A Multi-Region Cooperative Control Method for Hybrid Power Grids Based on Improved ADMM

Abstract

This invention discloses a multi-regional collaborative control method for hybrid power grids based on an improved ADMM (Automatic Generation Control Model), belonging to the field of multi-regional collaborative control technology for automatic generation control of new energy-AC / DC hybrid power grids. The method includes: dividing the hybrid power grid into multiple sub-regions and mapping them to form a collaborative control region; establishing a global objective function based on wind and solar power output characteristics, DC power modulation, and the state of charge constraints of the energy storage system, including frequency deviation, tie-line power deviation, and new energy absorption rate, and quantifying the frequency regulation resource coordination relationship within the collaborative control region; avoiding the adjustment conflicts and response lag problems of traditional algorithms; significantly improving the new energy absorption rate while reducing the overshoot of frequency deviation and tie-line power deviation; and ensuring system safety through hard constraints, with overall performance superior to existing control methods that only focus on local optimization or a single objective.

Description

Technical Field

[0001] This invention belongs to the field of multi-regional collaborative control technology of automatic generation control for new energy-AC / DC hybrid power grids, specifically involving a multi-regional collaborative control method for hybrid power grid AGC based on an improved ADMM. Background Technology

[0002] With the accelerated construction of new power systems, the installed capacity of wind power and photovoltaic power in new energy-AC / DC hybrid power grids continues to grow, and the scale of cross-regional DC transmission is constantly expanding. AGC multi-regional collaborative control has become the core direction for ensuring grid frequency stability and improving the capacity for new energy absorption. However, the output of new energy is highly volatile due to natural conditions, and the operating constraints of AC / DC systems are complex and changeable. In addition, the uneven distribution of frequency regulation resources and significant differences in interests among regions make it difficult for traditional single-region AGC control methods to balance global optimization and local interests across multiple regions. This leads to problems such as large frequency deviation overshoot, severe power fluctuations in tie lines, and low new energy absorption rates.

[0003] While the field of coordinated control for automatic generation control (AGC) in hybrid AC / DC grids has developed multiple technical approaches, key bottlenecks remain. Some coordinated algorithms based on the traditional Alternating Directional Multiplier Method (ADMM) can achieve distributed solutions and reduce computational burden, but they suffer from slow convergence and weak anti-interference capabilities when dealing with high-frequency fluctuations in renewable energy. Control methods based on intelligent optimization algorithms can improve local frequency regulation accuracy, but they do not fully consider AC / DC system constraints and regional interest balance, easily leading to cross-regional regulation conflicts. Hierarchical control architectures can divide control levels, but lack scientific basis for regional division and efficient multi-objective coordination mechanisms, making them difficult to adapt to the complex characteristics of hybrid grids. Furthermore, existing methods often focus on single objectives such as frequency stability or power balance, failing to adequately consider the systematic coupling of multiple objectives such as renewable energy absorption and AC / DC constraints. Simultaneously, the dynamic changes in the operating conditions of hybrid grids pose challenges to centralized control, including large information interaction volumes and poor adaptability. How to scientifically divide control areas based on grid characteristics, achieve organic integration of regional autonomous optimization and global coordinated control, and how to improve ADMM algorithms to adapt to the multi-constraint and strong fluctuation characteristics of hybrid grids remain unresolved key technical problems.

[0004] To address the above issues, it is necessary to develop a multi-regional collaborative control method for hybrid power grids based on an improved ADMM to solve existing problems and enhance the ability of new energy-AC / DC hybrid power grids to cope with uncertainties and improve operational efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-regional collaborative control method for hybrid power grid AGC based on an improved ADMM, so as to solve the problems of regulation conflict and response lag in the traditional ADMM algorithm.

[0006] Firstly, this application provides a multi-regional cooperative control method for hybrid power grid AGC based on an improved ADMM, including:

[0007] The hybrid power grid is divided into multiple sub-regions and mapped to form a collaborative control area;

[0008] Based on the output characteristics of wind power and photovoltaic power, DC power modulation and the state of charge constraints of energy storage systems, a global objective function including frequency deviation, tie-line power deviation and new energy absorption rate is established to quantify the frequency regulation resource coordination relationship in the coordinated control area.

[0009] The global objective function is decomposed into local objective functions corresponding to each of the cooperative control regions;

[0010] Each local objective function is iteratively updated and dualized until the convergence condition is met.

[0011] Output collaborative control commands for the collaborative control area.

[0012] In one implementation of the first aspect, dividing the hybrid power grid into multiple sub-regions and mapping them to form a coordinated control region includes:

[0013] Define the indicators of new energy enrichment, comprehensive load, and DC correlation.

[0014] Based on the aforementioned new energy enrichment index, load comprehensive index, and DC correlation index, a set of regions with similar characteristics is identified.

[0015] In one implementation of the first aspect, the definition of the new energy enrichment index, the comprehensive load index, and the DC correlation index includes:

[0016] The renewable energy enrichment index is defined based on the proportion of renewable energy installed capacity and the matching relationship between the average daily output and average daily load of renewable energy. The calculation formula is as follows:

[0017]

[0018] in, Indicator representing the enrichment degree of new energy sources. Indicates the weighting coefficient. Indicates the first Sub-regions Subregion The total installed capacity of new energy, Subregion Total installed power generation capacity Subregion The daily output of new energy sources Subregion The average daily load demand;

[0019] The load composite index is defined by weighting and integrating load density and intraday maximum load fluctuation. The calculation formula is as follows:

[0020]

[0021] in, This represents the comprehensive load index. Indicates the weighting coefficient. Subregion The maximum daily load, Subregion The area supplied by the power grid, Indicates the first Maximum power fluctuation amplitude of load in each sub-region within 24 hours;

[0022] The DC correlation index is defined based on the proportion of rated transmission power at DC landing points within the region and the relative density of external tie lines. The calculation formula is as follows:

[0023]

[0024] in, Indicators representing DC correlation degree Indicates the weighting coefficient. Subregion Rated transmission power of the internal DC landing point, This represents the sum of the rated power of all DC terminals in the entire power grid. This indicates the total number of DC landing points. Subregion The number of communication lines connecting to external areas. Subregion The total number of power grid nodes within the area.

[0025] In one implementation of the first aspect, identifying a set of regions with similar characteristics based on the new energy enrichment index, the comprehensive load index, and the DC correlation index includes: constructing a three-dimensional feature vector based on the new energy enrichment index, the comprehensive load index, and the DC correlation index, expressed as:

[0026]

[0027] in, Subregion 3D feature vectors;

[0028] After removing outliers, valid 3D feature vectors are obtained, and a feature matrix is ​​constructed, expressed as follows:

[0029]

[0030] in, Represents the characteristic matrix, This represents the first valid three-dimensional feature vector. This represents the second valid three-dimensional feature vector. Indicates the first One effective three-dimensional feature vector, Represents the transpose of a matrix. Represent a A real matrix with 3 rows and 3 columns. Indicates the number of valid three-dimensional feature vectors;

[0031] The formula for calculating the sum of squares within clusters corresponding to different numbers of clusters is as follows:

[0032]

[0033] in, Represents the number of clusters The sum of squares within the cluster, Indicates the first The center of each cluster is determined by the mean of all valid three-dimensional eigenvectors within the cluster. Calculations show that Denotes the Euclidean norm;

[0034] Three cluster center samples were randomly selected as the initial cluster centers. , , Calculate each sub-region The Euclidean distances to the three cluster centers define the sub-regions. Assigning to the nearest cluster is calculated using the following formula:

[0035]

[0036] in, Indicates the current iteration number. Indicates the first Each cluster passed through Cluster center after the next iteration;

[0037] Based on the assigned clusters, the center of each cluster is recalculated using the following formula:

[0038]

[0039] in, Indicates the recalculation of the first Each cluster passed through Cluster center after +1 iteration Indicates the first The number of samples in each cluster;

[0040] Convergence is determined based on the recalculated center of each cluster; the convergence is achieved when the change in the center of all clusters is less than a threshold. The iteration terminates when the threshold is reached; otherwise, it returns to continue iterating. The calculation formula is:

[0041]

[0042] in, Indicates the threshold;

[0043] After the iteration terminates, based on the final cluster centers of the three clusters... The size of the area, combined with its physical meaning, is mapped to the new energy rich area, the load center area, and the DC landing point area.

[0044] In one implementation of the first aspect, the wind power and photovoltaic output characteristics include: the actual output of wind power must satisfy:

[0045]

[0046] in, This indicates the minimum technical output of the wind farm. Indicates the first Each wind farm at any time Actual output This indicates the rated output of the wind farm;

[0047] The rate of change of wind power output satisfies the ramp rate constraint:

[0048]

[0049] in, Indicates the automatic power generation control step size. , They represent the first The limits for the upward and downward ramp rates of a wind farm. Indicates the first The wind farm at the previous control time ; output power;

[0050] The upper and lower limits of photovoltaic power output and the constraints imposed by the ramp rate are as follows:

[0051]

[0052] in, Indicates the first A photovoltaic power station at time Actual output Indicates the first The rated output of each photovoltaic power station Indicates the first Minimum technical output limit for a photovoltaic power station Indicates time Photovoltaic power output efficiency, , These represent the upward ramp rate limit and the downward ramp rate limit for photovoltaic power plants, respectively. Indicates the first The photovoltaic power station was in the previous control time. ; output power;

[0053] The DC power modulation includes: the DC line transmission power satisfying the modulation range constraint.

[0054]

[0055] in, Indicates the first DC lines at time Transmission power, , They represent the first Minimum and maximum transmission power limits for each DC line;

[0056] The transmission power of AC lines must meet the transmission capacity constraints:

[0057]

[0058] in, Indicates the first The communication and contact lines are always Transmission power, Indicates the maximum transmission capacity of the AC line;

[0059] The state-of-charge (POC) constraints of the energy storage system include: a range-constrained POC condition for the energy storage system based on ensuring battery life and frequency regulation reliability, with the constraint form as follows:

[0060]

[0061] in, Indicates the energy storage system at time The state of charge, , These represent the minimum and maximum allowable values ​​for the state of charge of the energy storage system, respectively.

[0062] Charge / discharge rate constraints are established based on charge / discharge power, charge / discharge efficiency, and rated capacity.

[0063]

[0064] in, Indicates the energy storage system at time The charging and discharging power, Indicates the energy storage system at time The charge and discharge efficiency, Indicates the rated capacity of the energy storage system. , These represent the minimum charge / discharge power limit and the maximum charge / discharge power limit of the energy storage system, respectively. This indicates the energy storage system at the previous control time. The state of charge.

[0065] In one implementation of the first aspect, establishing a global objective function including frequency deviation, tie-line power deviation, and renewable energy absorption rate, and quantifying the frequency regulation resource coordination relationship in the coordinated control area includes:

[0066] Based on the integral absolute error, the objective function for frequency deviation is determined, and its expression is:

[0067]

[0068] in, The objective function representing the frequency deviation is... Indicates time frequency deviation, Indicates the automatic power generation control cycle. The weighting coefficient representing the maximum frequency deviation. This represents the integral absolute error index, reflecting the continuous impact of frequency deviation. This indicates the maximum frequency deviation within the control cycle;

[0069] Based on the integral-weighted absolute power deviation, the objective function for tie-line power deviation is determined, and its expression is:

[0070]

[0071] in, This represents the objective function for tie-line power deviation. Indicates time The power deviation of the connecting line, Indicates the time weighting factor. Weighting coefficients representing the range of power fluctuations. This represents the integral-weighted absolute power deviation index, reflecting the weighted cumulative effect of power deviation. Indicates the maximum power deviation within the control cycle; Indicates the automatic power generation control cycle;

[0072] Based on the renewable energy absorption rate, the objective function for the renewable energy absorption rate is determined, and its expression is:

[0073]

[0074] in, The objective function represents the renewable energy consumption rate. This indicates the total number of new energy power stations. Indicates time No. The actual power absorption capacity of each new energy power station Indicates time No. The theoretical power generation capacity of a new energy power station.

[0075] In one implementation of the first aspect, the establishment of a global objective function including frequency deviation, tie-line power deviation and renewable energy absorption rate, and the quantification of the frequency regulation resource coordination relationship in the coordinated control area further include: building a three-level hierarchical structure of target layer, criterion layer and scheme layer;

[0076] The criterion layer judgment matrix is ​​constructed using the scaling method. :

[0077]

[0078] in, , , , , , All represent judgment matrices Matrix elements;

[0079] Calculate the judgment matrix Maximum eigenvalue The corresponding feature vectors are normalized to obtain the initial weights of the criterion layer. ;

[0080] Calculate the consistency index and consistency ratio:

[0081]

[0082]

[0083] in, Indicators of consistency Indicates the consistency ratio. This represents the random consistency index. To determine the order of a matrix, if the consistency ratio is less than a preset consistency ratio threshold, the consistency of the matrix is ​​considered to be satisfied and the initial weights are reliable.

[0084] The maximization objective is transformed into a minimization form to achieve aggregate optimization of multiple objectives. A single-objective function is constructed by combining the initial weights, and its expression is:

[0085]

[0086] in, This represents the global objective function.

[0087] In one implementation of the first aspect, the step of decomposing the global objective function into local objective functions corresponding to each of the cooperative control regions includes:

[0088] The decomposition expression of the global objective function is:

[0089]

[0090] In the formula, Represents the global objective function. Indicates the first The local objective function of the control region for each cooperative control region For coordinated control of the area Local decision variables, For power interaction variables between coordinated control regions;

[0091] Based on the power interaction variables between the cooperative control regions, a local objective function for each cooperative control region is constructed, expressed as follows:

[0092]

[0093] in, Indicates the collaborative control area The local objective function, Indicates the collaborative control area Local decision variables, , Indicates the collaborative control area The coefficient matrix, Indicates the collaborative control area The constraint constant term.

[0094] In one implementation of the first aspect, the iterative update and dual adjustment of each local objective function includes:

[0095] In the In the next iteration, each cooperative control region is based on the first... The dual variable of the next iteration Solve the local optimization independently and update the local decision variables. The expression is:

[0096]

[0097] in, Indicates the collaborative control area Local decision variables, Indicates the penalty factor. Represents the transpose of a matrix. Indicates the first The global interaction variable for the next iteration;

[0098] The global interaction variable is updated based on the local decision variables of each collaborative control region, expressed as follows:

[0099]

[0100] in, Indicates the first The global interaction variable for the next iteration; Represents a global interactive variable;

[0101] The dual variable is updated based on the deviation between the local decision variables and the global interaction variables in each collaborative control region. The expression is as follows:

[0102]

[0103] in, Indicates the first The dual variable of the next iteration;

[0104] The iteration continues by repeatedly updating local decision variables and updating global interaction variables until a preset convergence threshold is reached, at which point the iteration terminates. The expression is:

[0105]

[0106] in, This indicates the preset convergence threshold; For the first The first collaborative control area in the first Local decision variables after the next iteration;

[0107] Based on the dynamic adjustment of the penalty factor according to the fluctuation degree of new energy power output and the constraint tightness of DC system, the standard deviation of new energy power output prediction error is used. The formula for quantifying the fluctuation of new energy power output is as follows:

[0108]

[0109] in, Indicates the number of prediction periods. , They represent Real-time comparison of predicted and actual renewable energy output;

[0110] Define DC power constraint tightness coefficient The formula for quantifying DC constraint tightness is as follows:

[0111]

[0112] in, Indicates the actual DC transmission power. , Indicates the upper and lower limits of DC power modulation;

[0113] Based on the standard deviation of the new energy output prediction error and the DC power constraint tightness coefficient, the penalty factor is dynamically adjusted to achieve an adaptive step size. The calculation formula is as follows:

[0114]

[0115] in, Indicates adaptive step size, Indicates the initial penalty factor. This represents the maximum standard deviation of the prediction error for new energy power output. Indicates the first The standard deviation of the new energy output prediction error in the next iteration Indicates the step size and the number of iterations. Indicates the DC power constraint tightness coefficient;

[0116] The prediction error compensation term is calculated based on the deviation between the predicted and actual values ​​from the Long Short-Term Memory network. The calculation formula is as follows:

[0117]

[0118] in, This represents the prediction error compensation term. Indicates the future Forecast values ​​of renewable energy output for the specified time period;

[0119] Embed the prediction error compensation term into the collaborative control region The local objective function is expressed as:

[0120]

[0121] in, Indicates the first term after embedding the prediction error compensation term. The modified local objective function of the cooperative control region Indicates the first The original local objective function of each cooperative control region Indicates the compensation coefficient. Indicates the collaborative control area New energy output decision variables;

[0122] Based on the frequency modulation capacity input, response speed, and control effect of each coordinated control area, a regional contribution index is defined, and the calculation formula is as follows:

[0123]

[0124] in, Indicates the collaborative control area The regional contribution index, Indicates the collaborative control area The frequency modulation capacity input, Indicates the total global frequency modulation capacity. Indicates the collaborative control area frequency modulation response time, Indicates the collaborative control area The improvement in the global objective function. , , Indicates the weighting coefficient. This represents the total global improvement.

[0125] A contribution penalty term is added to the global objective function, calculated using the following formula:

[0126]

[0127]

[0128] in, This represents the global objective function after adding a contribution penalty term. The global objective function is... Indicates the penalty coefficient. This indicates the average contribution.

[0129] In one implementation of the first aspect, the convergence condition includes:

[0130] Define the convergence function as follows:

[0131]

[0132] in, Represents the Lyapunov function. , , These represent the collaborative control areas. The optimal solution for local decision variables, global interaction variables, and dual variables. Represented as the first The adaptive step size for the next iteration;

[0133] Substituting the iterative update formula of the alternating direction multiplier method, which includes adaptive step size, error compensation term, and contribution penalty term, into the Lyapunov function, we obtain the convergence condition:

[0134] Adaptive step size Convergence condition:

[0135] ;

[0136] in, , These represent the minimum penalty factor and the maximum penalty factor, respectively.

[0137] Convergence conditions for error compensation and contribution penalty:

[0138]

[0139] in, Indicates the compensation coefficient. This represents the penalty coefficient.

[0140] According to the above-mentioned technical solution proposed in this application, the following technical effects can be achieved: This multi-regional collaborative control method for hybrid power grids based on improved ADMM, compared with traditional methods that lack scientific basis for regional division and have vague descriptions of multi-element coupling relationships, constructs three-dimensional indicators of new energy enrichment, load integration, and DC correlation, and combines clustering algorithms to achieve accurate division of new energy enrichment areas, load center areas, and DC landing point areas. Simultaneously, it embeds wind power and photovoltaic output characteristics, DC power modulation, and energy storage system charge state constraints, establishing a global objective function containing frequency deviation, tie-line power deviation, and new energy absorption rate, comprehensively quantifying the regional frequency regulation resource coordination relationship, and solving the problem of one-sidedness in traditional modeling; based on new energy... The design incorporates adaptive step size to accelerate convergence by addressing source fluctuations and DC constraint tightness. It introduces a long short-term memory network (LSTM) prediction error compensation term to suppress fluctuation interference, adds a contribution penalty term to balance regional interests, and uses Lyapunov stability theory to ensure convergence. This approach retains the advantages of ADMM distributed solution while overcoming its adaptation bottleneck in hybrid power grids, avoiding the adjustment conflicts and response lag problems of traditional algorithms. Furthermore, it employs the analytic hierarchy process (AHP) to determine weights, integrating and optimizing multiple objectives such as frequency stability, power balance, and renewable energy absorption. This significantly improves renewable energy absorption rate while reducing frequency deviation and tie-line power deviation overshoot, and ensures system safety through hard constraints. Its overall performance surpasses existing control methods that only focus on local optimization or a single objective. Attached Figure Description

[0141] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below only involve some embodiments of this disclosure and are not a limitation of this disclosure. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0142] Figure 1 This is a flowchart illustrating the method described in this application; Figure 2 This is a graph illustrating the verification of the elbow rule in K-means clustering in this application. Figure 3 This is a flowchart of the K-means clustering algorithm according to an embodiment of this application; Figure 4 This is a flowchart of the improved alternating direction multiplier method according to an embodiment of this application; Figure 5 The graph shows the sensitivity analysis of the adaptive step size, LSTM compensation coefficient, and contribution penalty parameter of the improved alternating direction multiplier method in this application embodiment. Figure 6 This is a simulation model diagram of the IEEE 33-node system according to an embodiment of this application. Figure 7 This is a quasi-scenario load curve and AC / DC power curve diagram of an embodiment of this application; Figure 8 This is a key scenario 1 of the embodiments of this application, showing the fluctuation curve of new energy power output; Figure 9 This is a key scenario 2 of the embodiments of this application, showing the AC tie line fault and DC power modulation system disturbance curves; Figure 10 This is a comparison chart of frequency deviation in the baseline scenario, key scenario 1, key scenario 2, and complex scenario of the embodiments of this application. Figure 11 This is a comparison diagram of the tie-line power deviation in embodiments of this application; Figure 12 This is a bar chart comparing the renewable energy consumption rates in various scenarios according to embodiments of this application; Figure 13 This is a comparison chart of regional contribution deviation rate and energy storage loss cost in an embodiment of this application. Detailed Implementation

[0143] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0144] Unless otherwise defined, the technical or scientific terms used in this disclosure shall have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms “first,” “second,” and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as “comprising” or “including” mean that the element or object preceding the word encompasses the element or object listed following the word and its equivalents, without excluding other elements or objects. Terms such as “connected” or “linked” are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as “upper,” “lower,” “left,” and “right” are used only to indicate relative positional relationships, which may change accordingly when the absolute position of the described objects changes. It should be understood that the various steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0145] Firstly, according to one or more embodiments, a multi-region cooperative control method for AGC in hybrid power grids based on an improved ADMM is provided, such as... Figure 1 As shown, it includes:

[0146] S1. The sub-regions (smallest calculation units) of the hybrid power grid are pre-divided according to electrical boundaries and geographical range, and mapped to form new energy rich areas, load center areas and DC landing point areas. The sub-regions are the direct statistical objects of new energy richness indicators, load comprehensive indicators and DC correlation indicators, and are also the basic samples for subsequent cluster analysis. They do not have the preset control function attributes and only exist as units for data statistics and feature quantification.

[0147] S11. To accurately quantify the degree of regional renewable energy concentration and reflect the intensity of regional demand for renewable energy consumption and transmission, a renewable energy concentration index is defined. This index comprehensively considers the proportion of regional renewable energy installed capacity and the matching relationship between average daily renewable energy output and average daily load. The calculation formula is as follows:

[0148]

[0149] in, Indicator representing the enrichment degree of new energy sources. Indicates the weighting coefficient. Indicates the first Sub-regions Subregion The total installed capacity of new energy, Subregion Total installed power generation capacity Subregion The daily output of new energy sources Subregion The average daily load demand.

[0150] To comprehensively characterize regional load characteristics, integrate load scale and fluctuation features, and thus reflect the region's demand intensity for Automatic Generation Control (AGC) frequency regulation, a comprehensive load index is defined. This index is calculated by weighting and integrating load density and maximum daily load fluctuation, using the following formula:

[0151]

[0152] in, This represents the comprehensive load index. Indicates the weighting coefficient. Subregion The maximum daily load, Subregion The area supplied by the power grid, Indicates the first Maximum power fluctuation amplitude of load in each sub-region within 24 hours.

[0153] To accurately measure the degree of interconnection between a region and the DC system and highlight the region's pivotal role in cross-regional power transmission, a DC interconnection index is defined. This index comprehensively considers the proportion of rated transmission power at DC landing points within the region and the relative density of external tie lines. The calculation formula is as follows:

[0154]

[0155] in, Indicators representing DC correlation degree Indicates the weighting coefficient. Subregion Rated transmission power of the internal DC landing point, This represents the sum of the rated power of all DC terminals in the entire power grid. Subregion The number of communication lines connecting to external areas. Subregion The total number of power grid nodes within the area.

[0156] like Figure 2 , Figure 3 As shown, to achieve the scientific division of power grid areas, the three-dimensional feature vectors of each sub-region calculated above need to be input into a clustering algorithm to mine the inherent structure of the data and distinguish regions with different characteristics. The expression is as follows:

[0157]

[0158] in, Subregion The three-dimensional feature vector.

[0159] In one specific embodiment of the present invention, the K-means clustering algorithm is used to identify a set of regions with similar features by iteratively optimizing the similarity of samples within a cluster;

[0160] First, the original sample data required for the sub-region feature matrix, mainly including regional operational data, static data, and abnormal data (i.e., power grid monitoring and measurement data), undergoes data preprocessing to remove outliers caused by data acquisition or calculation errors. This ensures the validity and reliability of the samples participating in clustering, ultimately yielding the number of effective three-dimensional feature vector samples, and constructing the feature matrix: The expression is:

[0161]

[0162] in, Represents the characteristic matrix, This represents the first valid three-dimensional feature vector. This represents the second valid three-dimensional feature vector. Indicates the first One effective three-dimensional feature vector, Represents the transpose of a matrix. Represent a A real matrix with 3 rows and 3 columns. This represents the number of valid three-dimensional feature vectors.

[0163] To determine a reasonable number of clusters, the elbow rule is used for verification. The sum of squared squares (WCSS) within each cluster corresponding to different numbers of clusters k is calculated using the following formula:

[0164]

[0165] in, Represents the number of clusters The sum of squares within the cluster, This represents the center of the c-th cluster, which is composed of all sub-regions within the cluster. Mean of effective three-dimensional feature vectors Calculations show that This represents the Euclidean norm.

[0166] After determining the number of clusters, the iterative calculation step of K-means clustering is performed. Three cluster center samples are randomly selected as the initial cluster centers. , , For each sub-region Calculate its Euclidean distance to the centers of the three clusters, and assign it to the cluster with the closest distance. The calculation formula is as follows:

[0167] in, Indicates the current iteration number. Indicates the first Each cluster passed through Cluster center after the next iteration.

[0168] Based on the assigned clusters, the center of each cluster is recalculated using the following formula:

[0169]

[0170] in, Indicates the recalculation of the first Each cluster passed through Cluster center after the next iteration Indicates the first Number of samples in each cluster.

[0171] Convergence is determined based on the recalculated center of each cluster; the convergence is achieved when the change in the center of all clusters is less than a threshold. That is, satisfying:

[0172]

[0173] in, If the threshold is indicated, the iteration terminates; otherwise, return to the sample allocation step and continue iterating.

[0174] After the iteration terminates, based on the final cluster centers of the three clusters... The size of the region is combined with the physical meaning of the mapped region type to achieve collaborative control region division.

[0175] The collaborative control area comprises the renewable energy rich area, the load center area, and the DC landing point area. In the stages of clustering result mapping, collaborative control partition formation, and control command issuance and execution, that is, after clustering analysis of the feature vectors of the above sub-regions, three types of partitions with clear control functions are formed by mapping according to their dominant feature attributes. This collaborative control area is the execution unit of global optimization target decomposition and distributed collaborative control, and undertakes differentiated control responsibilities.

[0176] S2. Embedding wind / solar power output characteristics, DC power modulation, and energy storage SOC constraints, a global objective function is established, including frequency deviation, tie-line power deviation, and renewable energy absorption rate, to quantify the frequency regulation resource coordination relationship in each coordinated control area:

[0177] S21. For wind power output, its operating range is limited by the rated power and minimum technical output of the unit, that is, at any given time... The output must meet the following requirements:

[0178]

[0179] in, This indicates the minimum technical output of the wind farm. Indicates the first Each wind farm at any time Actual output This indicates the rated output of the wind farm;

[0180] Meanwhile, the rate of change of wind power output is limited by the mechanical characteristics of the unit and must also meet the gradeability constraint:

[0181] in, Indicates the AGC control step size. , They represent the first Uphill and downhill ramp rate limits for individual wind farms Indicates the first The wind farm at the previous control time . output power.

[0182] Considering the effects of light intensity and temperature on output, the upper and lower limits of output and the constraints imposed by the gradient rate are as follows:

[0183]

[0184] in, Indicates the first A photovoltaic power station at time Actual output Indicates the first The rated output of each photovoltaic power station Indicates the first Minimum technical output limit for a photovoltaic power station Indicates time Photovoltaic power output efficiency, , These represent the upward ramp rate limit and the downward ramp rate limit for photovoltaic power plants, respectively. Indicates the first The photovoltaic power station was in the previous control time. . output power.

[0185] S22. For DC systems, their power modulation capability is limited by factors such as converter station capacity and valve-side voltage. The transmitted power at any given time must meet the modulation range constraint.

[0186]

[0187] in, Indicates the first DC lines at time Transmission power, , They represent the first Minimum and maximum transmission power limits for each DC line.

[0188] The power transmission of AC lines is limited by constraints such as thermal stability and voltage stability, and must meet transmission capacity constraints.

[0189]

[0190] in, Indicates the first The communication and contact lines are always Transmission power, This indicates the maximum transmission capacity of the AC line.

[0191] S23. As a fast response resource for AGC, the state of charge (SOC) and charge / discharge rate of the energy storage system directly determine the frequency regulation capability. It is necessary to avoid overcharging, over-discharging or power exceeding the limit of the energy storage through dual constraints.

[0192] First, the energy storage SOC needs to be maintained within a reasonable range to ensure battery life and frequency regulation reliability. The constraint is as follows:

[0193]

[0194] in, Indicates the energy storage system at time The state of charge, , These represent the minimum and maximum permissible values ​​of the state of charge of the energy storage system, respectively.

[0195] Secondly, there is a dynamic correlation between the energy storage charging and discharging power and the State of Charge (SOC). The change in SOC is jointly determined by the charging and discharging power, charging and discharging efficiency, and rated capacity, forming a charging and discharging rate-related constraint.

[0196]

[0197] in, For energy storage systems at all times The charging and discharging power, For energy storage systems at all times The charge and discharge efficiency, For the rated capacity of the energy storage system, , These are the minimum and maximum charge / discharge power limits for the energy storage system. This indicates the energy storage system at the previous control time. The state of charge.

[0198] S24. The cumulative effect of frequency deviation over time is quantified using integral absolute error (IAE), and a maximum frequency deviation constraint is introduced to prevent instantaneous frequency exceeding the limit from triggering protection actions. The expression for the objective function is:

[0199]

[0200] in, The objective function representing the frequency deviation is... Indicates time frequency deviation, Indicates the AGC control cycle. The weighting coefficient representing the maximum frequency deviation. This represents the integral absolute error index, reflecting the continuous impact of frequency deviation. This indicates the maximum frequency deviation within the control cycle.

[0201] S25. In AC / DC hybrid power grids, tie lines are critical channels for power exchange between regions. Excessive power deviation can lead to power imbalance between regions and even trigger cascading failures. It is necessary to suppress tie line power fluctuations and control cumulative deviations through an objective function.

[0202] The cumulative time effect of power deviation is quantified using integral-weighted absolute power deviation (IPAE). Combined with power fluctuation range constraints, the objective function expression is as follows:

[0203]

[0204] in, This represents the objective function for tie-line power deviation. Indicates time The power deviation of the connecting line, The time-weighted coefficient, Weighting coefficients representing the range of power fluctuations. This is an integral-weighted absolute power deviation index, reflecting the weighted cumulative effect of power deviation. This indicates the maximum power deviation within the control cycle. Indicates the automatic power generation control cycle;

[0205] S26. The objective function for the renewable energy absorption rate focuses on maximizing the absorption rate, and its expression is:

[0206]

[0207] in, The objective function represents the renewable energy consumption rate. This indicates the total number of new energy power stations. Indicates time No. The actual power absorption capacity of each new energy power station Indicates time No. The theoretical power generation capacity of a new energy power station;

[0208] S27. In the multi-objective optimization of AGC in a new energy-AC / DC hybrid power grid, there is a significant coupling and trade-off relationship between the three objectives: frequency stability, power balance, and new energy absorption. The Analytic Hierarchy Process (AHP) is used to quantify the weights. This method transforms qualitative requirements into quantitative indicators by constructing a hierarchical structure. The specific implementation process is as follows: First, a three-level hierarchical structure is built: objective layer, criterion layer, and scheme layer. The objective layer is for AGC multi-objective optimization; the criterion layer corresponds to the three core objectives (frequency stability, power balance, and new energy absorption); and the scheme layer consists of differentiated control strategies for new energy-rich areas, load center areas, and DC landing point areas, ensuring that the hierarchical structure is deeply aligned with the actual control needs of the power grid.

[0209] Based on the requirements of power grid operation, a criterion-level judgment matrix is ​​constructed using the 1-9 scaling method (1 = equally important, 9 = extremely important). :

[0210]

[0211] in, , , , , , All represent judgment matrices Matrix elements.

[0212] By calculating the largest eigenvalue of the judgment matrix By matching the corresponding feature vectors and normalizing the feature vectors, the initial weights of the criterion layer are obtained. Simultaneously, a consistency verification mechanism is introduced to calculate the consistency index and consistency ratio:

[0213]

[0214]

[0215] in, Indicators of consistency Indicates the consistency ratio. This represents the random consistency index. =3 random consistency index =0.58. Only when When the value is less than 0.1, the consistency of the judgment matrix meets the requirements, the weight results are reliable, and the bias caused by subjective judgment is avoided.

[0216] To achieve multi-objective aggregation optimization, the maximization objective is transformed into a minimization form. A single-objective function is then constructed using the weights determined by AHP, expressed as:

[0217]

[0218] in, Represents the global objective function. The weights determined for AHP can be dynamically adjusted according to the power grid operation scenario.

[0219] S3. To address the limitations of traditional ADMM, three improvements are proposed: an adaptive step size is designed based on the fluctuation of new energy sources and the tightness of DC constraints to accelerate convergence; an LSTM prediction error compensation term is introduced to suppress fluctuation interference; and a contribution penalty term is added to balance regional interests.

[0220] The core idea of ​​the S311 and ADMM algorithms is to decompose the global optimization problem into local optimization problems in multiple cooperative control regions. By coordinating the constraints between these regions through dual variables, a distributed solution is achieved. For the constructed global objective function, combined with the power balance constraints of multiple cooperative control regions, the global optimization problem can be decomposed into three sub-problems: the renewable energy rich region, the load center region, and the DC landing point region. Its core process includes three steps: objective decomposition, iterative update, and dual adjustment.

[0221] In this embodiment, the decomposition expression of the global objective function is:

[0222]

[0223] In the formula, Represents the global objective function. Indicates the first Local objective function of the control region of each sub-region sub-region Local decision variables, For power interaction variables between coordinated control regions;

[0224] The global objective function Decomposed into local objective functions of each cooperative control region Introducing power interaction variables between coordinated control regions Construct the constrained optimization problem form:

[0225]

[0226] in, Indicates the collaborative control area The local objective function, Indicates the collaborative control area Local decision variables, , Indicates the collaborative control area The coefficient matrix, Indicates the collaborative control area The constraint constant term, This represents the power interaction variable between different collaborative control regions.

[0227] Iterative update: at the first In the next iteration, the dual variable of the previous iteration Solve the local optimization problem independently, and update the local decision variables. The expression is:

[0228]

[0229] in, Indicates the collaborative control area Local decision variables, Indicates the penalty factor. Represents the transpose of a matrix. Indicates the first The global interaction variable for the next iteration;

[0230] The coordinator is based on each cooperative control area Local decision variables Update global interaction variables The expression is:

[0231]

[0232] in, Indicates the first The global interaction variable for the next iteration. Represents a global interactive variable. This is a penalty factor used to adjust the degree of constraint satisfaction.

[0233] Dual adjustment: based on local decision variables With local decision variables The deviation is used to update the dual variable, achieving feedback coordination between the cooperative control region and the coordinator. The expression is:

[0234]

[0235] in, Indicates the first The dual variable of the next iteration.

[0236] Repeat the above steps until the convergence threshold is reached, at which point the iteration terminates. The expression is:

[0237]

[0238] in, This indicates the preset convergence threshold. For the first The first collaborative control area in the first Local decision variables after the next iteration.

[0239] like Figure 4 , Figure 5 (a) Figure 5 (b) Figure 5 As shown in (c), Figure 5 (a) in the figure represents the adaptive step-size sensitivity analysis. Figure 5 (b) in the figure represents the sensitivity analysis of the LSTM compensation coefficient. Figure 5 (c) in the figure represents the sensitivity analysis of the contribution penalty coefficient. Based on the fluctuation of new energy sources and the tightness of DC constraints, an adaptive step size is designed to accelerate convergence. An LSTM prediction error compensation term is introduced to suppress fluctuation interference, and a contribution penalty term is added to balance the benefits of the collaborative control region.

[0240] To address the limitations of traditional ADMM (Advanced Dynamic Management Model), and considering the characteristics of new energy-AC / DC hybrid power grids, a triple improvement strategy is proposed from three dimensions: convergence speed, anti-interference capability, and balancing of interests in the coordinated control area.

[0241] S321. The penalty factor is dynamically adjusted according to the fluctuation of new energy output and the tightness of DC system constraints to achieve rapid convergence when the fluctuation is large and stable iteration when the constraints are tight.

[0242] The quantification of the volatility of new energy sources is achieved by using the standard deviation of the prediction error of new energy output. To measure the degree of volatility,

[0243]

[0244] in, Indicates the number of prediction periods. , They represent Real-time forecast and actual values ​​of new energy power output.

[0245] DC constraint tightness quantification is achieved by defining the DC power constraint tightness coefficient. This reflects the degree to which the DC power approaches the modulation limit; the calculation formula is as follows:

[0246]

[0247] in, Indicates the actual DC transmission power. , Indicates the upper and lower limits of DC power modulation.

[0248] The adaptive step size combines both design principles to create a dynamic penalty factor, calculated using the following formula:

[0249]

[0250] in, Indicates adaptive step size, Indicates the initial penalty factor. This represents the maximum standard deviation of the prediction error for new energy power output. Indicates the first The standard deviation of the new energy output prediction error in the next iteration Indicates the step size and the number of iterations. This represents the DC power constraint tightness coefficient.

[0251] In this embodiment, S322, to reduce the impact of new energy power output prediction errors on the optimization results, a Long Short-Term Memory (LSTM) network prediction error compensation term is introduced. This term improves the algorithm's anti-interference capability by real-time correction of the local objective function. The LSTM is constructed using historical new energy power output data as input and outputs the predicted new energy power output value for the next T time period. The LSTM uses a 3-layer LSTM structure (20 neurons in the input layer, 32 neurons in the hidden layer, and 1 neuron in the output layer), trained using the Adam optimizer. The error compensation term includes:

[0252] The prediction error compensation term is calculated by determining the deviation between the LSTM predicted value and the actual value. The calculation formula is as follows:

[0253]

[0254] in, This represents the prediction error compensation term. Indicates the future Forecast values ​​of renewable energy output for the specified period.

[0255] Embed the prediction error compensation term into the collaborative control region The local objective function is expressed as:

[0256]

[0257] in, Indicates the first term after embedding the prediction error compensation term. The modified local objective function of the cooperative control region Indicates the first The original local objective function of each cooperative control region Indicates the compensation coefficient. Indicates the collaborative control area New energy output decision variables.

[0258] In this embodiment, S323 introduces a contribution penalty term for the collaborative control region to quantify the frequency modulation contribution of the collaborative control region and avoid imbalance of interests. The penalty coefficient is adjusted to achieve "matching of contribution and benefit".

[0259] Taking into account the frequency regulation capacity input, response speed, and regulation effect of the coordinated control area, a regional contribution index is defined, and the calculation formula is as follows:

[0260]

[0261] in, Indicates the collaborative control area The regional contribution index, Indicates the collaborative control area The frequency modulation capacity input, Indicates the total global frequency modulation capacity. Indicates the collaborative control area frequency modulation response time, Indicates the collaborative control area The improvement in the global objective function. , , Indicates the weighting coefficient. This represents the total improvement in the overall value.

[0262] A contribution penalty term is added to the global objective function to penalize regions with low contributions but high returns. The calculation formula is as follows:

[0263]

[0264]

[0265] in, This represents the global objective function after adding a contribution penalty term. The global objective function is... Indicates the penalty coefficient. This indicates the average contribution.

[0266] In this embodiment, the convergence conditions of the improved ADMM are derived based on Lyapunov stability theory, and the influence of key parameters on convergence is analyzed:

[0267] Define Lyapunov functions This is used to describe the changing trend of variable deviation during the iteration process:

[0268]

[0269] in, Represents the Lyapunov function. , , These represent the collaborative control areas. The optimal solution for local decision variables, global interaction variables, and dual variables. Represented as the first The adaptive step size for the next iteration;

[0270] Substituting the improved alternating direction multiplier method iterative update formula, which includes adaptive step size, error compensation term, and contribution penalty term, into the Lyapunov function... Through inequality derivation, the convergence condition can be obtained:

[0271] Adaptive step size Must meet:

[0272]

[0273] in, , These represent the minimum and maximum penalty factors, respectively; typically 5 and 100 are chosen to avoid... Too small a coefficient leads to slow convergence, while too large a coefficient causes oscillations; compensation coefficient With penalty coefficient Must meet

[0274]

[0275] in, Indicates the compensation coefficient. Indicates the penalty coefficient; prevents extraneous terms from excessively affecting the convexity of the objective function;

[0276] During the iterative update process, the local optimization problem of the cooperative control region must be a strictly convex function;

[0277] Under the above conditions, the Lyapunov stability theory is monotonically decreasing and has a lower bound of 0. Based on the Lyapunov stability theory, the improved ADMM algorithm converges.

[0278] To verify the effectiveness of this invention in practical applications, simulation verification was performed based on the IEEE 33-bus distribution system model, and the system parameters were reasonably preset. For example... Figure 6As shown, the IEEE 33-node distribution system comprises 33 nodes (labeled with numbers 1, 2, 3… in the diagram) and 32 branches, with a voltage level of 12.66kV. Area A is a renewable energy-rich area, Area B is a load center area, and Area C is a DC transmission point area. Based on the distribution of renewable energy, load characteristics, and electrical connections between nodes, the system is divided into three coordinated control areas: the renewable energy-rich area, the load center area, and the DC transmission point area. Power exchange occurs between these coordinated control areas via tie lines. The renewable energy-rich area primarily connects to distributed photovoltaic and small wind power generation equipment, the load center area concentrates a large number of industrial and commercial loads, and the DC transmission point area is responsible for the connection and distribution of DC transmission power.

[0279] Different types and capacities of generator sets are configured in each coordinated control area to meet the power supply and frequency regulation needs within the area. The specifications of each generator set are shown in Table 1.

[0280] Table 1 Generator Set Specifications

[0281]

[0282] To clearly present the distribution and variation characteristics of system load and AC / DC power under the baseline scenario, the curves of load power, DC power (A→C), and AC tie-line power (C→B) in area B are analyzed. The results are as follows: Figure 7 As shown, during the off-peak period (0:00-6:00), the load power in Zone B remained stable at 3GW, a relatively low level. During the morning peak period (8:00-11:00), the load power in Zone B rose to 6GW, meeting the increased electricity demand in the morning. During the evening peak period (18:00-21:00), the load power in Zone B further climbed to 6.5GW, reaching the peak load of the day. The DC power (A→C) remained stable throughout the day at 4GW, providing stable DC support for power transmission between zones. The AC tie-line power (C→B) also remained stable at 3GW, ensuring reliable AC power exchange between zones.

[0283] To verify the adaptability of the improved ADMM algorithm to the scenario of fluctuating power output from new energy sources, the traditional ADMM algorithm was selected as a comparison scheme. The power output characteristics of wind power and photovoltaic power and the algorithm's response to fluctuations were analyzed.

[0284] like Figure 8As shown, the actual wind power output fluctuates by ±20%, with severe fluctuations occurring between 10:00 and 12:00; the actual photovoltaic output fluctuates by ±15%, with severe fluctuations occurring between 14:00 and 15:00. Traditional ADMM algorithms, lacking a dynamic compensation mechanism for renewable energy fluctuations, are insufficient in their ability to regulate system frequency and power balance when facing these severe fluctuations, easily leading to increased frequency deviations and power exceeding limits. In contrast, the improved ADMM, utilizing a long short-term memory network to predict and compensate for renewable energy fluctuations, can proactively detect and respond to wind and photovoltaic output fluctuations, effectively mitigating the impact of renewable energy fluctuations on the system, enabling the system to maintain good frequency stability and power balance in this scenario.

[0285] In summary, the improved ADMM demonstrates stronger adaptability and better control effect in key scenario 1: the fluctuation of renewable energy output, verifying the effectiveness of the improved ADMM in dealing with renewable energy fluctuations.

[0286] To verify the control performance of the improved ADMM in key scenario 2: AC / DC system disturbance scenario, the traditional ADMM algorithm was selected as a comparison scheme for analysis, and the results are as follows: Figure 9 (a) Figure 9 As shown in (b) of the diagram.

[0287] like Figure 9 As shown in (a), DC power modulation begins at 12:00 with an initial power of 4GW. The improved ADMM can stabilize the DC power at 5.5GW after 30s, with a smoother modulation process and faster stabilization speed. In contrast, the traditional ADMM algorithm is inferior to the improved algorithm in terms of the smoothness and stabilization rate of power change during DC power modulation.

[0288] like Figure 9 As shown in (b), during the AC tie line fault process, a single-phase ground fault occurred at 16:00, and the fault was cleared after 0.1s. The improved ADMM can quickly restore the AC tie line power to the normal 3GW, and the power fluctuation is small during the restoration process. Under the traditional ADMM algorithm, the timeliness and stability of AC tie line power restoration are relatively poor.

[0289] In summary, the improved ADMM outperforms the traditional ADMM algorithm in terms of power regulation stability, stabilization speed, and fault recovery effect in both DC power modulation and AC tie-line fault recovery scenarios under AC / DC disturbance conditions. This verifies the effectiveness and superiority of the improved ADMM in dealing with AC / DC disturbances.

[0290] To verify the frequency regulation performance of the improved ADMM in multiple scenarios, the traditional ADMM algorithm was selected as a comparison scheme. The comparison and analysis were carried out on the baseline scenario, key scenario (new energy fluctuations, AC / DC disturbances) and complex scenario (multiple disturbances superimposed) from the dimensions of peak frequency deviation, recovery time and over-limit duration.

[0291] Table 2 Comparison of Frequency Regulation Performance in Various Scenarios

[0292]

[0293] Depend on Figure 10 (a) Figure 10 (b) Figure 10 (c) Figure 10 As shown in (d) and Table 2, in the baseline scenario, the peak frequency deviation of the traditional ADMM algorithm reaches 0.12Hz, and the recovery time to within the allowable deviation range of ±0.05Hz is approximately 8s. The improved ADMM, through adaptive step size and long short-term memory network compensation, achieves a peak frequency deviation of only 0.08Hz, reduces the recovery time to 5s, and avoids any limit violations throughout the process. When in a renewable energy fluctuation scenario with a 15% random fluctuation in renewable energy output, the traditional ADMM, lacking a dynamic compensation mechanism, experiences a sudden surge in the peak frequency deviation to 0.15Hz, with a limit violation time of approximately 2s. The improved ADMM, leveraging the prediction compensation of renewable energy fluctuations through the long short-term memory network, suppresses the peak frequency deviation to 0.09Hz, with a limit violation time of less than 0.5s. When in an AC / DC disturbance scenario with a simultaneous 10% step disturbance in AC / DC tie-line power, the peak frequency deviation of the traditional ADMM reaches 0.18Hz, with a recovery time exceeding 12s. The improved ADMM, through contribution penalty optimization of inter-regional power allocation, reduces the peak frequency deviation to 0.11Hz, and shortens the recovery time to 8s. Complex scenarios are superimposed with new energy fluctuations (15%), AC / DC disturbances (10%), and load mutations (8%). Under the traditional ADMM algorithm, the frequency deviation oscillates violently, with a peak value as high as 0.22Hz and an over-limit duration of more than 5s. The improved ADMM, with the synergistic effect of the triple improvement mechanism, controls the peak frequency deviation at 0.13Hz, with an over-limit duration of only 1.2s, and the recovery process is more stable.

[0294] To verify the improved ADMM's tie-line power regulation capability under AC / DC disturbance scenarios, the traditional ADMM algorithm was selected as a comparison scheme. Quantitative analysis was conducted from dimensions such as peak power deviation, settling time, and fluctuation amplitude. The results are as follows: Figure 11 As shown in Table 3.

[0295] Table 3 Comparison of tie-line power regulation performance in key scenario 2.

[0296]

[0297] In critical scenario 2, the system experiences a combined AC / DC disturbance at t=0.1s, with a 15% step fluctuation in DC transmission power and a 10% sudden change in AC tie-line load. After the fault is cleared at t=0.1s, the dynamic process of tie-line power deviation exhibits significant differences between the two algorithms: the traditional ADMM algorithm experiences violent power deviation oscillations, with peak values ​​reaching ±620MW, far exceeding the system safety threshold of ±200MW. The fluctuations persist for approximately 2.8s before gradually converging to a stable state, triggering multiple power limit alarms during this period, seriously threatening the safe operation of the tie-line. The improved ADMM, through adaptive step size adjustment, dynamically optimizes the iteration step size and contribution penalty mechanism to constrain the power allocation deviation between regions based on the AC / DC disturbance intensity. The peak power deviation is effectively suppressed within ±180MW, and it quickly stabilizes in only approximately 1.2s, with a smoother fluctuation process and no limit exceedances.

[0298] To verify the effectiveness of the improved ADMM algorithm in enhancing renewable energy absorption rates under different scenarios, the traditional ADMM algorithm was selected as a comparison scheme. A comparative analysis was conducted on the renewable energy absorption rates under the baseline scenario, renewable energy fluctuation scenario, AC / DC disturbance scenario, and multiple disturbance superposition scenario. The results are as follows: Figure 12 As shown.

[0299] like Figure 12 As shown, in the baseline scenario, the renewable energy absorption rate of the traditional ADMM algorithm is 92.3%, while that of the improved ADMM algorithm reaches 98.5%. In the renewable energy fluctuation scenario, the absorption rate of the traditional ADMM algorithm drops to 81.5%, while that of the improved ADMM algorithm remains at 93.8%. In the AC / DC disturbance scenario, the absorption rate of the traditional ADMM algorithm is 88.2%, while that of the improved ADMM algorithm is 95.1%. In the scenario with multiple disturbances, the absorption rate of the traditional ADMM algorithm is only 76.2%, while that of the improved ADMM algorithm is 90.1%.

[0300] In summary, the improved ADMM algorithm achieves a significantly higher renewable energy absorption rate than the traditional ADMM algorithm across all scenarios. In particular, in complex scenarios such as renewable energy fluctuations and multiple disturbances, the improved ADMM demonstrates stronger adaptability and better renewable energy absorption performance, verifying its effectiveness in improving renewable energy absorption levels.

[0301] To verify the performance of the improved ADMM algorithm in controlling regional contribution deviation rate and energy storage loss cost, the traditional ADMM algorithm was selected as a comparison scheme. The relevant indicators for renewable energy rich areas, load center areas, and DC landing point areas were analyzed, and the results are as follows: Figure 13 As shown.

[0302] like Figure 13As shown, in areas rich in renewable energy, the contribution deviation rate of the traditional ADMM algorithm is 28.5%, with a daily energy storage loss cost of 125,000 yuan; the improved ADMM algorithm significantly reduces the contribution deviation rate, with a daily energy storage loss cost of only 102,000 yuan. In load center areas, the contribution deviation rate of the traditional ADMM algorithm is as high as 31.2%, while the improved ADMM algorithm effectively controls it, bringing the daily energy storage loss cost close to zero. In DC landing point areas, the contribution deviation rate of the traditional ADMM algorithm is 26.8%, with a daily energy storage loss cost of 83,000 yuan; the improved ADMM algorithm reduces the contribution deviation rate while lowering the daily energy storage loss cost to 69,000 yuan.

[0303] In summary, the improved ADMM can effectively reduce the contribution deviation rate and energy storage loss cost in different regions. Compared with the traditional ADMM algorithm, it has significant advantages in ensuring the balance of contribution between regions and controlling the operating cost of energy storage, which verifies the effectiveness of the improved ADMM in improving the economy and fairness of multi-regional collaborative control.

[0304] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0305] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

[0306] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM, characterized in that: include: The hybrid power grid is divided into multiple sub-regions and mapped to form a collaborative control area; Based on the output characteristics of wind power and photovoltaic power, DC power modulation and the state of charge constraints of energy storage systems, a global objective function including frequency deviation, tie-line power deviation and new energy absorption rate is established to quantify the frequency regulation resource coordination relationship in the coordinated control area. The global objective function is decomposed into local objective functions corresponding to each of the cooperative control regions; Each local objective function is iteratively updated and dualized until the convergence condition is met. Output collaborative control commands for the collaborative control area.

2. The multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM according to claim 1, characterized in that: The step of dividing the hybrid power grid into multiple sub-regions and mapping them to form a collaborative control region includes: Define the indicators of new energy enrichment, comprehensive load, and DC correlation. Based on the aforementioned new energy enrichment index, load comprehensive index, and DC correlation index, a set of regions with similar characteristics is identified.

3. The multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM according to claim 2, characterized in that: The definitions of the new energy enrichment index, load comprehensive index, and DC correlation index include: The renewable energy enrichment index is defined based on the proportion of renewable energy installed capacity and the matching relationship between the average daily output and average daily load of renewable energy. The calculation formula is as follows: ； in, Indicator representing the enrichment of new energy sources. Indicates the weighting coefficient. Indicates the first Sub-regions Subregion The total installed capacity of new energy, Subregion Total installed power generation capacity Subregion The daily output of new energy sources Subregion The average daily load demand; The load composite index is defined by weighting and integrating load density and intraday maximum load fluctuation. The calculation formula is as follows: ； in, This represents the comprehensive load index. Indicates the weighting coefficient. Subregion The maximum daily load, Subregion The area supplied by the power grid, Indicates the first Maximum power fluctuation amplitude of load in each sub-region within 24 hours; The DC correlation index is defined based on the proportion of rated transmission power at DC landing points within the region and the relative density of external tie lines. The calculation formula is as follows: ； in, Indicators representing DC correlation degree Indicates the weighting coefficient. Subregion Rated transmission power of the internal DC landing point, This represents the sum of the rated power of all DC terminals in the entire power grid. This indicates the total number of DC landing points. Subregion The number of communication lines connecting to external areas. Subregion The total number of power grid nodes within the area.

4. A multi-regional cooperative control method for hybrid power grid AGC based on an improved ADMM, as described in claim 2 or 3, characterized in that: The step of identifying a set of regions with similar characteristics based on the new energy enrichment index, the comprehensive load index, and the DC correlation index includes: constructing a three-dimensional feature vector based on the new energy enrichment index, the comprehensive load index, and the DC correlation index, expressed as: ； in, Subregion 3D feature vectors; After removing outliers, valid 3D feature vectors are obtained, and a feature matrix is ​​constructed, expressed as follows: ； in, Represents the characteristic matrix, This represents the first valid three-dimensional feature vector. This represents the second valid three-dimensional feature vector. Indicates the first One effective three-dimensional feature vector, Represents the transpose of a matrix. Represent a A real matrix with 3 rows and 3 columns. Indicates the number of valid three-dimensional feature vectors; The formula for calculating the sum of squares within clusters corresponding to different numbers of clusters is as follows: ； in, Represents the number of clusters The sum of squares within the cluster, Indicates the first The center of each cluster is determined by the mean of all valid three-dimensional eigenvectors within the cluster. Calculations show that Denotes the Euclidean norm; Three cluster center samples were randomly selected as the initial cluster centers. , , Calculate each sub-region The Euclidean distances to the three cluster centers define the sub-regions. Assigning to the nearest cluster is calculated using the following formula: ； in, Indicates the current iteration number. Indicates the first Each cluster passed through Cluster center after the next iteration; Based on the assigned clusters, the center of each cluster is recalculated using the following formula: ； in, Indicates the recalculation of the first Each cluster passed through Cluster center after +1 iteration Indicates the first The number of samples in each cluster; Convergence is determined based on the recalculated center of each cluster; the convergence is achieved when the change in the center of all clusters is less than a threshold. The iteration terminates when the threshold is reached; otherwise, it returns to continue iterating. The calculation formula is: ； in, Indicates the threshold; After the iteration terminates, based on the final cluster centers of the three clusters... The size of the area, combined with its physical meaning, is mapped to the new energy rich area, the load center area, and the DC landing point area.

5. The multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM according to claim 1, characterized in that: The aforementioned wind power and solar power output characteristics include: the actual output of wind power must meet the following requirements: ； in, This indicates the minimum technical output of the wind farm. Indicates the first Each wind farm at any time Actual output This indicates the rated output of the wind farm; The rate of change of wind power output satisfies the ramp rate constraint: ； in, Indicates the automatic power generation control step size. , They represent the first The limits for the upward and downward ramp rates of a wind farm. Indicates the first The wind farm at the previous control time ; output power; The upper and lower limits of photovoltaic power output and the constraints imposed by the ramp rate are as follows: ； in, Indicates the first A photovoltaic power station at time Actual output Indicates the first The rated output of each photovoltaic power station Indicates the first Minimum technical output limit for a photovoltaic power station Indicates time Photovoltaic power output efficiency, , These represent the upward ramp rate limit and the downward ramp rate limit for photovoltaic power plants, respectively. Indicates the first The photovoltaic power station was in the previous control time. ; output power; The DC power modulation includes: the DC line transmission power satisfying the modulation range constraint. ； in, Indicates the first DC lines at time Transmission power, , They represent the first Minimum and maximum transmission power limits for each DC line; The transmission power of AC lines must meet the transmission capacity constraints: ； in, Indicates the first The communication and contact lines are always Transmission power, Indicates the maximum transmission capacity of the AC line; The state-of-charge (POC) constraints of the energy storage system include: a range-constrained POC condition for the energy storage system based on ensuring battery life and frequency regulation reliability, with the constraint form as follows: ； in, Indicates the energy storage system at time The state of charge, , These represent the minimum and maximum allowable values ​​for the state of charge of the energy storage system, respectively. A charge / discharge rate constraint is established based on charge / discharge power, charge / discharge efficiency, and rated capacity. ； in, Indicates the energy storage system at time The charging and discharging power, Indicates the energy storage system at time The charge and discharge efficiency, Indicates the rated capacity of the energy storage system. , These represent the minimum charge / discharge power limit and the maximum charge / discharge power limit of the energy storage system, respectively. This indicates the energy storage system at the previous control time. The state of charge.

6. The multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM according to claim 1, characterized in that: The establishment of a global objective function that includes frequency deviation, tie-line power deviation, and renewable energy absorption rate, and the quantification of the frequency regulation resource coordination relationship in the coordinated control area, include: Based on the integral absolute error, the objective function for frequency deviation is determined, and its expression is: ； in, This represents the objective function for frequency deviation. Indicates time frequency deviation, Indicates the automatic power generation control cycle. The weighting coefficient representing the maximum frequency deviation. This represents the integral absolute error index, reflecting the continuous impact of frequency deviation. This indicates the maximum frequency deviation within the control cycle; Based on the integral-weighted absolute power deviation, the objective function for tie-line power deviation is determined, and its expression is: ； in, This represents the objective function for tie-line power deviation. Indicates time The power deviation of the connecting line, Indicates the time weighting factor. Weighting coefficients representing the range of power fluctuations. This represents the integral-weighted absolute power deviation index, reflecting the weighted cumulative effect of power deviation. This indicates the maximum power deviation within the control cycle; Based on the renewable energy absorption rate, the objective function for the renewable energy absorption rate is determined, and its expression is: ； in, The objective function represents the renewable energy consumption rate. This indicates the total number of new energy power stations. Indicates time No. The actual power consumption of each new energy power station Indicates time No. The theoretical power generation capacity of a new energy power station.

7. A multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM according to claim 6, characterized in that: The establishment of a global objective function that includes frequency deviation, tie-line power deviation and renewable energy absorption rate, and the quantification of the frequency regulation resource coordination relationship in the coordinated control area, also includes: building a three-level hierarchical structure of target layer, criterion layer and scheme layer; The criterion layer judgment matrix is ​​constructed using the scaling method. : ； in, , , , , , All represent judgment matrices Matrix elements; Calculate the judgment matrix Maximum eigenvalue The corresponding feature vectors are normalized to obtain the initial weights of the criterion layer. ; Calculate the consistency index and consistency ratio: ； ； in, Indicators of consistency Indicates the consistency ratio. This represents the random consistency index. To determine the order of a matrix, if the consistency ratio is less than a preset consistency ratio threshold, the consistency of the matrix is ​​considered to be satisfied and the initial weights are reliable. The maximization objective is transformed into a minimization form to achieve aggregate optimization of multiple objectives. A single-objective function is constructed by combining the initial weights, and its expression is: ； in, This represents the global objective function.

8. The multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM according to claim 1, characterized in that: The step of decomposing the global objective function into local objective functions corresponding to each of the cooperative control regions includes: The decomposition expression of the global objective function is: ； In the formula, Represents the global objective function. Indicates the first The local objective function of the control region for each cooperative control region For coordinated control area Local decision variables, For power interaction variables between coordinated control regions; Based on the power interaction variables between the cooperative control regions, a local objective function for each cooperative control region is constructed, expressed as follows: ； in, Indicates the collaborative control area The local objective function, , Indicates the collaborative control area The coefficient matrix, Indicates the collaborative control area The constraint constant term.

9. A multi-regional cooperative control method for hybrid power grid AGC based on improved ADMM as described in claim 8, characterized in that: The iterative update and dual adjustment of each local objective function includes: In the In the next iteration, each cooperative control region is based on the first... The dual variable of the next iteration Solve the local optimization independently and update the local decision variables. The expression is: ； in, Indicates the collaborative control area Local decision variables, Indicates the penalty factor. Represents the transpose of a matrix. Indicates the first The global interaction variable for the next iteration; The global interaction variable is updated based on the local decision variables of each collaborative control region, expressed as follows: ； in, Indicates the first The global interaction variable for the next iteration; Represents a global interactive variable; The dual variable is updated based on the deviation between the local decision variables and the global interaction variables in each collaborative control region. The expression is as follows: ； in, Let represent the dual variable of the d-th iteration; The iteration continues by repeatedly updating local decision variables and updating global interaction variables until a preset convergence threshold is reached, at which point the iteration terminates. The expression is: ； in, This indicates the preset convergence threshold; For the first The first collaborative control area in the first Local decision variables after the next iteration; Based on the dynamic adjustment of the penalty factor according to the fluctuation degree of new energy power output and the constraint tightness of the DC system, the standard deviation of the new energy power output prediction error is used. The formula for quantifying the fluctuation of new energy power output is as follows: ； in, Indicates the number of prediction periods. , They represent Real-time comparison of predicted and actual renewable energy output; Define DC power constraint tightness coefficient The formula for quantifying DC constraint tightness is as follows: ； in, Indicates the actual DC transmission power. , Indicates the upper and lower limits of DC power modulation; Based on the standard deviation of the new energy output prediction error and the DC power constraint tightness coefficient, the penalty factor is dynamically adjusted to achieve an adaptive step size. The calculation formula is as follows: ； in, Indicates adaptive step size, Indicates the initial penalty factor. This represents the maximum standard deviation of the prediction error for new energy power output. Indicates the first The standard deviation of the new energy output prediction error in the next iteration Indicates the step size and the number of iterations; Indicates the DC power constraint tightness coefficient; The prediction error compensation term is calculated based on the deviation between the predicted and actual values ​​from the Long Short-Term Memory network. The calculation formula is as follows: ； in, This represents the prediction error compensation term. Indicates the future Forecast values ​​of renewable energy output for the specified time period; Embed the prediction error compensation term into the collaborative control region The local objective function is expressed as: ； in, Indicates the first term after embedding the prediction error compensation term. The modified local objective function of the cooperative control region Indicates the first The original local objective function of each cooperative control region Indicates the compensation coefficient. Indicates the collaborative control area New energy output decision variables; Based on the frequency modulation capacity input, response speed, and control effect of each coordinated control area, a regional contribution index is defined, and the calculation formula is as follows: ； in, Indicates the collaborative control area The regional contribution index, Indicates the collaborative control area The frequency modulation capacity input, Indicates the total global frequency modulation capacity. Indicates the collaborative control area frequency modulation response time, Indicates the collaborative control area The improvement in the global objective function. , , Indicates the weighting coefficient. This represents the total global improvement. A contribution penalty term is added to the global objective function, calculated using the following formula: ； ； in, This represents the global objective function after adding a contribution penalty term. Indicates the penalty coefficient. This indicates the average contribution.

10. A multi-regional cooperative control method for hybrid power grid AGC based on an improved ADMM, as described in claim 9, is characterized in that: The convergence conditions include: Define the convergence function as follows: ； in, Represents the Lyapunov function. , , These represent the collaborative control areas. The optimal solution for local decision variables, global interaction variables, and dual variables; Substituting the iterative update formula of the alternating direction multiplier method, which includes adaptive step size, error compensation term, and contribution penalty term, into the Lyapunov function... The convergence condition is obtained: Adaptive step size Convergence condition: ； in, , These represent the minimum penalty factor and the maximum penalty factor, respectively. Convergence conditions for error compensation and contribution penalty: ； in, Indicates the compensation coefficient. This represents the penalty coefficient.

Images