the entropy increase perturbation excitation mechanism, extract the inherent frequency characteristics of each power source from the chaotic state operating data to obtain the resonance coupling matrix, and use the resonance coupling matrix to generate a frequency synchronization coordination mechanism between power sources. The strategy optimization module is used to generate a preliminary load allocation scheme based on the frequency synchronization coordination mechanism, perform counterexample search on the preliminary load allocation scheme to generate allocation vulnerability identification data, and obtain an enhanced allocation strategy based on the allocation vulnerability identification data. The stress analysis module is used to obtain load capacity assessment data of each power source based on the enhanced allocation strategy, match and analyze the load demand data with the load capacity assessment data to generate a load transmission topology, and use the load transmission topology to obtain stress distribution data among each power source. The boundary detection module is used to generate an active boundary detection instruction set based on the pressure distribution data, use the active boundary detection instruction set to drive each power supply to gradually approach the operating boundary to obtain boundary response data, and perform extreme value analysis on the boundary response data to generate the load-bearing limit parameters of each power supply. The timing control module is used to set a unified timing adjustment benchmark based on the load limit parameters, use the timing adjustment benchmark to synchronize the power output adjustment process of each power source to obtain coordinated timing data, and generate timing synchronization control parameters based on the coordinated timing data. The instruction execution module is used to generate a final load distribution control instruction set based on the timing synchronization control parameters. Each power source executes the control instruction set to coordinate output and obtain actual operating status data, thereby completing the microgrid load distribution control.

[0008] A third aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of a microgrid load distribution control method based on entropy increase disturbance disclosed in the first aspect.

[0009] The beneficial effects of this invention are reflected in the following aspects: First, by applying chaos theory to microgrid control, the chaotic state of the system is actively stimulated through an entropy-increasing perturbation excitation mechanism. The inherent frequency characteristics and coupling relationships of each power source are extracted from the chaotic response, providing an analytical approach different from traditional models. This approach can discover hidden system characteristics and optimization potential. Simultaneously, the inherent synergy of multiple power sources is achieved through a frequency synchronization coordination mechanism, enhancing the coordinated operation capability of each power source. Second, a counterexample search technique is used to comprehensively identify vulnerabilities in the initial allocation scheme. Potential problems are actively exposed by constructing adversarial modes and conflict scenarios, changing the traditional passive approach to fault response. Combined with boundary active detection technology, the operating limits of each power source can be safely explored, fully tapping the potential of the equipment. This proactive optimization strategy improves the system's operating efficiency and reliability. Finally, a time-series synchronization control mechanism based on load-bearing limits is established. Through hierarchical time scales and phase difference configuration, coordinated action of multiple power sources is achieved, avoiding system shocks caused by simultaneous adjustments. The entire control process moves from disorder to order, from independence to collaboration, improving the intelligence level of load allocation and providing a new solution for the stable operation of microgrids under conditions of high-proportion renewable energy access.

[0010] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0011] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0012] Unless otherwise specified or defined, the same reference numerals in different figures represent the same or similar technical features, and different reference numerals may be used to represent the same or similar technical features.

[0013] Figure 1 This is a flowchart illustrating a microgrid load distribution control method based on entropy increase disturbance according to the present invention.

[0014] Figure 2 This is a structural block diagram of a microgrid load distribution control device based on entropy increase disturbance according to the present invention.

[0015] Figure 3 This is a schematic diagram of the structure of a computer device according to the present invention. Detailed Implementation

[0016] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0017] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0018] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0019] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0020] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0022] The technical solutions of the embodiments of this application will be described below.

[0023] like Figure 1 As shown, this embodiment of the invention provides a microgrid load sharing control method based on entropy increase disturbance, including the following steps S110-S170: Step S110: Obtain real-time power output data and load demand data of the power sources in the microgrid, perform chaotic measurement analysis on the real-time power output data to generate system entropy parameters, and construct an entropy increase disturbance excitation mechanism based on the system entropy parameters.

[0024] Specifically, the real-time output data acquisition of power sources within the microgrid employs a distributed data acquisition architecture. High-precision power measurement sensors are deployed at the output ends of various distributed power sources, such as photovoltaic (PV) generators, wind turbines, energy storage systems, and diesel generators, to achieve real-time monitoring. The power measurement sensors utilize a combination of current and voltage transformers based on the Hall effect principle, achieving a current measurement accuracy of 0.2% and a voltage measurement accuracy of 0.1%. The sampling frequency is set to 1kHz to ensure the capture of instantaneous fluctuations in power output. PV generator output data acquisition combines irradiance and temperature sensors to monitor the impact of changes in light intensity on power generation in real time. The irradiance measurement range is 0-1500 W / m² with an accuracy of ±2%, and the temperature measurement range is -40℃ to 85℃ with an accuracy of ±0.5℃. Wind turbine output data acquisition, in conjunction with wind speed and direction sensors, monitors the correlation between wind speed changes and power generation. The wind speed measurement range is 0-60 m / s with an accuracy of ±0.2 m / s, and the wind direction measurement accuracy is ±3°. The energy storage system's output data includes multi-dimensional parameters such as charging and discharging power, state of charge (SOC), and battery temperature, with SOC estimation accuracy controlled within ±2%. Data transmission adopts the IEC61850 communication protocol, constructing a high-speed data transmission network via fiber optic Ethernet, with transmission latency controlled within 10 milliseconds. Load demand data acquisition is achieved through smart meters and load monitoring terminals, covering different types of electricity demand such as residential, commercial, and industrial loads. The smart meter sampling interval is set at 15 minutes, with the sampling interval for key load nodes shortened to 1 minute. The load forecasting module uses the LSTM (Long Short-Term Memory) neural network algorithm to perform short-term load forecasting based on diverse inputs such as historical load data, weather forecast information, and holiday factors, with forecast time scales including 15 minutes, 1 hour, and 24 hours. High-frequency sampling is used on the power supply side to capture output fluctuation characteristics and perform chaotic analysis, while low-frequency sampling combined with the forecasting algorithm on the load side can meet load allocation requirements. The two are matched in time scale through the forecasting model. Through a distributed data acquisition architecture and intelligent forecasting algorithms, real-time power output data and load demand data reflecting the current operating status of the microgrid are acquired.

[0025] Chaotic metric analysis was performed on real-time power output data to generate system entropy parameters. The chaotic metric analysis adopted a nonlinear dynamics framework, treating the power output fluctuations of the microgrid system as chaotic behavior manifestations of a complex dynamic system. The analysis process first reconstructed the phase space of the collected multi-source power output data, then extended the one-dimensional time series to a high-dimensional phase space using a delayed coordinate embedding method. The delay time τ was determined by minimizing the mutual information function, and the embedding dimension m was calculated using the spurious nearest neighbor method. In a typical photovoltaic-wind power-storage microgrid system, the optimal delay time τ for photovoltaic power output data was 3 sampling periods, and the embedding dimension m was 5; for wind power output data, τ was 5 sampling periods, and m was 7, reflecting the higher chaotic characteristics of wind power output. The Lyapunov exponent was calculated using the Wolf algorithm, quantifying the degree of chaos by tracking the separation rate of neighboring orbits in the phase space. A positive Lyapunov exponent indicates the presence of chaotic behavior in the system, and a larger exponent indicates greater sensitivity to initial conditions. The correlation dimension is calculated using the Gaussian General (GP) algorithm. By analyzing the distribution density of points in phase space, the complexity of the system is reflected. The fractional characteristics of the correlation dimension are an important indicator of chaotic systems. Kolmogorov entropy is used to quantify the rate of information generation in the system, reflecting the unpredictability of its motion. Under typical operating conditions of alternating sunny and cloudy skies, the Lyapunov exponent of the photovoltaic system reaches 0.15, exhibiting significant chaotic characteristics. Nonlinear dynamic analysis methods are used to further explore the chaotic properties inherent in the microgrid output data.

[0026] In some embodiments, the step of generating system entropy parameters by performing chaotic metric analysis on the real-time output data includes: performing disordered metric analysis on the real-time output data to obtain chaotic feature vectors; performing entropy integration processing on the chaotic feature vectors to generate initial entropy data; and performing time series standardization processing on the initial entropy data to obtain system entropy parameters.

[0027] First, disorder quantification analysis is performed on real-time power output data to obtain chaotic feature vectors. The disorder quantification analysis employs a multi-scale entropy analysis method, coarsening the original power output time series at different time scales and calculating sample entropy values ​​at each scale. Sample entropy calculation uses a template matching algorithm, setting the template length m=2 and the similarity tolerance r=0.2 times the standard deviation. The regularity of the sequence is quantified by statistically analyzing the conditional probability of similar templates appearing. For photovoltaic power output data, sample entropy is calculated at four time scales: 1 minute, 5 minutes, 15 minutes, and 60 minutes, yielding a multi-scale entropy spectrum. Permutation entropy analysis converts continuous data into ordinal patterns through symbolization, calculating the probability distribution entropy values ​​for different permutation patterns. The embedding dimension is set to 3-7, and the delay time is set to 1-5. Approximate entropy calculation uses a sliding window technique, with a window length of 100 data points and a sliding step size of 10 data points, to track the dynamic changes in system disorder in real time. Fuzzy entropy introduces a fuzzy membership function to replace hard threshold judgment, using a Gaussian membership function, with the fuzzy factor set adaptively within the range of 0.1-0.3. During periods of severe fluctuations in wind power output, the permutation entropy value can reach above 0.95, indicating that the system is in a highly disordered state. Fluctuation complexity analysis employs the Lempel-Ziv algorithm, reflecting its complexity by calculating the compressibility of the sequence. Chaotic indices such as Lyapunov exponent, correlation dimension, Kolmogorov entropy, multi-scale entropy, permutation entropy, approximate entropy, fuzzy entropy, and LZ complexity are normalized to form a chaotic feature vector.

[0028] Then, the entropy integration process employs a weighted fusion algorithm, assigning weight coefficients based on the contribution of each chaotic indicator to the system's disorder. Weight determination utilizes an entropy weighting method combined with the analytic hierarchy process (AHP). First, the information entropy of each indicator is calculated to determine the objective weight; then, a judgment matrix is ​​constructed based on expert experience to determine the subjective weight; finally, a comprehensive weight is obtained through linear combination. In a typical microgrid system, the Lyapunov exponent has a weight of 0.25, the multi-scale entropy weight is 0.20, the permutation entropy weight is 0.15, and the weights of other indicators are distributed between 0.05 and 0.10. The integration calculation uses the trapezoidal integration method, numerically integrating discrete chaotic characteristic values ​​over time. The integration time window is set to 15 minutes, consistent with the power system dispatch cycle. To avoid the influence of outliers at a single moment, a moving average filter is used to smooth the integration results, with a filter window length of 5 data points. A dynamic weight adjustment mechanism adjusts the weights of each indicator in real time according to the system's operating conditions. When the proportion of renewable energy output is high, the weights of the Lyapunov exponent and permutation entropy are increased; when load fluctuations are severe, the weight of the multi-scale entropy is increased. The cumulative entropy calculation takes into account the influence of historical entropy values, employing an exponential decay factor (α=0.95) to weight historical data, ensuring that recent data has a dominant influence on the current entropy value. Through weighted integration, the multidimensional chaotic feature vectors are fused into scalar initial entropy values ​​that reflect the overall disorder of the system.

[0029] Finally, time-series standardization was performed on the initial entropy data to obtain the system entropy parameters. Standardization first involved outlier detection and correction, using the 3σ criterion to identify outliers exceeding the mean ± 3 standard deviations, and then correcting them using linear or spline interpolation. Trend decomposition employed the STL time-series decomposition algorithm to decompose the initial entropy data into trend, seasonal, and random components, separating the steady-state entropy reflecting the long-term evolution trend of the system and the dynamic entropy reflecting short-term fluctuations. Seasonal adjustment considered the diurnal and seasonal characteristics of the microgrid operation, using the X-12-ARIMA method to extract and eliminate seasonal factors. Normalization used the max-min standardization method, mapping the entropy values ​​to the [0,1] interval, where 0 represents a completely ordered state and 1 represents a state of maximum chaos. During a typical operating day, the system entropy value reached 0.7-0.8 in the early morning and evening due to rapid changes in photovoltaic output; during midday, when sunlight was stable, the system entropy value decreased to 0.2-0.3. The moving average processing employs the exponentially weighted moving average (EWMA) method with a smoothing factor λ=0.3, eliminating high-frequency noise while preserving the dynamic characteristics of the entropy value. The standardized system entropy parameters include four dimensions: instantaneous entropy, average entropy, rate of change of entropy, and amplitude of entropy fluctuation, forming a complete description of the system's chaotic state. Through time series standardization, comparable and stable system entropy parameters are generated.

[0030] An entropy-increasing perturbation excitation mechanism is constructed based on the system entropy parameter. The core idea of ​​this mechanism is to actively introduce controlled perturbations to increase the chaos level of the system and stimulate it to explore a better operating state. The perturbation intensity is designed using an adaptive control strategy, dynamically adjusting the perturbation amplitude based on the deviation between the current system entropy and the target entropy. When the system entropy is below 0.3, it indicates that the system may be trapped in a local optimum, requiring a stronger perturbation to stimulate system activity; when the system entropy is above 0.8, it indicates that the system is too chaotic, requiring a reduction in perturbation intensity to maintain system stability. The perturbation signal is generated using chaotic mapping methods, including classic chaos generators such as the Logistic mapping (with control parameter μ adjusted in the range of 3.57-4.0), the Henon mapping (parameters a=1.4, b=0.3), and the Lorenz system. The chaotic characteristics of the perturbation are controlled by adjusting the mapping parameters. The selection of disturbance injection points is based on sensitivity analysis results, prioritizing control variables that significantly affect the system entropy, such as a 5%-10% disturbance amplitude added to the charging and discharging power of the energy storage system, or a controllable load switching timing offset of ±15 minutes. The disturbance timing design employs an intermittent excitation mode, with a disturbance duration of 5-10 minutes and an interval of 15-30 minutes, ensuring the disturbance effect while avoiding system fatigue. A multi-point coordinated disturbance mechanism uses correlation analysis to determine the coupling relationship between different disturbance points, designing orthogonal disturbance sequences to reduce mutual interference. The disturbance frequency is set within the range of 0.1-1Hz based on the system's inertial characteristics to ensure that the disturbance signal can effectively propagate and affect the system state. In a microgrid application scenario in an industrial park, at 10:00 AM, the system detected an entropy value dropping to 0.25, indicating that the output of photovoltaic and wind power was stabilizing but the dispatch mode was fixed. At this time, the entropy increase disturbance excitation mechanism was automatically triggered, applying a ±8% amplitude disturbance based on the Logistic mapping to the charging and discharging power of the energy storage system, and introducing a ±20-minute time-shift disturbance generated by the Henon mapping to the controllable load switching sequence. The disturbance signal frequency was set to 0.5Hz. After 8 minutes of continuous action, the system entropy value gradually rose to about 0.45, successfully stimulating the system to break out of the original operating mode and exploring a better energy dispatch scheme. Through adaptive adjustment of disturbance intensity, chaotic signal generation, and multi-point collaborative injection, an entropy increase disturbance excitation mechanism based on system entropy value feedback was constructed.

[0031] Step S120: Randomize the output perturbation of each power source based on the entropy increase perturbation excitation mechanism to generate chaotic state operation data, extract the inherent frequency characteristics of each power source from the chaotic state operation data to obtain the resonance coupling matrix, and use the resonance coupling matrix to generate a frequency synchronization coordination mechanism between power sources.

[0032] Specifically, by applying random disturbances based on entropy increase mechanism to each power source of the microgrid to induce chaotic operation of the system, the inherent oscillation characteristics of each power source are identified from the chaotic response and a resonant coupling relationship is constructed, ultimately forming a multi-power source frequency synchronization and coordination control strategy.

[0033] In some embodiments, the step of generating chaotic operating data by randomly perturbing the output of each power source based on the entropy increase perturbation excitation mechanism includes: generating a random perturbation parameter set based on the entropy increase perturbation excitation mechanism; inputting the random perturbation parameter set into each power source to obtain a power source-specific perturbation command; using the power source-specific perturbation command to drive each power source model to execute output perturbation to obtain perturbation response data; and integrating and processing the perturbation response data to generate chaotic operating data.

[0034] First, using the chaotic mapping module in the entropy-increasing perturbation excitation mechanism, a set of random perturbation parameters satisfying different statistical characteristics is generated through parameterized configuration. The Logistic mapping parameter μ is randomly selected between 3.75 and 4.0 to generate perturbation sequences with different degrees of chaos; the Henon mapping parameter a varies within the range of 1.2 to 1.4, while parameter b is fixed at 0.3, producing a two-dimensional chaotic attractor projection sequence; the Lorenz system parameters σ=10, ρ=28, and β=8 / 3 are used to obtain the three-dimensional chaotic trajectory through numerical integration. The random perturbation parameter set includes parameters in five dimensions: perturbation amplitude, perturbation frequency, perturbation phase, perturbation duration, and perturbation interval. The disturbance amplitude is determined as a percentage of the rated capacity of each power supply, typically ranging from 5% to 15%. The disturbance frequency covers a wide frequency band of 0.01Hz to 10Hz, with a focus on the low-frequency oscillation range of 0.1Hz to 2Hz. The disturbance phase is uniformly distributed between 0 and 2π to ensure phase dispersion of disturbances from different power supplies. The disturbance duration adopts an exponential distribution with an average duration of 5 minutes. The disturbance interval adopts a Poisson distribution with an average interval of 10 minutes. The parameter set size is determined based on the number of power supplies and the duration, with a typical configuration generating 500-1000 sets of disturbance parameters for each power supply.

[0035] Then, the random disturbance parameter set is allocated to each power source to obtain power source-specific disturbance commands. The allocation strategy considers the technical characteristics and operational constraints of various power sources, and a constrained random allocation algorithm is used to ensure the executability of the disturbance commands. The disturbance commands for photovoltaic power sources mainly affect the duty cycle of the DC / DC converter and the modulation depth of the DC / AC inverter, with the disturbance amplitude limited to a range that does not affect MPPT efficiency; the disturbance commands for wind turbine generators are converted into pitch angle adjustment and generator torque setpoints, taking into account the response speed limitations of mechanical components; the disturbance commands for energy storage systems are directly mapped to charge and discharge power commands, and constraint checks are performed based on the current SOC state; the disturbance commands for diesel generators affect the speed governor and excitation regulator, with the disturbance change rate limited to the equipment's tolerance range. Command generation uses a real-time operating system to ensure that the timing accuracy of the disturbance commands reaches the millisecond level. Through constraint mapping and real-time scheduling, a sequence of dedicated disturbance commands that can be directly executed by each power source is generated.

[0036] Next, power supply-specific disturbance commands are used to drive each power supply to execute output disturbances and acquire disturbance response data. This is achieved through the remote control function of the microgrid SCADA system. After safety verification, the disturbance commands are sent to the local controllers of each power supply. Real-time monitoring during execution ensures that the disturbance occurs within safe boundaries, with protection limits set for voltage deviation ±7% and frequency deviation ±0.5Hz. The disturbance is automatically terminated when these limits are exceeded. Response data acquisition adopts a distributed measurement architecture, with synchronous phasor measurement units (PMUs) configured at the output of each power supply to achieve synchronous acquisition of electrical quantities. The acquired data includes electrical parameters such as three-phase voltage, three-phase current, active power, reactive power, power factor, frequency, and phase angle, as well as operating status parameters such as temperature, speed, and SOC. Data timestamp accuracy reaches the microsecond level, ensuring time alignment of data from different measurement points. During the disturbance process, the complete dynamic process of the system from steady state to disturbance excitation and then to the new steady state is recorded. Typical disturbance response data contains millions of data points.

[0037] Finally, the disturbance response data is integrated and processed to generate chaotic state operation data. First, time alignment is performed, unifying the scattered data to the same time reference based on a GPS synchronized clock. The data cleaning process identifies and processes outliers, missing values, and noise interference, using the 3σ criterion to remove outliers and linear interpolation to fill in short-term missing data. Data standardization employs zero-mean normalization to eliminate the influence of different physical quantities' dimensions. Feature enhancement processing calculates derived features such as the rate of change of power output, acceleration, and cross-correlation coefficients for each power source, enriching the data's information dimensions. Data segmentation divides the long-term series into multiple segments according to the changes in the system's operating state, with each segment corresponding to a quasi-steady-state or transient process. Through data integration and preprocessing, a structured, standardized, and information-rich chaotic state operation dataset is generated.

[0038] In some embodiments, the step of extracting the inherent frequency features of each power source from the chaotic state operating data to obtain the resonant coupling matrix includes: performing a frequency domain transformation on the chaotic state operating data to obtain frequency spectrum distribution data; identifying the dominant frequency of each power source from the frequency spectrum distribution data to generate frequency feature identifiers; and performing coupling strength analysis based on the frequency feature identifiers to construct the resonant coupling matrix.

[0039] Frequency domain transformation is performed on chaotic operating data to obtain frequency spectrum distribution data. A multi-resolution analysis framework is adopted, selecting appropriate processing methods based on the physical characteristics and analytical requirements of different frequency bands, and using different analytical accuracies in different frequency ranges. The low-frequency band (0.01-1Hz) contains the main oscillation modes of the microgrid, employing high-resolution Welch power spectrum estimation with a window length set to 1 / 8 of the data length and an overlap rate of 75%. Noise is suppressed through multi-window averaging, achieving a frequency resolution of 0.01Hz. The mid-frequency band (1-10Hz) covers the dynamic response characteristics of the control loop, using standard FFT combined with a Hamming window function to reduce spectral leakage and balance computational efficiency and accuracy requirements. The high-frequency band (10-100Hz) mainly reflects switching frequencies and higher harmonics, employing fast algorithms to reduce computational complexity. Cross-spectral analysis calculates the cross-power spectral density between different power supply outputs, preserving amplitude and phase information in complex form to reveal the coupling strength and phase relationship between power supplies at various frequency points. The coherence function is calculated to assess the degree of linear correlation at different frequency components, with a value ranging from 0 to 1. Frequency points with a coherence coefficient greater than 0.8 are considered to have strong coupling, 0.5-0.8 are considered to have moderate coupling, and values ​​below 0.5 are considered to have weak coupling or no coupling. Detrending and windowing preprocessing are performed before frequency domain transformation to eliminate the influence of DC components and boundary effects. Through multi-resolution frequency domain transformation and cross-spectral analysis, fine frequency spectrum distribution data of the power output of each power source are obtained.

[0040] Frequency characteristic identifiers are generated by identifying the dominant frequencies of each power source from the frequency spectrum distribution data. An adaptive threshold peak detection algorithm is employed. First, the statistical characteristics of the power spectrum are calculated, and frequency points with power spectral density exceeding three times the average value and exhibiting local maxima are selected as candidate peaks. The peak selection process considers the physical meaning of the frequency and system characteristics, eliminating non-intrinsic frequency components such as the 50Hz power frequency and its integer multiples of harmonics (100Hz, 150Hz, etc.), retaining characteristic frequencies reflecting the dynamic characteristics of the power source. For photovoltaic power sources, the identified dominant frequency of 0.35Hz corresponds to the search oscillation of MPPT control, 0.52Hz reflects the dynamic response of the DC / DC converter, and 0.78Hz is related to the inverter current loop. For wind turbines, the 0.15Hz ultra-low frequency oscillation originates from the mechanical characteristics of the wind turbine rotor, 0.28Hz reflects the adjustment process of pitch angle control, and 0.45Hz is related to the generator electromagnetic torque control. For energy storage systems, 0.65Hz corresponds to the transition process of charge-discharge switching, and 1.2Hz and 1.8Hz reflect the response characteristics of the voltage control loop and current control loop, respectively. The frequency characteristic identifier is represented in a structured vector form, containing four core attributes: frequency value fi, normalized power spectral density amplitude Pi, initial phase φi, and quality factor Qi. The quality factor reflects the sharpness of the frequency components; a higher Q value indicates a more stable oscillation mode. Through systematic spectrum analysis and physical interpretation, frequency characteristic identifiers accurately describing the oscillation characteristics and control response of each power supply are generated.

[0041] Coupling strength analysis is performed based on the frequency characteristic identifiers to construct a resonance coupling matrix. First, modal analysis is used to decouple the frequency characteristic identifiers, constructing a state-space model of the multi-power system. The natural frequencies, damping ratios, and mode shape vectors of each oscillation mode are obtained by solving the system characteristic equations. In a three-power system including photovoltaic, wind power, and energy storage, the first mode exhibits a globally in-phase oscillation mode of 0.25Hz, with all power sources participating in the oscillation with similar amplitudes and phases. The second mode is a regional oscillation mode of 0.48Hz, with photovoltaic and wind power exhibiting anti-phase oscillation characteristics, while energy storage remains relatively stable. The third mode is a local oscillation mode of 0.82Hz, mainly dominated by the energy storage system. Participation factor analysis quantifies the contribution of each power source to different oscillation modes, observability analysis assesses the ability to observe each mode from the system output, and the mode damping ratio reflects the attenuation characteristics of the oscillation. These modal parameters together constitute standard coupling data describing the system's oscillation characteristics. Next, frequency anomaly signals are injected into the standard coupling data to obtain anomaly response data. The abnormal signal design covers typical disturbance conditions that may be encountered in actual operation: step-type frequency offset simulates frequency jumps caused by sudden switching of large-capacity loads, injecting a step signal with an amplitude of 0.2Hz and a duration of 5 seconds; pulse-type disturbance simulates instantaneous impacts caused by line faults or equipment tripping, using a pulse signal with an amplitude of 0.5Hz and a width of 2 seconds; random disturbance simulates continuous random interference such as wind speed fluctuations and cloud cover, using a Gaussian white noise signal with a standard deviation of 0.1Hz. Photovoltaic systems, due to their fast control response but low inertia, exhibit rapid but large-amplitude frequency fluctuations under disturbances; wind power systems, due to mechanical inertia, show relatively slow frequency changes but longer durations; energy storage systems, with their rapid power regulation capabilities, exhibit excellent frequency support characteristics. Through comprehensive abnormal injection tests, an abnormal response dataset reflecting the response characteristics of each power source under disturbance conditions is obtained. Then, anti-interference analysis is performed on the abnormal response data to generate robust coupling parameters. The anti-interference analysis, based on modal participation factors and abnormal response characteristics, calculates the degree of mutual influence of each power source under disturbance. By analyzing the correlation of frequency trajectories of various power sources in the abnormal response data, the dynamic coupling strength between power sources is extracted. Photovoltaic-wind power exhibits highly synchronized frequency changes during disturbances, with a correlation coefficient exceeding 0.85. The frequency responses of photovoltaic-energy storage show complementary characteristics, with energy storage rapidly compensating for the frequency shift of photovoltaic power. Wind power-energy storage coupling manifests as energy storage actively tracking the slow changes in wind power. Stability assessment calculates the degree of preservation of modal characteristics before and after the disturbance, while recovery capability assessment obtains the time constant by fitting the decay curve after disturbance elimination. Finally, robust coupling parameters are generated to quantitatively characterize the interaction strength between power sources under normal and abnormal operating conditions. Finally, a resonant coupling matrix is ​​constructed using these robust coupling parameters.The matrix construction directly uses robust coupling parameters as matrix elements, where the off-diagonal element Cij represents the coupling strength between power source i and power source j, and its value reflects the degree of mutual influence between the two in frequency oscillations. The coupling parameter between photovoltaic and wind power is the largest, reflecting the strong interaction between new energy power sources; the coupling parameter between energy storage and other power sources is moderate, reflecting the coordinating role of energy storage in the system. The diagonal elements are set as the self-regulation capability parameters of each power source, which are related to the modal damping characteristics. The symmetry of the matrix is ​​guaranteed by the physical characteristics of the interaction between power sources, and the positive definiteness ensures the stability of subsequent control design. By directly using robust coupling parameters that have been verified against interference, a resonant coupling matrix that accurately reflects the frequency coupling characteristics of multiple power sources in the microgrid is constructed.

[0042] A frequency synchronization coordination mechanism among power sources is generated using a resonant coupling matrix. Based on the dynamic relationships between power sources revealed by the resonant coupling matrix, a distributed cooperative control strategy is designed to achieve frequency synchronization among multiple power sources. The coordinated control employs a distributed algorithm based on consistency theory, where each power source adjusts its frequency control parameters according to the coupling strength with its neighboring power sources. The control law is designed in a proportional-integral form, with the proportional gain determined based on the coupling coefficient and the integral gain set according to the system damping requirements. In a strongly coupled photovoltaic-wind power pair, the frequency control loop of the photovoltaic inverter introduces feedforward compensation for the wind power frequency deviation, with the compensation coefficient equal to the coupling coefficient; similarly, the wind power converter introduces a feedforward term for the photovoltaic frequency to achieve bidirectional frequency support. The energy storage system, as the main actuator for frequency regulation, dynamically allocates frequency regulation power based on its coupling relationship with other power sources. The virtual synchronous generator (VSG) control introduces a coupling matrix to correct the virtual inertia and damping coefficients, enhancing the synchronization stability of the multi-power source system. The adaptive adjustment mechanism updates the coupling matrix based on real-time operating data, ensuring dynamic optimization of the coordinated control. In a microgrid application in an industrial park, after implementing a frequency synchronization coordination mechanism based on a resonant coupling matrix, the frequency fluctuation range of the multi-power source system decreased from ±0.3Hz to ±0.1Hz, and the frequency recovery time was shortened from 8 seconds to 3 seconds. Through accurate modeling of the coupling matrix and optimized design of coordinated control, a power source frequency synchronization coordination mechanism was generated to achieve frequency synchronization among multiple power sources and improve system stability.

[0043] Step S130: Generate a preliminary load allocation scheme based on the frequency synchronization coordination mechanism, perform a counterexample search on the preliminary load allocation scheme to generate allocation vulnerability identification data, and obtain an enhanced allocation strategy based on the allocation vulnerability identification data.

[0044] Specifically, a preliminary load allocation scheme for the microgrid is formulated using a frequency synchronization coordination mechanism. This mechanism reveals the strong coupling characteristics of photovoltaic (PV) and wind power (coupling coefficient 0.85), the frequency support capability of energy storage, and the dynamic response characteristics of each power source, providing a foundation for load allocation. The allocation principle follows a hierarchical objective of prioritizing frequency stability followed by economic efficiency, prioritizing power sources with fast frequency response and strong regulation capabilities to handle load fluctuations. Based on the coupling strength displayed by the resonant coupling matrix, strongly coupled power source groups are treated as a whole for load allocation, avoiding oscillations caused by improper allocation. Due to its fast response characteristics (response time less than 100 milliseconds), the energy storage system is designated as the main force for frequency regulation, handling high-frequency load fluctuations; PV and wind power handle the base load based on their real-time output and predicted curves; diesel generators serve as backup support, starting when renewable energy output is insufficient. The allocation algorithm employs a multi-objective optimization method, with the objective function including minimizing frequency deviation, minimizing generation cost, and minimizing grid loss. Constraints cover technical limitations such as the upper and lower limits of each power source's output, ramp rate limits, and energy storage SOC constraints. Under typical operating conditions, photovoltaic power accounts for 35% of the base load, wind power accounts for 30%, energy storage accounts for 25% of the regulating load, and diesel generators maintain 10% spinning reserve. By determining the output plans and load-bearing ratios of each power source at different times, a preliminary load allocation scheme was generated.

[0045] In some embodiments, the step of performing a counterexample search on the initial load allocation scheme to generate allocation vulnerability identification data includes: constructing an adversarial allocation mode opposite to the initial load allocation scheme; simulating a conflict scenario based on the adversarial allocation mode to obtain conflict response results; analyzing the conflict response results to generate allocation vulnerability markers; and generating allocation vulnerability identification data based on the allocation vulnerability markers.

[0046] An adversarial load allocation mode, the opposite of the initial scheme, was constructed to identify potential flaws in the original scheme through reverse thinking. The adversarial mode is designed based on the "worst-case" principle, reversing the allocation logic in the initial scheme: in the original scheme, photovoltaics prioritize daytime loads, while the adversarial mode forces photovoltaics to bear high loads during low-output periods such as cloudy days or early mornings and evenings; in the original scheme, energy storage mainly regulates high-frequency fluctuations, while the adversarial mode allows energy storage to charge and discharge at full power for extended periods, rapidly depleting its State of Charge (SOC); in the original scheme, wind and photovoltaics work together, while the adversarial mode sets them to operate in opposite phases. Extreme operating conditions are constructed, including low-probability but high-risk scenarios such as all renewable energy sources simultaneously operating at minimum output, a sudden 50% load increase, and multiple power source failures. Adversarial parameters are generated using a genetic algorithm, aiming to maximize system instability and searching for parameter combinations that expose allocation flaws. Boundary condition testing pushes various constraints to their limits, such as bringing the energy storage SOC close to 0% or 100% and causing frequent start-stop cycles of diesel generators. Timing adversarial measures disrupt the original coordination sequence, creating conflicts during the transition period of allocation switching. The adversarial mode library ultimately includes 5-10 typical adversarial scenarios, covering a variety of potential problems such as insufficient capacity, delayed response, and coordination failure, forming an adversarial allocation mode that comprehensively challenges the initial allocation scheme.

[0047] A conflict scenario simulation of a microgrid system was conducted based on an adversarial allocation mode, recording the system response under various extreme conditions. The simulation platform was built using MATLAB / Simulink, including detailed dynamic models of various power sources, control system models, and grid models. The simulation step size was set to 50 microseconds to capture rapid transient processes. During the execution of the conflict simulation, load was forcibly allocated according to the adversarial mode, while monitoring the dynamic changes of key indicators such as frequency, voltage, and power balance. In the scenario of "all renewable energy output at low levels + sudden load increase," the frequency dropped to 49.2Hz within 2 seconds, triggering the low-frequency protection boundary. In the scenario of "energy storage SOC depletion + wind power fluctuations," the frequency exhibited continuous oscillations of ±0.8Hz, exceeding the allowable range. In the scenario of "rapid photovoltaic fluctuations + wind power reversal," a risk of voltage collapse due to power deficit occurred. The acquisition of response data covered electrical parameters, control variables, and protection action signals, forming a multi-dimensional system state record. The capture of abnormal events included protection device actions, control saturation, and communication delay exceeding limits; these events directly reflected the vulnerabilities of the allocation scheme. The conflict response results were ultimately obtained by accumulating multi-dimensional system state records and abnormal event data through conflict scenario simulation.

[0048] A thorough analysis of the conflict response results was conducted to identify and label the vulnerability characteristics of the allocation scheme. The analytical framework included three dimensions: stability assessment, robustness assessment, and resilience assessment. Stability analysis used the Lyapunov exponent to determine the system's stability margin under disturbances; a positive exponent indicated system instability and was labeled "stability vulnerable." Robustness analysis assessed the system's sensitivity to parameter changes, identifying highly sensitive parameter combinations using a sensitivity matrix; combinations with sensitivity exceeding a set threshold were labeled "robust vulnerability." Resilience analysis measured the recovery time and path after a fault; recovery times exceeding 10 seconds or the occurrence of recovery oscillations were labeled "resilient vulnerability." Vulnerability quantification employed a risk scoring mechanism, comprehensively considering the probability of occurrence and the severity of impact; risk scores exceeding 0.7 were classified as high-risk vulnerabilities. Correlation analysis identified causal relationships between vulnerabilities, such as insufficient energy storage capacity leading to decreased frequency regulation capability, which in turn triggered system oscillations. Temporal characteristic analysis revealed that some vulnerabilities only occurred during specific time periods, such as the power balance vulnerability during the early morning photovoltaic startup phase. Through multi-dimensional vulnerability analysis and risk quantification assessment, allocation vulnerability labels were generated.

[0049] Based on vulnerability markers, various allocation defects are systematically organized and categorized to generate structured allocation vulnerability identification data. The vulnerability classification adopts a hierarchical architecture: the primary classification includes capacity vulnerabilities, time-series vulnerabilities, coordination vulnerabilities, and extreme vulnerabilities; the secondary classification is further refined, such as capacity vulnerabilities including subcategories like insufficient total capacity, insufficient reserve capacity, and insufficient regulation capacity. Each vulnerability entry includes multi-dimensional information such as vulnerability description, triggering conditions, impact assessment, and related vulnerabilities. Priority ranking is determined based on risk scores and remediation difficulty, with high-risk and easily remediatable vulnerabilities being prioritized. Quantitative descriptions of vulnerabilities use standardized templates to clearly define the numerical boundaries of vulnerabilities, such as "when photovoltaic output is 20% lower than the installed capacity and the load is 30% higher than the average, a power gap of 150kW occurs." The relationships between vulnerabilities are represented by a directed graph, revealing the propagation path and cumulative effect of vulnerabilities. Statistical analysis shows that time-series vulnerabilities account for the highest proportion (35%), reflecting the complexity of coordination control; although capacity vulnerabilities account for a lower proportion (20%), their impact is the most severe. The structured storage of vulnerability identification data uses JSON format, facilitating subsequent automated processing and strategy generation.

[0050] Based on the various defects exposed in the allocation vulnerability identification data, targeted improvement measures were designed to form a strengthened allocation strategy. The strategy formulation adopted a "vulnerability-countermeasure" mapping method, with each type of vulnerability corresponding to specific strengthening measures. For capacity-related vulnerabilities, the strengthening strategy includes: dynamically adjusting the reserve capacity ratio and adaptively determining reserve demand based on new energy prediction errors; establishing a multi-level reserve system with tiered configurations of rapid reserve (energy storage), spinning reserve (diesel engines), and cold reserve (interruptible loads); optimizing energy storage charging and discharging strategies, pre-charging during non-critical periods to ensure sufficient adjustment capacity during critical periods. For timing-related vulnerabilities, the strengthening strategy introduces predictive control, starting backup power 5-15 minutes in advance to avoid power gaps; designing a smooth switching mechanism, transitioning through a ramp function when switching between different operating modes; and establishing a priority queue to ensure the power supply sequence of critical loads. For coordination-related vulnerabilities, the strengthening strategy employs a distributed consensus algorithm to ensure the synchronization of multiple power source actions; introduces virtual inertia control to enhance the system's frequency support capability; and designs a conflict resolution mechanism for automatic arbitration when multiple power source commands conflict. For extreme vulnerabilities, the reinforcement strategies include emergency control plans, pre-setting emergency response procedures for extreme operating conditions; building islanded operation capabilities to ensure the autonomous operation of the local power grid; and designing a black-start scheme for rapid recovery after system crashes. The reinforcement measures for various vulnerabilities were integrated and optimized through system integration, ultimately resulting in a reinforcement allocation strategy.

[0051] Step S140: Based on the enhanced allocation strategy, obtain the load-bearing capacity assessment data of each power source, match and analyze the load demand data with the load-bearing capacity assessment data to generate a load transmission topology, and use the load transmission topology to obtain the pressure distribution data among each power source.

[0052] Specifically, an enhanced allocation strategy is used to comprehensively assess the load-bearing capacity of each power source in the microgrid. This strategy clarifies that energy storage systems need to maintain rapid standby capability, photovoltaic and wind power need to account for the impact of forecast errors, and diesel generators need to maintain a moderate spinning reserve. These requirements directly affect the actual available capacity of each power source. The assessment framework employs a multi-dimensional analysis method. The technical dimension assessment includes hard indicators such as rated capacity, adjustable capacity, ramp rate, and response time; the operational dimension assessment covers real-time factors such as current output level, remaining regulation margin, cumulative operating time, and maintenance status; and the reliability dimension assessment considers statistical indicators such as historical failure rate, mean time between failures (MTBF), and recovery time. The load-bearing capacity assessment of photovoltaic systems requires combining irradiance prediction and temperature correction, and adjusting for actual environmental conditions based on the rated capacity under standard test conditions. Wind power system assessment incorporates wind speed-power curves and turbulence intensity, considering key nodes such as cut-in wind speed, rated wind speed, and cut-out wind speed. Energy storage system assessment focuses on the current State of Charge (SOC) and charging / discharging power limitations, establishing a mapping relationship between SOC and power load-bearing capacity. Diesel generator assessment includes factors such as cold start time, hot standby status, and fuel reserves. A dynamic assessment mechanism updates every 15 minutes to ensure the timeliness of load-bearing capacity data. The comprehensive assessment results, after standardization and real-time updates, yield load-bearing capacity assessment data for each power source.

[0053] In some embodiments, the step of matching and analyzing the load demand data with the load capacity assessment data to generate a load transmission topology includes: performing difference analysis on the load demand data and the load capacity assessment data to obtain load gap data; constructing load transfer paths between power sources based on the load gap data; analyzing each transfer path to generate a conduction impedance coefficient; and using the conduction impedance coefficient to generate a load transmission topology.

[0054] Real-time load demand data is compared point-by-point with the load-bearing capacity assessment data of each power source to calculate the supply-demand difference and identify load gaps. The difference analysis employs a time-series alignment method to ensure that demand data and load-bearing capacity data are compared at the same time point. Analysis under normal operating conditions shows that during periods of sufficient sunlight in the daytime, photovoltaic (PV) capacity exceeds allocated demand, resulting in a positive surplus; in the evening, PV output decreases while load demand remains high, resulting in a negative gap. Gaps are classified into instantaneous gaps and persistent gaps. Instantaneous gaps are caused by short-term load spikes or fluctuations in power output, typically lasting less than 5 minutes; persistent gaps are caused by insufficient power capacity or persistently high load, requiring the use of backup resources. Gaps are quantified not only by power output but also by energy accumulation; the energy gap is obtained by integrating the power gap over time. Spatial distribution analysis identifies the distribution characteristics of gaps at different nodes, with gaps being more severe at nodes closer to the load center. Temporal distribution analysis reveals the periodicity of the gaps, with the gaps being most prominent during the morning and evening peak hours on weekdays. The severity of the power shortage is classified into three levels: minor shortages can be resolved through inter-power source adjustments, moderate shortages require the activation of backup power, and severe shortages may trigger load control. Through systematic analysis and classification of the supply-demand difference, load shortage data was obtained.

[0055] Based on the identified load gap data, load transfer paths between power sources are constructed to achieve supply and demand rebalancing. Path construction follows the principles of proximity transfer, minimum loss, and reliability priority. When a photovoltaic (PV) gap occurs, the preferred path is to obtain support from energy storage systems, which offer fast response and high transfer efficiency; the second preferred path is to adjust from wind power systems, but the output stability of wind power itself needs to be considered; the alternative path is to start diesel generators, which, although slower in response, offer high reliability. Path determination considers electrical distance rather than physical distance, and the electrical coupling strength between power sources is calculated using an impedance matrix. A multi-path parallel mechanism allows load transfer from multiple power sources simultaneously, improving the reliability and flexibility of the transfer. Dynamic path adjustment optimizes the transfer scheme based on real-time operating status, automatically switching to a backup path when a path becomes congested. Path capacity limits ensure that the transfer process does not exceed the carrying capacity of lines and equipment. Priority settings ensure that the transfer paths for critical loads are always unobstructed. The bidirectional design of the paths not only supports gap filling but also surplus absorption, forming a flexible energy mutual assistance network. The topology of the transfer paths exhibits a mesh-like characteristic, with at least one feasible path between any two power sources. Through gap data-driven path planning and network construction, load transfer paths between power sources were formed.

[0056] Impedance characteristic analysis was performed on load transfer paths between power sources to quantify the conduction difficulty of each path. Based on the physical characteristics and operating parameters of the transfer paths, the conduction impedance coefficient was calculated using a comprehensive impedance model: Z_ij = α·R_ij + β·X_ij + γ·C_ij, where Z_ij represents the conduction impedance from power source i to power source j; R_ij is the electrical impedance component, reflecting the line resistance and transmission loss of the path; X_ij is the control impedance component, reflecting the response delay and adjustment complexity of the path; C_ij is the economic impedance component, quantifying the transmission cost of the path; and α, β, and γ are weighting coefficients. For the photovoltaic-energy storage transfer path, due to the short physical distance and simple control, its electrical impedance R_pv-es is relatively small, and the control impedance X_pv-es is also low due to the fast response characteristics of energy storage. The wind power-energy storage path has a relatively large R_wind-es due to the longer distance; the control impedance X_diesel of diesel engine-related paths is generally high, reflecting the complexity of the start-up and shutdown process. The weighting coefficients are adjusted according to the system operating mode. During normal operation, α=0.3, β=0.5, and γ=0.2. In emergency situations, β is increased to 0.7 to prioritize response speed. The time-varying characteristics of impedance are reflected through online updates; changes in energy storage SOC affect the impedance values ​​of relevant paths. Per-unit processing uses the system average impedance as a benchmark to ensure the comparability of different types of impedance. Based on a comprehensive evaluation of the transfer paths and parameter calculations, the conduction impedance coefficient is finally generated.

[0057] A topology reflecting the load conduction relationship between power sources is constructed using the conduction impedance coefficient. The topology construction employs a weighted directed graph model, where nodes represent various power sources and loads, edges represent transfer paths, and edge weights are equal to the reciprocal of the conduction impedance coefficient, representing conduction capacity. The hierarchical design of the topology divides power sources into a main layer, a support layer, and a backup layer. The main layer includes photovoltaic and wind power, undertaking the base load; the support layer primarily consists of energy storage, providing rapid regulation; and the backup layer includes diesel generators as emergency backup. Adjacency matrix representation facilitates computer processing and path search, with matrix elements directly reflecting the connection relationships and conduction strength between nodes. A shortest path algorithm is used to find the optimal conduction path, comprehensively considering both minimum impedance and maximum reliability. The dynamic characteristics of the topology are reflected through a time-varying adjacency matrix, where the connection strength between nodes adjusts accordingly with changes in system operating status. Critical node identification is achieved by calculating node degree centrality, betweenness centrality, and other indicators; energy storage systems are typically critical nodes. Topology optimization improves reliability by increasing redundant paths and reducing the dependence of critical nodes. Through quantitative characterization of conduction capacity and network organization, a load conduction topology is generated.

[0058] Power flow analysis and load distribution calculations are performed using the generated load conduction topology to obtain the pressure state of each power source. The load conduction topology provides complete network structure and conduction capacity information. Based on the node connection relationships in the topology, a set of power balance equations is established, and the voltage and power of each node are solved using the Newton-Raphson method. The pressure index is calculated as P_i = (L_i / C_i) × (1 + σ_i), where P_i is the pressure index of power source i, L_i is the actual load borne by the power source calculated from the topology, C_i is the rated capacity, and σ_i is the dynamic adjustment load factor. The pressure conduction in the topology network follows the law ΔP_j = Σ(K_ij × (P_i - P_j)), where the conduction factor K_ij is directly taken from the edge weights of the load conduction topology. By tracing the pressure conduction process through the topology path, when the photovoltaic pressure is too high, according to the topology connection relationships, the pressure is preferentially conducted to energy storage, and after the energy storage is saturated, it continues to diffuse to wind power and diesel engines. The pressure distribution calculation considers the hierarchical structure of the topology. The main layer bears steady-state pressure, the support layer absorbs dynamic pressure, and the backup layer remains in a low-pressure standby state. Time evolution analysis, based on the dynamic characteristics of the topology, tracks the migration process of pressure in the network. Through power flow calculation and pressure transmission analysis under topological constraints, pressure distribution data among power sources are obtained.

[0059] Step S150: Generate a boundary active detection instruction set based on pressure distribution data, use the boundary active detection instruction set to drive each power source to gradually approach the operating boundary to obtain boundary response data, and perform extreme value analysis on the boundary response data to generate the load-bearing limit parameters of each power source.

[0060] Specifically, based on the current pressure status and pressure gradient of each power source as shown by the pressure distribution data, an active detection strategy is designed to identify the operating boundaries of each power source. The pressure distribution data indicates that the pressure index of the energy storage system fluctuates drastically, requiring focused detection of its charge-discharge switching boundaries; photovoltaic systems have significant detection potential during the low pressure period at midday, but require cautious detection during the high pressure periods in the morning and evening; wind power pressure is related to wind speed, and detection needs to consider wind speed prediction; diesel generators have lower pressure and possess significant potential for boundary detection. The detection command design adopts a progressive strategy to avoid system instability caused by aggressive detection. Command parameters include detection direction, step size, duration, and backoff conditions. The detection direction is determined based on the current pressure status: low-pressure power sources detect in the direction of increasing output, while high-pressure power sources detect the lower boundary of decreasing output. The step size design uses an adaptive algorithm, with an initial step size of 2% of the rated capacity, dynamically adjusted based on response sensitivity. The duration is matched to the system time constant to ensure that steady state is reached before proceeding to the next detection step. Backoff conditions include safety constraints such as frequency deviation exceeding limits, voltage exceeding limits, and excessively high temperature. The arrangement of the detection sequence considers the mutual influence between power sources to avoid the superposition effect caused by simultaneous detection of multiple power sources. The time window was chosen during a period of stable load to minimize the impact of external disturbances on the detection. A boundary active detection command set was generated through pressure-state-driven detection strategy formulation and parameter optimization.

[0061] In some embodiments, the step of using the boundary active detection instruction set to drive each power source to gradually approach the operating boundary and obtain boundary response data includes: increasing the output of each power source in a step-by-step manner based on the boundary active detection instruction set; monitoring the power source status and obtaining response characteristic data during the step-by-step output increase; performing nonlinear analysis on the response characteristic data to identify abrupt change points and recording abrupt change point parameters; and generating boundary response data based on the abrupt change point parameters.

[0062] The system executes step-by-step increment commands from the boundary active detection command set, driving each power source to gradually increase its output level. The step-by-step increment uses a step control method, with each increment determined according to the detection command set. For example, the energy storage system increases charging / discharging power by 5kW per step, the photovoltaic system increases by 3% per step by adjusting the MPPT bias, the wind power system increases output by 2% per step by adjusting the pitch angle, and the diesel generator increases load by 4% per step through the speed governor. The step interval is set to 2 minutes to ensure the system reaches a new steady-state equilibrium. The control during the increment process uses closed-loop feedback, monitoring frequency and voltage deviations in real time. When the deviation exceeds a set threshold, the increment is automatically slowed down or paused. For coordinated detection of multiple power sources, a polling mechanism is used. After one power source completes one detection step, other power sources maintain their current state to avoid mutual interference. The detection range starts from the current operating point and gradually approaches the theoretical limit: energy storage is detected until the SOC approaches its upper or lower limit, photovoltaic system until the inverter capacity limit, wind power system until the rated power or cut-off wind speed, and diesel generator system until the rated power or temperature limit. The anomaly handling mechanism operates continuously during the detection process, immediately halting detection and reverting to a safe operating point upon the occurrence of any anomaly. Ultimately, through precise execution and safety assurance of step-by-step control, a gradual increase in the output of each power source is achieved.

[0063] Throughout the step-by-step increase in power output, a comprehensive monitoring system is deployed to collect real-time status parameters of each power source. Monitoring covers multiple dimensions, including electrical, thermal, and mechanical parameters. Electrical parameters include output power, voltage, current, power factor, and harmonic content, with a sampling frequency increased to 1kHz to capture rapid changes. Thermal parameters monitor the temperature of key components such as energy storage batteries, inverter radiators, and diesel engine cylinder blocks, with the rate of temperature rise serving as a crucial monitoring indicator. Mechanical parameters primarily target rotating equipment, including wind turbine speed, vibration amplitude, and diesel engine speed fluctuations. Data acquisition synchronization is ensured using a GPS clock, and all measurement points have a unified timestamp. Anomaly detection algorithms operate in real-time, employing various methods such as limit exceedance detection, rate of change detection, and trend prediction. Data storage utilizes a circular buffer mechanism, retaining the most recent 30 minutes of high-precision data for analysis. Feature extraction focuses on boundary-related sensitive parameters, such as efficiency inflection points, temperature rise acceleration points, and vibration surge points. Nonlinear characteristics of the response are identified using phase space reconstruction methods, revealing the dynamic changes of the system near its boundaries. Through comprehensive monitoring and feature extraction, a wealth of response characteristic data was finally obtained.

[0064] Nonlinear analysis was performed on the collected response characteristic data to identify abrupt changes in system behavior. Multiple methods were used to cross-verify the nonlinear analysis, ensuring the accuracy of abrupt change identification. Bifurcation analysis, by plotting parameter-state bifurcation diagrams, identified critical points where system stability changed. When the photovoltaic inverter output approached its capacity limit, the voltage-power curve exhibited a reversal, indicating the attainment of the static stability boundary. Phase plane analysis projected multidimensional state variables onto a two-dimensional plane to observe the topological changes in the system trajectory. When the energy storage system approached its SOC limit, the charge-discharge switching trajectory changed from a smooth transition to a sharp jump. Lyapunov exponent calculation assessed the system's sensitivity to initial conditions; a change from negative to positive exponent indicates the system has entered a chaotic state, a characteristic exhibited by wind turbines when approaching the cut-off wind speed. Catastrophe theory applied the cusp catastrophe model to describe the abrupt change in output power with control parameters, identifying the power collapse point of the diesel generator under overload. The records of abrupt change point parameters included complete information such as the power level at which the abrupt change occurred, the corresponding control parameter values, environmental conditions, and precursor characteristics. In multi-source systems, it is also necessary to identify coupling abrupt changes, that is, a sudden change in one source triggers a chain reaction in other sources. Through system analysis of nonlinear characteristics and critical point identification, key abrupt change parameters were recorded.

[0065] Based on the identified and recorded mutation point parameters, a response dataset reflecting the boundary characteristics of each power source is compiled. The structured organization of boundary response data includes two categories: static boundaries and dynamic boundaries. Static boundaries reflect steady-state operating limits. For energy storage systems, static boundaries include maximum continuous charge / discharge power, available SOC range, and temperature limits. For photovoltaic systems, static boundaries are determined by inverter capacity, MPPT range, and voltage limits. For wind power, static boundaries include minimum output at the cut-in wind speed, rated output at the rated wind speed, and shutdown boundaries at the cut-out wind speed. For diesel engines, static boundaries mainly consist of rated power and minimum stable operating power. Dynamic boundaries describe limitations during transient processes, including maximum power change rate, frequency response limits, and fault ride-through capability. The time-varying characteristics of the boundaries are represented by parametric curves, such as the curve of available energy storage power changing with SOC and the derating curve of photovoltaic output with temperature. Safety margin quantification is based on the distance from the mutation point to the operating point, providing a reference for subsequent control strategies. The interrelationships between boundaries are described by an correlation matrix; tightening of one boundary may lead to relaxation of others. The impact of environmental factors on the boundaries is reflected by correction coefficients; under high-temperature environments, the boundaries of each power source generally contract. After systematically organizing the mutation point parameters and extracting boundary features, the boundary response data was finally generated.

[0066] Extreme value analysis was performed on the boundary response data to determine the load-bearing capacity limits of each power source under different operating conditions. The extreme value analysis employed statistical extreme value theory, fitting the boundary data with the generalized extreme value distribution (GEV) to obtain the limit values ​​at different confidence levels. For energy storage systems, the analysis showed that under standard conditions, the maximum continuous discharge power was 120% of the rated power, but the duration did not exceed 15 minutes; the maximum instantaneous power could reach 150% of the rated value, with a duration not exceeding 1 second; the usable SOC range was 10%-90%, which could be extended to 5%-95% in extreme cases, but this would affect battery life. The load-bearing limit of photovoltaic systems was mainly limited by inverter capacity, with a short-term overload capacity of up to 110% of the rated power; under low irradiance conditions, the MPPT efficiency decreased, resulting in the actual usable power being only 85% of the theoretical value. Wind turbines maintained rated output above rated wind speed through pitch angle control, but could output 105% of the rated power for short periods under gusts; the minimum stable output at low wind speeds was 5% of the rated power. Diesel generators possess strong overload capacity, capable of continuous operation at 110% rated power for 1 hour and 120% for 10 minutes; however, increased fuel consumption and maintenance costs must be considered. Probability distribution analysis of limiting parameters reveals the reliability level of load-bearing capacity; P95 represents the load-bearing capacity achievable with a 95% probability. Time-series correlation analysis reveals a memory effect in some limiting parameters, where high-load operation in the early stages reduces subsequent load-bearing capacity. Through statistical analysis and probabilistic assessment of boundary data, the load-bearing limiting parameters for each power source were obtained.

[0067] Step S160: Based on the load limit parameters, a unified timing adjustment benchmark is set, and the timing adjustment benchmark is used to synchronize the power output adjustment process of each power source to obtain coordinated timing data. Timing synchronization control parameters are generated based on the coordinated timing data.

[0068] Specifically, a unified timing adjustment benchmark for the microgrid is established based on the load-bearing limit parameters of each power source. These load-bearing limit parameters indicate that energy storage systems have millisecond-level response capabilities but limited duration; the regulation speed of photovoltaic and wind power is limited by physical processes; and diesel generators have longer start-up times but better stability. These differentiated characteristics need to be coordinated through a unified benchmark. The timing benchmark design adopts a layered time scale: the fast adjustment layer uses a 100-millisecond base period, primarily responding to instantaneous power fluctuations by the energy storage system; the medium-speed adjustment layer uses a 1-second base period, where photovoltaic MPPT and wind turbine pitch angle adjustments are handled; and the slow adjustment layer uses a 10-second base period, where diesel generator start-up, shutdown, and power ramp-up are completed. The reference clock uses a GPS-synchronized clock source with microsecond-level accuracy, ensuring time synchronization of all power controllers. The adjustment margin is set based on the load-bearing limit parameters: energy storage reserves a 20% power margin for fast response, photovoltaic and wind power maintain a 5% adjustment margin, and diesel generators maintain 15% reserve capacity. Timing constraints include minimum action interval, maximum adjustment rate, and coordination waiting time to prevent equipment damage caused by frequent adjustments. The priority mechanism is determined based on response speed and adjustment cost, with energy storage prioritizing high-frequency components, new energy sources handling mid-frequency regulation, and diesel engines providing low-frequency support. A unified timing adjustment benchmark is established through time-scale hierarchical and priority settings under limiting parameter constraints.

[0069] In some embodiments, the step of synchronizing the power output adjustment process of each power source using the timing adjustment benchmark to obtain coordinated timing data includes: generating power output adjustment instructions for each power source based on the timing adjustment benchmark; timing-marking each power output adjustment instruction to obtain timing identifier data; synchronizing and arranging the timing identifier data to generate a coordinated execution sequence; and performing timing peak-shifting optimization on the coordinated execution sequence to obtain coordinated timing data.

[0070] Based on the hierarchical cycle and priority rules defined in the unified timing adjustment benchmark, corresponding output adjustment commands are generated for each power source. The command generation process first decomposes the total regulation demand, breaking it down into different frequency components through spectrum analysis. High-frequency components above 100Hz are allocated to the energy storage system, mid-frequency components (1-100Hz) are allocated to photovoltaic and wind power, and low-frequency components below 1Hz are allocated to diesel generators. The adjustment commands for the energy storage system adopt a direct power control mode, with the command format including parameters such as target power value, execution time, and duration. Photovoltaic adjustment commands are implemented through MPPT reference voltage bias, and include voltage adjustment amount and adjustment rate limit. Wind power adjustment commands are converted into pitch angle control parameters, taking into account mechanical response delay. Diesel generator adjustment commands include start / stop commands and power setpoints, along with preheating and synchronization requirements. The timing of command generation is determined by a prediction algorithm. Energy storage commands are generated in real time, new energy commands are generated 1-2 seconds in advance, and diesel generator commands are generated 30 seconds in advance to allow for start-up time. The calculation of command amplitude considers the distance from the current operating point to the limit, ensuring that the carrying capacity is not exceeded. A redundant command mechanism generates two sets of commands (primary and backup) for critical regulation tasks to improve reliability. After demand decomposition and parameter calculation under the timing reference framework, power output adjustment commands for each power source are generated.

[0071] The generated power output adjustment commands for each power source are uniformly time-marked to ensure precise time positioning. The time-marking employs a dual identification method using absolute timestamps and relative timing. The absolute timestamp is based on a GPS synchronized clock, accurate to the microsecond level, and formatted as "year-month-day hour:minute:second.microsecond". The relative timing uses the start point of the scheduling cycle as a reference to identify the execution position of the command within the cycle. The marking process considers the response delay characteristics of each power source. The energy storage system has a response delay of approximately 50 milliseconds, which is pre-compensated based on the command timestamp; the photovoltaic MPPT response delay is 200-500 milliseconds, dynamically compensated based on real-time measurements; the wind turbine pitch angle adjustment delay is 1-3 seconds, and the diesel engine start-up delay is 30-60 seconds; these delays are all reflected in the time-marking. Priority marking is performed synchronously with time-marking, with higher-priority commands receiving priority execution. In case of conflicts at the same time, lower-priority commands are automatically postponed. A consistency check of the markings ensures that there are no timing conflicts or logical contradictions, such as when the same power source receives opposing adjustment commands at the same time. A version control mechanism assigns a unique identifier to each command, supporting command updates and revocations. The structured storage of tagged data uses a time-series database, supporting fast time-range query and sorting operations. Time-series identification data is obtained through precise time positioning and attribute labeling of instructions.

[0072] The power supply adjustment commands with timing identifiers are uniformly arranged to form a coordinated execution sequence on the timeline. Synchronization arrangement first sorts all commands in ascending order of execution timestamps, forming a global command queue. The time intervals between adjacent commands in the queue are analyzed to identify potential execution conflicts. When the execution times of multiple commands overlap, the execution order is adjusted according to priority rules. High-frequency regulation commands for energy storage systems maintain their original timing, serving as the first line of defense for rapid response; medium-frequency regulation commands for photovoltaic and wind power are inserted without affecting energy storage regulation; low-frequency support commands for diesel engines are scheduled for execution after the load change trend is clear. The evaluation of parallel execution capability determines which commands can be executed simultaneously; electrically independent power supply commands can be executed in parallel, while those with power coupling need to be executed serially. Synchronization points are set every second, ensuring all power supplies are aligned at the synchronization point for coordinated control. The integrity check of the execution sequence ensures that all necessary regulation needs are covered, with no regulation gaps. The executability verification of the sequence is completed through simulation pre-running to check for resource conflicts or timing lock-ups. The generation of backup sequences considers the failure of the main sequence, providing a degraded operation plan. Based on the orderly organization and conflict resolution of the time sequence identifier data, a coordinated execution sequence is finally obtained.

[0073] The coordinated execution sequence is optimized for timing-based peak shifting to avoid system shocks caused by simultaneous large-scale adjustments from multiple power sources. The objective function of peak shifting optimization is set to minimize the peak value of the system power change rate, with constraints including the completion of adjustment tasks, not exceeding the capacity limits of each power source, and meeting real-time balance requirements. The optimization algorithm uses dynamic programming, discretizing continuous time into 100-millisecond time slices and optimizing the timing of command execution within each time slice. Rapid adjustment commands from the energy storage system are given priority, but the number of consecutive large-scale adjustments is limited, with at least a 500-millisecond interval between two large-scale adjustments. Solar and wind power adjustment commands are executed separately; when solar power increases output, wind power remains stable or makes small adjustments. Diesel generator start-ups and shutdowns are scheduled during off-peak periods to reduce system disturbance. Power change rate smoothing is achieved by inserting transition commands, converting step changes into ramp changes. The peak shifting interval is determined based on the system time constant, ensuring that the next adjustment is executed only after the previous one has stabilized. Multi-objective optimization also considers economic costs, prioritizing the use of low-cost adjustment resources while meeting technical constraints. After peak shifting and smoothing of the execution sequence, coordinated timing data is generated.

[0074] Based on the coordinated timing data, timing synchronization parameters that can be directly used in the control system are generated. First, key time nodes are extracted from the coordinated timing data, including the start time, peak time, and stabilization time of each power source, forming a time node sequence. The synchronization control cycle parameter is determined based on the greatest common divisor principle: a cycle of 100 milliseconds for the fast control layer, 1 second for the medium-speed control layer, and 10 seconds for the slow control layer ensures coordination at each level. Phase relationship parameters describe the relative timing of each power source's actions: the energy storage leading response phase is set to 0°, the photovoltaic lagging phase to 30°, the wind power lagging phase to 45°, and the diesel engine lagging phase to 90°, achieving natural peak shifting through phase difference. Trigger condition parameters define the thresholds for each power source's start-up regulation, including frequency deviation thresholds, power deficit thresholds, and SOC thresholds, with differentiated thresholds for different power sources to achieve tiered responses. Coordination waiting time parameters specify the interlocking relationships between power sources; for example, diesel engine start-up requires waiting for the energy storage response to stabilize, with a waiting time set to 2 seconds. Communication synchronization parameters include data refresh rate, transmission delay compensation, and packet loss retransmission strategies to ensure reliable transmission of control commands. The parameters are encoded in a standardized format, supporting parsing by control systems from different manufacturers. The completeness of the parameter set includes both normal operation and fault handling schemes. Through parameterized transformation and standardized encapsulation of coordinated timing data, timing synchronization control parameters are obtained.

[0075] Step S170: Generate the final load distribution control instruction set based on the timing synchronization control parameters. Each power source executes the control instruction set to coordinate output and obtain actual operating status data, thereby completing the microgrid load distribution control.

[0076] Specifically, the final control commands for each power source are generated using timing synchronization control parameters. These parameters define the hierarchical control cycles: 100 milliseconds for the fast layer, 1 second for the medium-speed layer, and 10 seconds for the slow layer; the phase difference configurations: 0° for energy storage, 30° for photovoltaic, 45° for wind power, and 90° for diesel engines; and the differentiated trigger thresholds for each power source. These parameters are directly translated into the timing arrangement of the control commands. Command generation employs a template-based method, with each type of power source corresponding to a specific command template. The energy storage system command template includes fields such as charge / discharge power settings, power change rate limits, and SOC protection boundaries; the photovoltaic command template covers MPPT operating point adjustment, active / reactive power ratio, and inverter protection parameters; the wind power command template includes elements such as pitch angle settings, speed control targets, and power limit curves; and the diesel engine command template contains items such as start / stop control, load rate settings, and grid synchronization parameters. The timestamps of the commands are calculated based on the phase relationship of the synchronization control parameters, ensuring that each power source executes actions according to the designed timing sequence. The triggering logic converts threshold parameters into conditional judgment statements, embedding them into the commands to achieve autonomous triggering. The instructions are encoded using the IEC 61850 standard format to ensure compatibility with equipment from various manufacturers. A verification mechanism checks the integrity and consistency of the generated instructions to prevent logical conflicts or parameter exceeding limits. The instruction set is organized into a queue structure according to execution sequence, supporting real-time scheduling and dynamic updates. The parsing, conversion, and template encapsulation of synchronization parameters generate the final load distribution control instruction set.

[0077] Each power controller receives and parses the final load distribution control command set and executes coordinated output actions according to a predetermined sequence. The execution process strictly follows the time arrangement in the command set. The energy storage system responds first, adjusting its charging and discharging power within milliseconds to absorb or release power to maintain system balance. The photovoltaic system begins adjustment after a 30° phase delay, smoothly increasing or decreasing output by changing the MPPT operating point. The wind power system operates after a 45° phase delay, with the blade angle controller adjusting the blade angle according to the command. The diesel generator starts with a 90° phase delay when necessary and connects to the grid after a synchronization process. The execution monitoring system tracks the operational status of each power source in real time, recording the deviation between the actual execution time and the planned time. Closed-loop control ensures accurate command execution, automatically compensating for deviations when detected. The coordination mechanism handles mutual influences during execution, such as voltage fluctuations caused by rapid energy storage response interfering with the photovoltaic MPPT, which is eliminated through feedforward compensation. The anomaly handling process handles unexpected situations such as equipment failure or communication interruption, automatically switching to standby control mode. Data acquisition during execution includes electrical parameters such as power, voltage, current, and frequency, as well as operating parameters such as temperature, speed, and SOC. The data acquisition frequency is set according to the variation characteristics of different parameters, with fast variables sampled at 1kHz and slow variables sampled at 1Hz. Based on chaotic exploration driven by entropy increase disturbances, coupled control with frequency synchronization coordination, vulnerability repair through counterexample search, boundary identification through active detection, and coordinated execution with time synchronization, intelligent load distribution control of multiple power sources in the microgrid is achieved, moving from disorder to order and from independence to collaboration. Each power source operates stably according to the optimized control strategy and generates actual operating status data, thus completing the microgrid load distribution control.

[0078] To implement the microgrid load distribution control method based on entropy increase disturbance corresponding to the above method embodiments, and to achieve the corresponding functions and technical effects. See Figure 2 , Figure 2 This diagram illustrates a structural block diagram of a microgrid load distribution control device 200 based on entropy increase disturbance provided in an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The microgrid load distribution control device 200 based on entropy increase disturbance provided in this embodiment includes: The data acquisition module 201 is used to acquire real-time power output data and load demand data of the power sources in the microgrid, perform chaotic measurement analysis on the real-time power output data to generate system entropy parameters, and construct an entropy increase disturbance excitation mechanism based on the system entropy parameters. Chaotic coordination module 202 is used to generate chaotic state operation data by randomly perturbing the output of each power source based on the entropy increase perturbation excitation mechanism, extract the inherent frequency characteristics of each power source from the chaotic state operation data to obtain the resonance coupling matrix, and use the resonance coupling matrix to generate a frequency synchronization coordination mechanism between power sources. The strategy optimization module 203 is used to generate a preliminary load allocation scheme based on the frequency synchronization coordination mechanism, perform counterexample search on the preliminary load allocation scheme to generate allocation vulnerability identification data, and obtain an enhanced allocation strategy based on the allocation vulnerability identification data. The pressure analysis module 204 is used to obtain load-bearing capacity assessment data of each power source based on the enhanced allocation strategy, match and analyze the load demand data with the load-bearing capacity assessment data to generate a load transmission topology, and use the load transmission topology to obtain pressure distribution data among each power source. Boundary detection module 205 is used to generate a boundary active detection instruction set based on the pressure distribution data, use the boundary active detection instruction set to drive each power supply to gradually approach the operating boundary to obtain boundary response data, and perform extreme value analysis on the boundary response data to generate the bearing limit parameters of each power supply. The timing control module 206 is used to set a unified timing adjustment benchmark based on the load limit parameter, use the timing adjustment benchmark to synchronize the power output adjustment process of each power source to obtain coordinated timing data, and generate timing synchronization control parameters based on the coordinated timing data. The instruction execution module 207 is used to generate a final load distribution control instruction set based on the timing synchronization control parameters. Each power source executes the control instruction set to coordinate output and obtain actual operating status data, thereby completing the microgrid load distribution control.

[0079] The aforementioned microgrid load distribution control device 200 based on entropy increase disturbance can implement the microgrid load distribution control method based on entropy increase disturbance described in the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0080] like Figure 3 As shown, the third embodiment of the present invention also provides a computer device, including a memory 301, a processor 302, and a computer program stored in the memory 301 and executable on the processor 302, characterized in that the processor 302 executes the program to implement the steps of the microgrid load distribution control method based on entropy increase disturbance described in the first embodiment of the present invention.

[0081] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

[0082] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A microgrid load distribution control method based on entropy increase disturbance, characterized in that, include: Real-time power output data and load demand data of power sources in the microgrid are acquired, and chaotic measurement analysis is performed on the real-time power output data to generate system entropy parameters. Based on the system entropy parameters, an entropy increase disturbance excitation mechanism is constructed. Based on the entropy increase perturbation excitation mechanism, randomized power output perturbation is performed on each power source to generate chaotic state operation data. The inherent frequency characteristics of each power source are extracted from the chaotic state operation data to obtain the resonance coupling matrix. The resonance coupling matrix is ​​used to generate a frequency synchronization coordination mechanism between power sources. A preliminary load allocation scheme is generated based on the frequency synchronization coordination mechanism. Counterexample search is performed on the preliminary load allocation scheme to generate allocation vulnerability identification data. An enhanced allocation strategy is obtained based on the allocation vulnerability identification data. Based on the enhanced allocation strategy, load-bearing capacity assessment data of each power source is obtained. The load demand data and the load-bearing capacity assessment data are matched and analyzed to generate a load transmission topology. The load transmission topology is then used to obtain pressure distribution data among the power sources. Based on the pressure distribution data, a boundary active detection instruction set is generated. The boundary active detection instruction set is used to drive each power supply to gradually approach the operating boundary to obtain boundary response data. Extreme value analysis is performed on the boundary response data to generate the load-bearing limit parameters of each power supply. Based on the load limit parameters, a unified timing adjustment benchmark is set, and the timing adjustment benchmark is used to synchronize the power output adjustment process of each power source to obtain coordinated timing data. Timing synchronization control parameters are generated based on the coordinated timing data. Based on the timing synchronization control parameters, a final load distribution control instruction set is generated. Each power source executes the control instruction set to coordinate output and obtain actual operating status data, thereby completing the microgrid load distribution control.

2. The method according to claim 1, characterized in that, The step of generating system entropy parameters by performing chaotic measurement analysis on the real-time output data includes: The real-time output data is subjected to disordered quantitative analysis to obtain chaotic feature vectors; Initial entropy data is generated by performing entropy integration based on the chaotic feature vector. The initial entropy data is subjected to time series standardization to obtain system entropy parameters.

3. The method according to claim 1, characterized in that, The process of generating chaotic operating data by randomly perturbing the output of each power source based on the entropy increase perturbation excitation mechanism includes: A random perturbation parameter set is generated based on the aforementioned entropy-increasing perturbation excitation mechanism; Input the random disturbance parameter set into each power source to obtain a power source-specific disturbance command; The power supply-specific disturbance command is used to drive each power supply model to perform output disturbance and obtain disturbance response data; The disturbance response data is integrated and processed to generate chaotic state operation data.

4. The method according to claim 1, characterized in that, The step of extracting the inherent frequency features of each power source from the chaotic operating data to obtain the resonant coupling matrix includes: Frequency domain transformation is performed on the chaotic state operation data to obtain frequency spectrum distribution data; Identify the frequency generation characteristic identifiers of each power source's dominant frequency from the frequency spectrum distribution data; Based on the frequency characteristic identifier, the coupling strength is analyzed, and a resonance coupling matrix is ​​constructed.

5. The method according to claim 1, characterized in that, The step of performing a counterexample search on the preliminary load allocation scheme to generate allocation vulnerability identification data includes: Construct an adversarial load allocation mode that is the opposite of the initial load allocation scheme; Based on the aforementioned adversarial allocation mode, conflict scenario simulations are performed to obtain conflict response results; The conflict response results are analyzed to generate vulnerability markers for allocation; Based on the allocation vulnerability markers, allocation vulnerability identification data is generated.

6. The method according to claim 1, characterized in that, The step of matching and analyzing the load demand data with the load capacity assessment data to generate a load transmission topology includes: Perform a difference analysis between the load demand data and the load capacity assessment data to obtain load gap data; Construct a load transfer path between power sources based on the load gap data; Analyze each transfer path to generate conduction impedance coefficients; The load conduction topology is generated using the aforementioned conduction impedance coefficient.

7. The method according to claim 1, characterized in that, The step of using the boundary active detection instruction set to drive each power supply to gradually approach the operating boundary and obtain boundary response data includes: The power output of each power source is increased in a stepwise manner based on the aforementioned boundary active detection command set; During the stepped power output increase process, the power supply status is monitored to obtain response characteristic data; Nonlinear analysis is performed on the response characteristic data to identify abrupt change points, and the parameters of the abrupt change points are recorded; Boundary response data is generated based on the mutation point parameters.

8. The method according to claim 1, characterized in that, The process of synchronizing the power output adjustment process of each power source using the timing adjustment reference to obtain coordinated timing data includes: Based on the timing adjustment benchmark, power output adjustment commands for each power source are generated; Timing identification data is obtained by timing marking the power output adjustment commands of each power source; The time sequence identifier data is synchronized and arranged to generate a coordinated execution sequence; The coordinated execution sequence is optimized for timing shifting to obtain coordinated timing data.

9. A microgrid load distribution control device based on entropy increase disturbance, characterized in that, include: The data acquisition module is used to acquire real-time power output data and load demand data of the power sources in the microgrid, perform chaotic measurement analysis on the real-time power output data to generate system entropy parameters, and construct an entropy increase disturbance excitation mechanism based on the system entropy parameters. The chaotic coordination module is used to generate chaotic state operating data by randomly perturbing the output of each power source based on the entropy increase perturbation excitation mechanism, extract the inherent frequency characteristics of each power source from the chaotic state operating data to obtain the resonance coupling matrix, and use the resonance coupling matrix to generate a frequency synchronization coordination mechanism between power sources. The strategy optimization module is used to generate a preliminary load allocation scheme based on the frequency synchronization coordination mechanism, perform counterexample search on the preliminary load allocation scheme to generate allocation vulnerability identification data, and obtain an enhanced allocation strategy based on the allocation vulnerability identification data. The stress analysis module is used to obtain load capacity assessment data of each power source based on the enhanced allocation strategy, match and analyze the load demand data with the load capacity assessment data to generate a load transmission topology, and use the load transmission topology to obtain stress distribution data among each power source. The boundary detection module is used to generate an active boundary detection instruction set based on the pressure distribution data, use the active boundary detection instruction set to drive each power supply to gradually approach the operating boundary to obtain boundary response data, and perform extreme value analysis on the boundary response data to generate the load-bearing limit parameters of each power supply. The timing control module is used to set a unified timing adjustment benchmark based on the load limit parameters, use the timing adjustment benchmark to synchronize the power output adjustment process of each power source to obtain coordinated timing data, and generate timing synchronization control parameters based on the coordinated timing data. The instruction execution module is used to generate a final load distribution control instruction set based on the timing synchronization control parameters. Each power source executes the control instruction set to coordinate output and obtain actual operating status data, thereby completing the microgrid load distribution control.

10. A computer device, characterized in that, It includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and, when executing the computer program, implement the method as described in any one of claims 1 to 8.

Images