Multi-resource coordinated operation method and system for improving distributed photovoltaic bearing capacity

Abstract

The invention discloses a multi-resource coordinated operation method and system for improving distributed photovoltaic bearing capacity, and the method comprises the steps: firstly collecting power grid load data, carrying out the clustering analysis of the power grid load data through employing an improved fuzzy C-means clustering algorithm, and recognizing the power load characteristics of a power distribution network; then constructing an active-reactive power hybrid control model of a power grid system based on the identified power load characteristics of the power distribution network; and finally, solving the active-reactive power hybrid control model of the power grid system to obtain an active power and reactive power coordination control strategy for improving the distributed photovoltaic bearing capacity of the power distribution network. Through coordination of the active power control strategy and the reactive power control strategy, under the condition of considering daily load and photovoltaic power generation uncertainty, the photovoltaic access capacity of the power distribution network is remarkably improved, the bearing capacity of the power distribution network for high-permeability photovoltaic power generation is improved, and therefore operation and planning of a power distribution system are optimized.

Description

Technical Field

[0001] The present invention belongs to the field of power system engineering, and particularly relates to a multi-resource coordinated operation method and system for improving the bearing capacity of distributed photovoltaic power generation. Background Art

[0002] With the deepening of the construction of new distribution networks, the penetration rate of solar photovoltaic resources in distribution networks has been continuously increasing. The maximum number of solar photovoltaics connected to the low-voltage distribution network system depends on the bearing capacity of the low-voltage distribution network. Among them, the bearing capacity HC refers to the maximum solar photovoltaic PV power generation capacity that the distribution network can carry without violating the operating constraints of the distribution network system. If the power generation capacity of the connected solar photovoltaic resources exceeds the bearing capacity of the low-voltage distribution network system, it may bring negative impacts such as over-limit voltage amplitude, voltage imbalance, increased losses, harmonics, and feeder overload to the distribution network system; moreover, solar photovoltaic power generation depends on daily solar radiation and climatic conditions, and its power generation and load have uncertainties, and this uncertainty must also be considered in the determination of the bearing capacity of the distribution network system. Summary of the Invention

[0003] The purpose of the present invention is to provide a multi-resource coordinated operation method and system for improving the bearing capacity of distributed photovoltaic power generation in view of the above problems existing in the prior art.

[0004] To achieve the above purpose, the technical solution of the present invention is as follows:

[0005] In the first aspect, the present invention proposes a multi-resource coordinated operation method for improving the bearing capacity of distributed photovoltaic power generation, including:

[0006] S1. Collect grid load data, and use an improved fuzzy C-means clustering algorithm to perform clustering analysis on the grid load data to identify the power load characteristics of the distribution network;

[0007] S2. Based on the identified power load characteristics of the distribution network, construct an active-reactive power hybrid control model for the grid system. The active power control is achieved by optimizing the use of transferable and interruptible flexible loads, and the reactive power control is achieved by optimizing the layout of static var compensators;

[0008] S3. Solve the active-reactive power hybrid control model of the grid system to obtain an active power and reactive power coordinated control strategy for improving the bearing capacity of the distribution network for distributed photovoltaic power generation. The distributed photovoltaic power generation bearing capacity is the maximum capacity of distributed photovoltaics that the grid can accept under the conditions that no device is continuously overloaded and the voltage, short-circuit current, and harmonics of any node do not exceed the standards.

[0009] In the above S2, the active-reactive power hybrid control model constructs the following objective function with the goal of maximizing the access capacity of photovoltaic resources and minimizing the total expected cost of the power grid system:

[0010] Min{-w HC ×HC + w BC ×EC};

[0011]

[0012] In the above formula, w HC is the weight coefficient of the photovoltaic access capacity, HC is the photovoltaic access capacity, w EC is the weight coefficient of the total expected cost of the system, and EC is the total expected cost of the system; is the photovoltaic capacity installed at node i, ρ s is the occurrence probability of scenario s, is the variable operating cost of the static var compensator, is the binary decision variable indicating whether a static var compensator is installed at node i, is the reactive power support of the static var compensator at node i at time t under scenario s, C INC is the incentive payment cost for the interruptible load, is the load power curtailed at node i at time t under scenario s, is the cost of purchasing electricity from the market at node i at time t under scenario s, is the active power purchased from the market at node i at time t under scenario s.

[0013] In the above S2, the constraint conditions of the active-reactive power hybrid control model include:

[0014] Active-reactive power constraint:

[0015]

[0016] Voltage balance constraint:

[0017]

[0018] Linearization constraint:

[0019]

[0020] 0 ≤ ΔP i，i′f，t，s ≤ ΔS i，i′ ;

[0021] 0 ≤ ΔQ i，i′f，t，s ≤ ΔS i，i′ ;

[0022]

[0023] Photovoltaic constraint:

[0024]

[0025] Static var compensator constraint:

[0026]

[0027] Flexible load constraint:

[0028]

[0029] In the above formula, is the active power purchased by node i from the market at time t under scenario s, is the photovoltaic output power on node i at time t under scenario s, is the active power flow in the downstream direction, is the active power flow in the upstream direction, R i，i′ is the resistance between nodes i and i′, I2 i，i′t，s is the square of the current between nodes i and i′ at time t under scenario s, is the reactive power purchased by node i from the market at time t under scenario s, is the reactive power support of the static var compensator on node i at time t under scenario s, is the reactive power flow in the downstream direction, is the reactive power flow in the upstream direction, X i，i′ is the reactance between nodes i and i′, is the load reactive power on node i at time t, V Rated is the rated voltage, is the maximum current between nodes i and iτ; V2 i，t，s is the square of the voltage on node i at time t under scenario s, V2 i′t，s is the square of the voltage on node iτ at time t under scenario s, is the square of the impedance between nodes i and iτ; is the square of the rated voltage on node i, f is the block, ΔS i，i′ is the upper limit of the piecewise linearization of the active power and reactive power between nodes i and i′, ΔP i，i′f，t，s is the active power of the f-th block between nodes i and i′ at time t under scenario s, ΔQ i，i′f，t，s is the reactive power of the f-th block between nodes i and i′ at time t under scenario s, F is the number of linearization blocks; is the installed photovoltaic capacity on node i, is the output factor of the photovoltaic at time t under scenario s; is the capacity of the static var compensator installed at node i, is the maximum capacity of the static var compensator at node i, is a binary decision variable indicating whether to install a static var compensator at node i, is the total number of static var compensators allowed to be installed at node i, The absolute value of is an auxiliary variable of the static var compensator at node i; is the maximum proportion of transferable load at node i, is the initial load demand of node i at time t, is the transferable load demand of node i at time t under scenario s, is the load shedding of node i at time t under scenario s, α curt is the maximum proportion of interruptible load at node i, is the final load of node i at time t under scenario s.

[0030] The said S1 includes:

[0031] S11. Obtain the power grid node load data set X = {x1, x2,..., x n}, initialize the settings of the improved fuzzy C - means clustering algorithm, randomly select C = {c1, c2,..., c k} power grid load data as the initial clustering centroids, k is the total number of specified clustering centroids, set the maximum number of iterations of the clustering algorithm, and initialize the iteration γ = 1;

[0032] S12. Use the Pearson correlation coefficient and Euclidean distance to calculate the local density, and optimize the distance measure between the data point and the clustering centroid through the local density;

[0033]

[0034]

[0035] In the above formula, is the distance measure between the data point x i and the clustering centroid , dx i , is the Euclidean distance between the data point X i and the clustering centroid , is the local density of the data point X i , N a (X i ) is the set of data points closest to the data point X i , d(X i , Xj ) is the optimized distance between data points X i and X j , and d(X j , v) is the distance between data points X i and X j under other metrics, which is used to measure the similarity between data points. is the Euclidean distance between data points X i and X j , and δ(X i , X j ) is the Pearson correlation coefficient between data points X i and X j . n is the number of sample data, x is is the load value of the i-th data point at the S-th observation, and x js is the load value of the j-th data point at the S-th observation;

[0036] S13. Using the distance measure between the data point and the cluster centroid, calculate the membership degree between each data point and the cluster centroid;

[0037]

[0038] In the above formula, is the membership degree between data point X i and the cluster centroid , is the distance from data point X i to other cluster centroids , and m is the fuzziness parameter;

[0039] S14. Update the position of each cluster centroid using the calculation result of the membership degree to obtain a new cluster centroid:

[0040]

[0041] In the above formula, is the updated cluster centroid;

[0042] S15. Judge whether the membership degree error between two iterations reaches the stop condition or the maximum number of iterations. If the stop condition or the maximum number of iterations is reached, the iteration stops and the clustering result is output. If the stop condition or the maximum number of iterations is not reached, let γ = γ + 1, and return to S12 to continue the iteration;

[0043] The stop condition is:

[0044]

[0045] In the above formula, ∈ is the membership degree error between two iterations, and τ is the error threshold.

[0046] In a second aspect, the present invention proposes a multi - resource coordinated operation system for enhancing the carrying capacity of distributed photovoltaic power generation, which includes a load characteristic clustering module, a model construction module, and a model solution module;

[0047] The load characteristic clustering module is used to collect grid load data, perform clustering analysis on the grid load data by using an improved fuzzy C - means clustering algorithm, and identify the electrical load characteristics of the distribution network;

[0048] The model construction module is used to construct an active - reactive power hybrid control model of the grid system based on the identified electrical load characteristics of the distribution network. The active power control is achieved by optimizing the use of transferable and interruptible flexible loads, and the reactive power control is achieved by optimizing the layout of static var compensators;

[0049] The model solution module is used to solve the active - reactive power hybrid control model of the grid system, and obtain the active power and reactive power coordinated control strategies for improving the carrying capacity of the distribution network for distributed photovoltaic power generation. The carrying capacity of distributed photovoltaic power generation is the maximum capacity of distributed photovoltaic power that the grid can accept under the conditions that no device is continuously overloaded and the voltage, short - circuit current, and harmonics at any node do not exceed the standards.

[0050] The model construction module includes an objective function construction unit;

[0051] The objective function construction unit is used to construct the following objective function with the goal of maximizing the access capacity of photovoltaic resources and minimizing the total expected cost of the grid system:

[0052] Min{-w HC ×HC + w EC ×EC};

[0053]

[0054] In the above formula, w HC is the weight coefficient of the photovoltaic access capacity, HC is the photovoltaic access capacity, w EC is the weight coefficient of the total expected cost of the system, and EC is the total expected cost of the system; is the photovoltaic capacity installed at node i, ρ s is the occurrence probability of scenario s, is the variable operating cost of the static var compensator, is a binary decision variable indicating whether a static var compensator is installed at node i, is the reactive power support of the static var compensator at node i at time t in scenario s, C INC is the incentive payment cost of the interruptible load, is the reduced load power on node i at time t under scenario s, is the cost of node i purchasing electricity from the market at time t under scenario s, is the active power purchased by node i from the market at time t under scenario s.

[0055] The model construction module further includes an active-reactive power constraint construction unit, a voltage balance constraint construction unit, a linearization constraint construction unit, a photovoltaic constraint construction unit, a static var compensator constraint construction unit, and a flexible load constraint construction unit;

[0056] The active-reactive power constraint construction unit is used to construct the following active-reactive power constraints:

[0057]

[0058] The voltage balance constraint construction unit is used to construct the following voltage balance constraints:

[0059]

[0060] The linearization constraint construction unit is used to construct the following linearization constraints:

[0061]

[0062] 0 ≤ ΔP i，i′f，t，s s ≤ ΔS i，i′ ;

[0063] 0 ≤ ΔQ i，i′f,t,s ≤ ΔS i，i′ ;

[0064]

[0065] The photovoltaic constraint construction unit is used to construct the following photovoltaic constraints:

[0066]

[0067] The static var compensator constraint construction unit is used to construct the following static var compensator constraints:

[0068]

[0069] The flexible load constraint construction unit is used to construct the following flexible load constraints:

[0070]

[0071] In the above formula, is the active power purchased by node i from the market at time t under scenario s, is the photovoltaic output power at node i at time t under scenario s, is the active power flow in the downstream direction, is the active power flow in the upstream direction, R i，i′ is the resistance between nodes i and i′, I2 i，i′t，s is the square of the current between nodes i and i′ at time t under scenario s, is the reactive power purchased by node i from the market at time t under scenario s, is the reactive power support of the static var compensator at node i at time t under scenario s, is the reactive power flow in the downstream direction, is the reactive power flow in the upstream direction, X i，i′ is the reactance between nodes i and i′, is the load reactive power at node i at time t, V Rated is the rated voltage, is the maximum current between nodes i and i′; V2 i，t，s is the square of the voltage at node i at time t under scenario s, V2 i′t，s is the square of the voltage at node i′ at time t under scenario s, is the square of the impedance between nodes i and i′; is the square of the rated voltage at node i, f is the block, ΔS i，i′ is the upper limit of the piecewise linearization of the active power and reactive power between nodes i and i′, ΔP i，i′f，t，s is the active power of the f - th block between nodes i and i′ at time t under scenario s, ΔQ i，i′f，t，s is the reactive power of the f - th block between nodes i and i′ at time t under scenario s, F is the number of linearization blocks; is the installed photovoltaic capacity at node i, is the output factor of the photovoltaic at time t under scenario s; is the capacity of the static var compensator installed at node i, is the maximum capacity of the static var compensator at node i, is the binary decision variable indicating whether a static var compensator is installed at node i, is the total number of static var compensators allowed to be installed at node i, The absolute value of is the auxiliary variable of the static var compensator at node i; is the maximum proportion of transferable load at node i, is the initial load demand at node i at time t, is the transferable load demand at node i at time t under scenario s, The reduced load of node i at time t under scenario s, α curt is the maximum proportion of the interruptible load on node i, and the final load of node i at time t under scenario s.

[0072] The load characteristic clustering module includes a clustering initialization unit, a distance measure optimization unit, a membership degree calculation unit, a clustering centroid update unit, and a loop iteration unit;

[0073] The clustering initialization unit is used to obtain the power grid node load data set X = {x1, x2,..., x n}, initialize the settings of the improved fuzzy C - means clustering algorithm, randomly select C = {c1, c2,..., c k} power grid load data as the initial clustering centroids, k is the total number of specified clustering centroids, set the maximum number of iterations of the clustering algorithm, and initialize the iteration γ = 1;

[0074] The distance measure optimization unit is used to calculate the local density using the Pearson correlation coefficient and the Euclidean distance, and optimize the distance measure between the data point and the clustering centroid through the local density;

[0075]

[0076] In the above formula, is the distance measure between the data point X i and the clustering centroid dx i , is the Euclidean distance between the data point X i and the clustering centroid , is the local density of the data point X i N a (X i ) is the set of data points closest to the data point X i , d(X i , X j ) is the optimized distance between the data points X i and X j , d(X j , v) is the distance between the data points X i and X j under other metric standards, used to measure the similarity between data points, is the Euclidean distance between the data points X i and X j , δ(X i , X j ) is the distance between the data points X i and X jThe Pearson correlation coefficient between them, n is the number of sample data, and x is is the load value of the i-th data point at the s-th observation, and x js is the load value of the j-th data point at the s-th observation;

[0077] The membership degree calculation unit is used to calculate the membership degree between each data point and the clustering centroid by using the distance measure between the data point and the clustering centroid;

[0078]

[0079] In the above formula, is the data point X i and the clustering centroid between the membership degrees, is the data point X i to other clustering centroids distance, m is the fuzziness parameter;

[0080] The clustering centroid update unit is used to update the position of each clustering centroid by using the calculation result of the membership degree to obtain a new clustering centroid:

[0081]

[0082] In the above formula, is the updated clustering centroid;

[0083] The loop iteration unit is used to determine whether the membership degree error between two iterations reaches the stop condition or the maximum number of iterations. If the stop condition or the maximum number of iterations is reached, the iteration stops and the clustering result is output. If the stop condition or the maximum number of iterations is not reached, let γ = γ + 1 and return to the distance measure optimization unit to continue the iteration;

[0084] The stop condition is:

[0085]

[0086] In the above formula, ∈ is the membership degree error between two iterations, and τ is the error threshold.

[0087] In the third aspect, the present invention proposes a multi-resource coordinated operation device for enhancing the bearing capacity of distributed photovoltaics, including a processor and a memory;

[0088] The memory is used to store computer program code and transmit the computer program code to the processor;

[0089] The processor is used to execute the foregoing method for enhancing the grid bearing capacity based on coordinated active and reactive power according to the instructions in the computer program code.

[0090] Fourthly, the present invention provides a computer storage medium, on which a computer program is stored;

[0091] When the computer program is executed by a processor, the steps of the above-mentioned multi-resource coordinated operation method for improving the bearing capacity of distributed photovoltaic are realized.

[0092] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0093] 1. The present invention provides a multi-resource coordinated operation method and system for improving the bearing capacity of distributed photovoltaic. The method first collects grid load data, and uses an improved fuzzy C-means clustering algorithm to cluster and analyze the grid load data to identify the power load characteristics of the distribution network; then, based on the identified power load characteristics of the distribution network, an active-reactive power hybrid control model of the grid system is constructed. The active power control is realized by optimizing the use of transferable and interruptible flexible loads, and the reactive power control is realized by optimizing the layout of static var compensators; finally, by solving the active-reactive power hybrid control model of the grid system, an active power and reactive power coordinated control strategy for improving the bearing capacity of the distribution network to distributed photovoltaic is obtained. The bearing capacity of distributed photovoltaic is the maximum capacity of distributed photovoltaic that the grid can accept under the condition that any device does not overload continuously and the voltage, short-circuit current, and harmonics at any node do not exceed the standard. On the one hand, when constructing the active-reactive power hybrid control model of the grid system, the method particularly considers the expected cost, which can help the system effectively improve the technical stability of the grid system under the conditions of load demand, photovoltaic power generation fluctuations, and operation cost uncertainties, ensure that key performances such as voltage level and power balance are optimized, avoid unnecessary additional expenses, reduce operation and investment costs, make photovoltaic access and system configuration more economical, and improve the robustness of the model; on the other hand, by coordinating the active power control strategy and the reactive power control strategy, the method can significantly increase the photovoltaic access capacity of the distribution network and improve the bearing capacity of the distribution network to high-penetration photovoltaic power generation under the consideration of daily load and photovoltaic power generation uncertainties, thereby optimizing the operation and planning of the distribution system.

[0094] 2. The present invention proposes a multi - resource coordinated operation method and system for enhancing the bearing capacity of distributed photovoltaic power generation. This method uses an improved fuzzy C - means clustering algorithm to perform clustering analysis on grid load data. On the one hand, by combining the Euclidean distance with the Pearson correlation coefficient, the calculation of local density is optimized, which helps to more accurately determine the clustering centers of data points, classify load types, and improve the accuracy of load classification. Especially in the clustering analysis of load characteristics, it can more effectively handle the correlation problems in high - dimensional data, enhance the accuracy and convergence speed of clustering, and contribute to power system dispatching and demand - side management. On the other hand, in each iteration, the membership degree and clustering centroids are updated according to the optimized distance measure, which can ensure the accuracy of the final clustering centroids. Especially in the processing of multi - period power load data, it can better reduce the influence of the time - series and volatility of load data, making the clustering results more capable of reflecting the actual fluctuations of the load. For different load characteristics, such as peak and valley loads, the identification is more accurate. Through the improved iterative clustering process, convergence can be completed in a shorter time, reducing the occurrence of local optimal solutions, thus making the clustering results more stable and enhancing the efficiency and accuracy of clustering. Brief Description of the Drawings

[0095] Figure 1 It is the overall flowchart of the method described in the present invention.

[0096] Figure 2 It is the structure diagram of the system described in the present invention.

[0097] Figure 3 It is the structure diagram of the equipment described in Embodiment 3. Detailed Embodiments

[0098] The following further elaborates on the present invention in detail in combination with the detailed embodiments and the drawings.

[0099] The present invention proposes a multi - resource coordinated operation method and system for enhancing the bearing capacity of distributed photovoltaic power generation. Considering the uncertainties of photovoltaic power generation and load demand, multiple scenarios are set to simulate the operation conditions under different seasons, different time periods, and peak and valley loads. Through technical performance indicators such as voltage deviation, power loss, photovoltaic access capacity, and system stability, as well as economic performance indicators such as expected cost, operation cost, and investment return, the active power control strategy and reactive power control strategy are coordinated to construct and implement the hybrid optimization of the two powers in the distribution network system, and improve the bearing capacity of the distribution network for high - penetration photovoltaic power generation.

[0100] Embodiment 1:

[0101] In this embodiment, the IEEE 15-node and 33-node distribution network systems are taken as the research objects. According to different photovoltaic penetration rates and load scenarios, a method for improving the grid carrying capacity by coordinating active and reactive power is adopted to simulate the operation of the distribution network system.

[0102] As Figure 1 shown, a multi-resource coordinated operation method for improving the carrying capacity of distributed photovoltaics is carried out in the following steps:

[0103] 1. Conduct a demand analysis of the distribution network system, determine the key nodes and load points in the distribution network, analyze the existing photovoltaic power generation access situation and potential access capacity, and determine the number and initial positions of static var compensators to be installed;

[0104] Clarify the identification business problems and define the optimization objectives of the distribution network system;

[0105] The requirements of the distribution network system include clarifying the functions and non-functional requirements that the system needs to achieve according to the overall objectives of the project. The main objectives are to optimize the access capacity of photovoltaics, the scheduling of flexible loads, and the configuration of static var compensators to improve the stability and economy of the power grid. Therefore, the identification business problems of the distribution network system include determining the main problems in the current operation of the distribution network system, such as insufficient photovoltaic access capacity, large voltage fluctuations, high system operation costs, etc., and clarifying the impact of these problems on system performance and economic benefits; the defined optimization objectives are to clarify the demand objectives, such as minimizing the expected cost, improving the photovoltaic access capacity, enhancing voltage stability, etc.;

[0106] Determine the inputs and outputs of the distribution network system;

[0107] The input data analysis includes load data, such as historical load data, load curves, seasonal fluctuations, etc., photovoltaic power generation data, such as predicted values of photovoltaic power generation, volatility and uncertainty analysis, etc., equipment parameters, such as configuration parameters of static var compensators, shiftable and interruptible situations of flexible loads, etc., and power grid operation data, such as voltage, power, losses, etc.; the output target analysis includes the optimization results of photovoltaic access capacity and location, flexible load scheduling strategies for shiftable loads and interruptible loads, system expected costs, economic benefits, and technical performance indicators;

[0108] Confirmation of the requirements of relevant parties;

[0109] Clarify the requirements of the operator in the distribution network system, such as minimizing operating costs, maximizing revenues, etc., the requirements of users, such as ensuring that the electricity needs of users can be reliably met, voltage stability, reliable power supply, etc., and the regulatory requirements, such as meeting the relevant regulations, standards and technical requirements for the operation of the power system. Conduct a priority ranking of the requirements. For example, based on business needs, technical needs and economic benefits, rank the priorities of the requirements, and give priority to handling the requirements that have the greatest impact on system performance, cost-effectiveness and stability, so as to ensure that key issues are resolved first.

[0110] 2. Based on the results of the demand analysis of the distribution network system, collect grid load data, and use the improved fuzzy C-means clustering algorithm to perform clustering analysis on the grid load data to identify the power load characteristics of the distribution network.

[0111] System initialization: Obtain the topological structure and related parameters of the distribution network, collect historical data of photovoltaic power generation and grid load data, and determine the types and distribution of flexible loads.

[0112] First, preprocess the original grid load data, including:

[0113] Suppose there is one year's worth of grid load data, collected hourly every day, forming a 366×24 matrix A, which contains 366 power load curves.

[0114] Use the following formula to calculate the average value and standard deviation of the grid load data for each day in matrix A, and eliminate the load data with errors exceeding 3 times the standard deviation to form a new matrix B:

[0115]

[0116] In the above formula, is the average value of the data in the s-th row of the original load data matrix, that is, the average value of the grid load data for a certain day. This symbol is usually used in the data preprocessing stage to calculate the average value of each row of data for subsequent normalization processing or outlier removal. m is the number of samples, and A′ js is the data in the j-th column of the s-th row of the original load data matrix, and E js is the remaining error, and θ is the standard deviation.

[0117] Normalize the matrix B after eliminating the error data, so that the grid load data in matrix B is converted into a normal distribution matrix C with zero mean and unit variance, eliminating the dimensional difference of different load data:

[0118]

[0119] In the above formula, x scaleis the normalized data value, also known as standardized data, x is the data in the new matrix, μ is the mean of the data in the new matrix, and sa is the standard deviation of the data in the new matrix; the purpose of normalization is to convert data with different dimensions into the same standard for processing such as cluster analysis and model training;

[0120] Extract key feature indicators of the power grid load data in matrix C, such as daily load rate, peak period load rate, etc., and perform dimensionality reduction processing to form a new feature matrix D;

[0121] Then, perform load analysis on the preprocessed power grid load data feature matrix D. By improving the fuzzy C-means clustering algorithm, achieve precise cluster analysis of the power load data and identify the power load characteristics of the distribution network;

[0122] Obtain the power grid node load data set X = {x1, x2,..., x n}, initialize the settings of the improved fuzzy C-means clustering algorithm, randomly select C = {c1, c2,..., c k} power grid load data as the initial cluster centroids, k is the total number of specified cluster centroids, set the maximum number of iterations of the clustering algorithm, and initialize the iteration γ = 1;

[0123] Use the Pearson correlation coefficient and Euclidean distance to calculate the local density, and optimize the distance measure between the data points and the cluster centroids through the local density;

[0124]

[0125] In the above formula, is the distance measure between the data point X i and the cluster centroid , dx i , is the Euclidean distance between the data point X i and the cluster centroid , is the local density of the data point X i , N a (X i ) is the set of data points closest to the data point X i , d(X i , X j ) is the optimized distance between the data point X i and X j , d(X j , v) is the distance between the data point X i and X j under other metric standards, used to measure the similarity between data points, is the data point X i and X jThe Euclidean distance between δ(X i , X j ) is the Pearson correlation coefficient between data points X i and X j . n is the number of sample data, x is is the load value of the i-th data point at the S-th observation, and x js is the load value of the j-th data point at the S-th observation;

[0126] Using the distance measure between the data point and the cluster centroid, calculate the membership degree between each data point and the cluster centroid;

[0127]

[0128] In the above formula, is the membership degree between data point X i and the cluster centroid , is the distance from data point X i to other cluster centroids , and m is the fuzziness parameter;

[0129] Use the calculation result of the membership degree to update the position of each cluster centroid to obtain a new cluster centroid:

[0130]

[0131] In the above formula, is the updated cluster centroid;

[0132] Judge whether the membership degree error between two iterations reaches the stop condition or the maximum number of iterations. If the stop condition or the maximum number of iterations is reached, the iteration stops and the clustering result is output. If the stop condition or the maximum number of iterations is not reached, let γ = γ + 1, return the distance measure between the locally density-optimized data point and the cluster centroid, and continue the iteration;

[0133] The stop condition is:

[0134]

[0135] In the above formula, ∈ is the membership degree error between two iterations, and τ is the error threshold;

[0136] Finally, use the clustering evaluation index to evaluate the clustering recognition result of the power load characteristics of the distribution network;

[0137]

[0138] In the above formula, J CH is the CH clustering evaluation index, J XB is the XB clustering evaluation index, Bk is the between-class scatter, and W k is the within-class scatter; the larger the CH index and the smaller the XB index, the better the clustering effect. When evaluating the clustering effect, these two indicators do not have a strictly fixed range. Their values mainly depend on the data distribution, clustering algorithm, and the number of clusters, and need to be analyzed in combination with the characteristics of specific data;

[0139] Among them, the between-class scatter is calculated using the following formula:

[0140]

[0141] In the above formula, k is the number of clustering categories, and n i is the number of samples in the i-th clustering category, and d(C i , C all ) is the distance between the class center of the i-th clustering category and the overall data center; the between-class scatter represents the degree of dispersion between different classes, usually calculated by the distance between the center of each class and the center of the overall data set. The larger the between-class scatter, the higher the separation degree between different classes, and the better the clustering effect;

[0142] The within-class scatter is calculated using the following formula:

[0143]

[0144] In the above formula, d(x j , C i ) is the distance between the sample point X i in the i-th clustering category and the class center c j of the i-th clustering category; the within-class scatter represents the degree of tightness between data points in each class, usually obtained by calculating the distance between each sample within the class and the clustering center of the class. The smaller the within-class scatter, the higher the degree of aggregation within the class, and the better the clustering effect;

[0145] If the CH index is small and the XB index is large, it indicates that the clustering effect is not good, and improvements can be made from several aspects, including: optimizing the distance measure: In addition to the traditional Euclidean distance, try to combine other distance metric methods, such as Pearson correlation coefficient, Mahalanobis distance, etc., to better reflect the characteristics of the data. For example, in time series data, the Pearson correlation coefficient can better reflect the correlation between different time points; combining weights: Set weights for different feature dimensions to enhance the contribution of features to distance measurement; choosing the appropriate number of clusters: Too few clusters may lead to excessive within-class differences and inaccurate clustering, while too many clusters will result in small differences between classes. Therefore, the optimal number of clusters can be selected through parameter tuning, such as the Elbow Method or Silhouette Score; improving the clustering algorithm: If the traditional fuzzy C-means clustering algorithm performs poorly, especially when the data has a strong non-linear structure, other clustering algorithms can be tried, such as K-means, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), or hierarchical clustering; using heuristic algorithms: Combining with clustering, such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), etc., can help improve the clustering effect; feature extraction and selection: If the features of the original data are not sufficient to characterize the classification effect, the clustering effect can be enhanced by adding features or dimensionality reduction. Common methods include Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), etc.; data normalization: Ensure that all features are on the same scale to avoid some features overly influencing the clustering effect due to different scales; increasing the data sample size: If the data volume is too small, the model may not be able to well reflect the true structure of the data. Appropriately increasing the data sample size can improve the clustering effect;

[0146] Compared with the traditional fuzzy C-means clustering algorithm, this method improves the clustering accuracy and convergence speed by improving the distance measurement method, combining the Pearson correlation coefficient and the Euclidean distance; the improved fuzzy C-means clustering algorithm optimizes the calculation of local density using the Euclidean distance combined with the Pearson correlation coefficient, which helps to more accurately determine the clustering centers of data points and divide the load types. Especially in the clustering analysis of load characteristics, it can more effectively handle the correlation problems in high-dimensional data; in each iteration, the membership degree and clustering centroids are updated according to the optimized distance measure, which can ensure the accuracy of the final clustering centroids. Especially in the processing of multi-period power load data, it can better reduce the influence of the time series and volatility of load data, making the clustering results better reflect the actual fluctuations of the load, and being more accurate in identifying different load characteristics, such as peak and valley loads; through the improved iterative clustering process, it can converge in a shorter time, reduce the occurrence of local optimal solutions, thereby making the clustering results more stable and improving the efficiency and accuracy of clustering.

[0147] 3. Based on the identified power load characteristics of the distribution network, construct an active-reactive power hybrid control model for the power grid system;

[0148] Considering the uncertainties of photovoltaic power generation and load demand, set multiple scenarios to simulate the operation under different seasons, different time periods, and peak and valley loads. Evaluate the performance of the system under different load and photovoltaic power generation scenarios through technical performance indicators such as voltage deviation, power loss, photovoltaic access capacity, and system stability, as well as economic performance indicators such as expected cost, operating cost, and return on investment. By coordinating the active power control strategy and the reactive power control strategy, construct and implement the hybrid optimization of the two powers in the distribution network system to improve the carrying capacity of the distribution network for high-penetration photovoltaic power generation, thereby optimizing the operation and planning of the distribution system;

[0149] Among them, for active power control, flexible loads are utilized, including transferable loads such as electric vehicle charging and air-conditioning loads, and interruptible loads such as industrial loads and non-critical household appliances, to achieve the optimal management function of active power; develop load management strategies to optimize active power by adjusting the load distribution, deploy a load management system to achieve real-time monitoring and scheduling of flexible loads, and dynamically adjust the load distribution according to the changes in load demand and photovoltaic power generation;

[0150] Reactive power control usually requires installing equipment and incurring additional investment costs. By optimizing the layout of the static var compensator (SVC), achieve the optimal distribution function of reactive power; use optimization algorithms such as mixed-integer linear programming (MILP) to determine the optimal installation location and capacity of the static var compensator, considering minimizing the investment and operating costs of the static var compensator equipment while meeting the voltage stability and reactive power requirements. According to the optimization results, install static var compensator equipment at key nodes in the distribution network, and set and debug the static var compensator equipment to ensure its normal operation;

[0151] The active-reactive power hybrid control model constructs the following objective function with the goal of maximizing the access capacity of photovoltaic resources and minimizing the total expected cost of the power grid system:

[0152] Min{-w HC ×HC + w EC ×EC};

[0153]

[0154] In the above formula, w HC is the weight coefficient of the photovoltaic access capacity, HC is the photovoltaic access capacity, that is, the sum of the photovoltaic capacities installed on all nodes, w EC is the weight coefficient of the total expected cost of the system, and EC is the total expected cost of the system; is the photovoltaic capacity installed at node i, ρs is the occurrence probability of scenario s, is the variable operating cost of the static var compensator, is a binary decision variable indicating whether to install a static var compensator at node i, is the reactive power support of the static var compensator at node i at time t under scenario s, C INC is the incentive payment cost for interruptible loads, is the curtailed load power at node i at time t under scenario s, is the cost of purchasing electricity from the market at node i at time t under scenario s, is the active power purchased from the market at node i at time t under scenario s;

[0155] The expected cost is particularly considered in the objective function. The expected cost is modeled and optimized for the uncertainty of the system operating cost when optimizing the distributed PV access capacity and the static var compensator configuration, aiming to reduce the total cost while enhancing the PV access capacity of the system. The optimization of the expected cost can help the power system reduce the operating cost under the condition of the uncertainty of the operating cost. Through the optimization of the expected cost, the power grid system can reduce the operating and investment costs while achieving the technical goals, making the PV access and system configuration more economically viable and ensuring the economy of the power grid system; The optimization of the expected cost also considers the uncertainties of PV power generation and load demand, and can help the system maintain a more stable performance when facing load demand and PV power generation fluctuations, avoid unnecessary additional expenses, effectively enhance the technical stability of the power grid system while ensuring the optimal cost, ensure the optimization of key performances such as voltage level and power balance, and improve the robustness of the model;

[0156] The constraint conditions of the active-reactive power hybrid control model need to meet the technical constraints of the distribution network system, including the system voltage stability requirements to ensure that the voltage deviation is within the allowable range, the power balance requirements to ensure the active and reactive power balance of each node, and the PV access limit requirements considering the technical feasibility of the access capacity and location; economic constraints, including the investment budget limit considering cost factors such as static var compensator configuration and equipment investment, and the operating cost limit to minimize the energy loss and maintenance cost in the system operation; including active-reactive power constraints, voltage balance constraints, linearization constraints, PV constraints, static var compensator constraints, and flexible load constraints;

[0157] The active-reactive power constraints include:

[0158] The active power balance equation constraint to ensure the active power balance of each node, that is, the input active power is equal to the output active power plus the power loss on the resistor:

[0159]

[0160] Reactive power balance equation constraint to ensure the reactive power balance at each node, that is, the input reactive power is equal to the output reactive power plus the power loss on the reactance.

[0161]

[0162] Active-reactive power limit equation to limit the maximum values of active and reactive power flows:

[0163]

[0164] Voltage balance constraints include:

[0165]

[0166] Linearization constraints include:

[0167] Linearized apparent power equation constraint:

[0168]

[0169] Linearized active and reactive power equation constraints:

[0170]

[0171] Piecewise linearization equation constraints for active and reactive power:

[0172] 0 ≤ ΔP i，i′f，t，s ≤ ΔS i，i′ ;

[0173] 0 ≤ ΔQ i，i′f，t，s ≤ ΔS i，i′ ;

[0174] Linearized upper limit equation constraint for apparent power:

[0175]

[0176] Photovoltaic constraints include:

[0177] Constraint to ensure that the photovoltaic access capacity is non - negative:

[0178]

[0179] Photovoltaic output factor constraint, which associates the photovoltaic capacity at each node with the photovoltaic output factor for calculating the photovoltaic power generation:

[0180]

[0181] Static var compensator constraints include:

[0182] The capacity limit equation constraint of the static var compensator restricts the capacity and total number of static var compensators at each node:

[0183]

[0184] The static var compensator support range constraint describes the allowable range of reactive power support provided by the static var compensator:

[0185]

[0186] The flexible load constraint describes the limitations of shiftable and interruptible loads:

[0187]

[0188] In the above formula, is the active power purchased by node i from the market at time t under scenario s, is the photovoltaic output power at node i at time t under scenario s, is the active power flow in the downstream direction, that is, in the power grid, the active power flowing from the current node to the next node or from the power source end to the load end, is the active power flow in the upstream direction, that is, in the power grid, the active power flowing from the previous node to the current node or from the load end to the power source end, R i，i′ is the resistance between nodes i and i′, I2 i，i′t，s is the square of the current between nodes i and i′ at time t under scenario s, is the reactive power purchased by node i from the market at time t under scenario s, is the reactive power support of the static var compensator at node i at time t under scenario s, is the reactive power flow in the downstream direction, is the reactive power flow in the upstream direction, X i，i′ is the reactance between nodes i and i′, is the load reactive power at node i at time t, V Rated is the rated voltage, is the maximum current between nodes i and i′; V2 i，t，s is the square of the voltage at node i at time t under scenario s, V2 i′t，s is the square of the voltage at node i′ at time t under scenario s, is the square of the impedance between nodes o and i′; is the square of the rated voltage at node i, f is the block, ΔS i，i′ is the upper limit of the piecewise linearization of the active power and reactive power between nodes o and i′, ΔP i，i′f，t，sThe active power of the f-th block between nodes o and i′ at time t under scenario s, ΔQ i，i′f，t，s The reactive power of the f-th block between nodes i and i′ at time t under scenario s, F is the number of linearized blocks; The installed photovoltaic capacity at node i, The output factor of the photovoltaic at time t under scenario s; The capacity of the static var compensator installed at node i, The maximum capacity of the static var compensator at node i, The binary decision variable indicating whether a static var compensator is installed at node i, The total number of static var compensators allowed to be installed at node i, The absolute value of The auxiliary variable of the static var compensator at node i; The maximum proportion of transferable load at node i, The initial load demand at node i at time t, The transferable load demand at node i at time t under scenario s, The load shedding at node i at time t under scenario s, α curt The maximum proportion of interruptible load at node i, The final load at node i at time t under scenario s.

[0189] 4. Solve the active-reactive power hybrid control model of the power grid system to obtain the active power and reactive power coordinated control strategy for improving the distributed photovoltaic bearing capacity of the distribution network, that is, the configuration scheme of photovoltaic access capacity, flexible load scheduling, and static var compensator. The distributed photovoltaic bearing capacity is the maximum distributed photovoltaic capacity that the power grid can accommodate under the conditions that any equipment is continuously not overloaded and the voltage, short-circuit current, and harmonics of any node do not exceed the standards.

[0190] 5. Use simulation tools to conduct multi-scenario simulation analysis, collect the operation data under each scenario, evaluate the obtained active power and reactive power coordinated control strategy using technical indicators and economic indicators. Based on the evaluation results, further optimize the expected cost model and flexible resource scheduling scheme of the system to ensure that the entire system optimization and decision-making process can be effectively implemented and achieve the best results. The technical indicators include voltage deviation, power loss, photovoltaic access ability, system stability, etc., and the economic indicators include expected cost, operation cost, return on investment, etc. Through multiple evaluations and optimization iterations, the system finally reaches the optimal state in terms of technical performance and economic benefits, improving the robustness and stability of the system under different scenarios;

[0191] When analyzing the evaluation results, according to the simulation results under different scenarios, focus on the following aspects:

[0192] Technical performance issues: If the evaluation results show that there are voltage over-limit, excessive power loss or insufficient PV access capacity in certain scenarios, it is necessary to analyze whether the economic performance is insufficient. For example, in certain scenarios, the expected cost is high or the system operating expenses are too high, so it is necessary to focus on optimizing the configuration or scheduling strategy of related equipment;

[0193] Optimize PV access capacity and location: If the evaluation results show that the PV access capacity is too large, resulting in voltage fluctuations and power imbalance in the distribution network system, or the PV access location of some nodes is inappropriate, resulting in high power loss, the decision on PV access capacity and access location can be adjusted. If the access capacity is too low or the access node selection is not ideal, the system performance can be optimized by increasing the access capacity or reselecting a more suitable node;

[0194] Optimize static VAR compensator configuration: If the system experiences voltage fluctuations or poor voltage stability in high-load scenarios or when PV power generation fluctuates greatly, the number, location or capacity of static VAR compensators can be reconfigured based on the evaluation results to enhance voltage regulation capabilities and ensure that the grid voltage remains within a stable range;

[0195] Adjust the flexibility resource scheduling strategy: Analyze the scheduling effect of flexible loads, such as electric vehicle charging and shiftable loads, based on the evaluation results. If the scheduling of flexible resources in certain scenarios fails to effectively reduce the load pressure or fails to fully utilize the photovoltaic power generation during the period of high photovoltaic power generation, optimize the scheduling decision of flexible resources by adjusting the period of shiftable loads, increasing the interruptibility of loads, or optimizing load shifting plans;

[0196] Adjust the expected cost model: If the evaluation results show that the expected cost is too high in certain scenarios, the cost model needs to be optimized. The cost weight coefficients in different scenarios can be adjusted to better reflect the equipment maintenance costs, flexible resource scheduling costs and photovoltaic power abandonment costs in actual operation. In addition, the electricity purchase costs during peak load periods can be re-evaluated, and the electricity purchase costs can be reduced through appropriate flexible load scheduling to improve the economy of the system;

[0197] Optimize equipment investment and operation strategies: In the evaluation results, if the utilization rate of some equipment is low or the return on investment is insufficient, it is necessary to optimize the investment strategy of the distribution network system, reduce unnecessary equipment configuration, reallocate equipment investment, ensure the optimal utilization of resources, and re-evaluate the economy and importance of each equipment in various operation scenarios by combining the actual operation data in different scenarios, so as to optimize investment and operation decisions;

[0198] Improve the ability to handle scenarios: If the system performs poorly in some extreme scenarios, such as during peak load periods or when photovoltaic power generation fluctuates violently, then scheduling strategies to handle these extreme scenarios can be added to enhance the system's ability to handle uncertainties by improving the adjustment ability of flexible resources, increasing energy storage systems, or optimizing the scheduling logic of equipment;

[0199] Feedback and optimize the overall system design: Re-evaluate the overall decision-making, including the selection of reactive power compensation equipment such as photovoltaic access capacity, flexible load scheduling, equipment investment configuration, and before static reactive power compensation. Combine the evaluation results under multiple scenarios to ensure that the entire system can achieve optimal performance both technically and economically under different load demands and photovoltaic power generation conditions.

[0200] Embodiment 2:

[0201] As Figure 2 shown, a multi-resource coordinated operation system for improving the bearing capacity of distributed photovoltaics includes a load characteristic clustering module, a model construction module, and a model solution module;

[0202] The load characteristic clustering module is used to collect grid load data, perform clustering analysis on the grid load data using an improved fuzzy C-means clustering algorithm, and identify the power load characteristics of the distribution network;

[0203] The model construction module is used to construct an active-reactive power hybrid control model of the grid system based on the identified power load characteristics of the distribution network. The active power control is achieved by optimizing the use of transferable and interruptible flexible loads, and the reactive power control is achieved by optimizing the layout of static reactive power compensators;

[0204] The model solution module is used to solve the active-reactive power hybrid control model of the grid system to obtain the active power and reactive power coordinated control strategies for improving the bearing capacity of the distribution network for distributed photovoltaics. The bearing capacity of distributed photovoltaics is the maximum capacity of distributed photovoltaics that the grid can accept under the conditions that no equipment is continuously overloaded and the voltage, short-circuit current, and harmonics at any node do not exceed the standards.

[0205] The load characteristic clustering module includes a clustering initialization unit, a distance measure optimization unit, a membership degree calculation unit, a clustering centroid update unit, and a loop iteration unit;

[0206] The clustering initialization unit is used to obtain the grid node load data set X = {x1, x2,..., x n}, initialize the settings of the improved fuzzy C-means clustering algorithm, randomly select C = {c1, c2,..., c k} grid load data as the initial clustering centroids, k is the total number of specified clustering centroids, set the maximum number of iterations of the clustering algorithm, and initialize the iteration γ = 1;

[0207] The distance measure optimization unit is used to calculate the local density by using the Pearson correlation coefficient and the Euclidean distance, and optimize the distance measure between the data point and the cluster centroid through the local density;

[0208]

[0209] In the above formula, is the distance measure between the data point X i and the cluster centroid , dx i , is the Euclidean distance between the data point X i and the cluster centroid , is the local density of the data point X i , N a (X i ) is the set of data points closest to the data point X i , d(X i , X j ) is the optimized distance between the data point X i and X j , d(X j , v) is the distance between the data point X i and X j under other measurement criteria, which is used to measure the similarity between data points, is the Euclidean distance between the data point X i and X j , δ(X i , X j ) is the Pearson correlation coefficient between the data point X i and X j , n is the number of sample data, x is is the load value of the i-th data point at the s-th observation, x js is the load value of the j-th data point at the S-th observation;

[0210] The membership degree calculation unit is used to calculate the membership degree between each data point and the cluster centroid by using the distance measure between the data point and the cluster centroid;

[0211]

[0212] In the above formula, is the membership degree between the data point X i and the cluster centroid , is the distance from the data point X i to other cluster centroids , m is the fuzziness parameter;

[0213] The clustering centroid update unit is used to update the position of each clustering centroid by using the calculation result of the membership degree, and obtain a new clustering centroid:

[0214]

[0215] In the above formula, is the updated clustering centroid;

[0216] The loop iteration unit is used to determine whether the membership degree error between two iterations reaches the stop condition or the maximum number of iterations. If the stop condition or the maximum number of iterations is reached, the iteration stops and the clustering result is output. If the stop condition or the maximum number of iterations is not reached, let γ = γ + 1, and return to the distance measure optimization unit to continue the iteration;

[0217] The stop condition is:

[0218]

[0219] In the above formula, ∈ is the membership degree error between two iterations, and τ is the error threshold.

[0220] The model construction module includes an objective function construction unit;

[0221] The objective function construction unit is used to construct the following objective function with the goal of maximizing the access capacity of photovoltaic resources and minimizing the total expected cost of the power grid system:

[0222] Min{-w HC ×HC + w EC ×EC};

[0223]

[0224] In the above formula, w HC is the weight coefficient of the photovoltaic access capacity, HC is the photovoltaic access capacity, w EC is the weight coefficient of the total expected cost of the system, and EC is the total expected cost of the system; is the photovoltaic capacity installed at node i, ρ s is the occurrence probability of scenario s, is the variable operating cost of the static var compensator, is a binary decision variable indicating whether a static var compensator is installed at node i, is the reactive power support of the static var compensator at node i at time t under scenario s, C INC is the incentive payment cost of the interruptible load, is the load power curtailed at node i at time t under scenario s, is the cost for node i to purchase electricity from the market at time t under scenario s. is the active power purchased by node i from the market at time t under scenario s.

[0225] The model construction module further includes an active-reactive power constraint construction unit, a voltage balance constraint construction unit, a linearization constraint construction unit, a photovoltaic constraint construction unit, a static var compensator constraint construction unit, and a flexible load constraint construction unit;

[0226] The active-reactive power constraint construction unit is used to construct the following active-reactive power constraints:

[0227]

[0228] The voltage balance constraint construction unit is used to construct the following voltage balance constraints:

[0229]

[0230] The linearization constraint construction unit is used to construct the following linearization constraints:

[0231]

[0232] The photovoltaic constraint construction unit is used to construct the following photovoltaic constraints:

[0233]

[0234] The static var compensator constraint construction unit is used to construct the following static var compensator constraints:

[0235]

[0236] The flexible load constraint construction unit is used to construct the following flexible load constraints:

[0237]

[0238] In the above formula, is the active power purchased by node i from the market at time t under scenario s, is the photovoltaic output power at node i at time t under scenario s, is the active power flow in the downstream direction, is the active power flow in the upstream direction, R i，i′ is the resistance between nodes i and i′, I2 i，i′t，s is the square of the current between nodes i and i′ at time t under scenario s, is the reactive power purchased by node i from the market at time t under scenario s, For the reactive power support of the static var compensator at node i at time t under scenario s, For the reactive power flow in the downstream direction, For the reactive power flow in the upstream direction, X i，i′ Is the reactance between nodes i and i′, Is the load reactive power at node i at time t, V Rated Is the rated voltage, Is the maximum current between nodes i and i′; V2 i，t，s Is the square of the voltage at node i at time t under scenario s, V2 i′t，s Is the square of the voltage at node i′ at time t under scenario s, Is the square of the impedance between nodes i and i′; Is the square of the rated voltage at node i, f is the block, ΔS i，i′ Is the upper limit of the piecewise linearization of the active power and reactive power between nodes i and i′, ΔP i，i′f，t，s Is the active power of the f-th block between nodes i and i′ at time t under scenario s, ΔQ i，i′f，t，s Is the reactive power of the f-th block between nodes i and i′ at time t under scenario s, F is the number of linearization blocks; Is the installed PV capacity at node i, Is the output factor of the PV at time t under scenario s; Is the capacity of the static var compensator installed at node i, Is the maximum capacity of the static var compensator at node i, Is a binary decision variable indicating whether a static var compensator is installed at node i, Is the total number of static var compensators allowed to be installed at node i, The absolute value of Is an auxiliary variable of the static var compensator at node i; Is the maximum proportion of transferable load at node i, Is the initial load demand of node i at time t, Is the transferable load demand of node i at time t under scenario s, Is the load shedding of node i at time t under scenario s, α curt Is the maximum proportion of interruptible load at node i, Is the final load of node i at time t under scenario s.

[0239] Example 3:

[0240] As Figure 3 Shown, a multi-resource coordinated operation device for improving the bearing capacity of distributed PV includes a processor and a memory;

[0241] The memory is used to store computer program code and transmit the computer program code to the processor;

[0242] The processor is used to execute the multi-resource coordinated operation method for improving the bearing capacity of distributed photovoltaics described in Embodiment 1 according to the instructions in the computer program code.

[0243] Embodiment 4:

[0244] A computer storage medium, on which a computer program is stored;

[0245] When the computer program is executed by a processor, the steps of the multi-resource coordinated operation method for improving the bearing capacity of distributed photovoltaics described in this solution are implemented.

Claims

1. A multi-resource coordinated operation method for improving distributed photovoltaic carrying capacity, characterized in that: The method comprises: S1. Collect grid load data, use improved fuzzy C-means clustering algorithm to perform cluster analysis on grid load data, and identify the power load characteristics of the distribution network; S2. Based on the identified power load characteristics of the distribution network, construct an active-reactive power hybrid control model of the power grid system, wherein the active power control is achieved by optimizing the use of transferable and interruptible flexible loads, and the reactive power control is achieved by optimizing the layout of static VAR compensators; S3. Solve the active-reactive power hybrid control model of the power grid system to obtain the active power and reactive power coordinated control strategy for improving the distribution network's carrying capacity for distributed photovoltaics. The distributed photovoltaic carrying capacity is the maximum capacity of distributed photovoltaics that the power grid can accommodate under the condition that any device is continuously not overloaded and any node voltage, short-circuit current, and harmonics do not exceed the standard.

2. A multi-resource coordinated operation method for improving distributed photovoltaic carrying capacity according to claim 1, characterized in that: In S2, the active-reactive power hybrid control model constructs the following objective function with the goal of maximizing the access capacity of photovoltaic resources and minimizing the total expected cost of the power grid system: Min{-in HC ×HC+in BC ×EC}; In the above formula, w HC is the weight coefficient of photovoltaic access capacity, HC is the photovoltaic access capacity, w EC is the weight coefficient of the total expected cost of the system, EC is the total expected cost of the system; is the photovoltaic capacity installed at node i, ρ s is the occurrence probability of scenario s, is the variable operating cost of the static VAR compensator, is a binary decision variable marking whether to install a static VAR compensator at node i, is the reactive power support of the static VAR compensator at node i at time t under scenario s, C INC Paying the cost of interruptible load incentives, is the load power cut on node i at time t in scenario s, is the cost of node i purchasing electricity from the market at time t under scenario s, is the active power purchased by node i from the market at time t in scenario s.

3. A multi-resource coordinated operation method for improving distributed photovoltaic carrying capacity according to claim 1, characterized in that: In S2, the constraints of the active-reactive power hybrid control model include: Active-reactive power constraint: Voltage balance constraints: Linearization constraints: 0≤ΔP i，i′f，t，s ≤ΔS i，i′ ; 0≤ΔQ i，i′f，t，s ≤ΔS i，i′ ; Photovoltaic constraints: Static VAR compensator constraints: Flexible load constraints: In the above formula, is the active power purchased by node i from the market at time t in scenario s, is the photovoltaic output power at node i at time t under scenario s, is the active power flow in the downstream direction, is the active power flow in the upstream direction, R i，i′ is the resistance between nodes i and i′, I2 i，i′t，s is the square of the current between nodes i and i′ at time t in scenario s, is the reactive power purchased by node i from the market at time t under scenario s, is the reactive power support of the static VAR compensator on node i at time t under scenario s, is the reactive power flow in the downstream direction, is the reactive power flow in the upstream direction, X i，i′ is the reactance between nodes i and i′, is the reactive power of the load on node i at time t, V Rated is the rated voltage, is the maximum current between nodes i and i′; V2 i，t，s V2 is the square of the voltage on node i at time t in scenario s, i′t，s is the square of the voltage on node i′ at time t in scenario s, is the square of the impedance between nodes i and i′; is the square of the rated voltage on node i, f is the block, ΔS i，i′ is the upper limit of the block linearization of active power and reactive power between nodes i and i′, ΔP i，i′f，t，s is the active power of the fth block between nodes i and i′ at time t in scenario s, ΔQ i，i′f，t，s is the reactive power of the fth block between nodes i and i′ at time t in scenario s, where F is the number of linearized blocks; is the photovoltaic capacity installed at node i, is the photovoltaic output factor at time t under scenario s; is the capacity of the static VAR compensator installed on node i, is the maximum capacity of the static VAR compensator at node i, is a binary decision variable marking whether to install a static VAR compensator at node i, is the total number of static VAR compensators allowed to be installed on node i, The absolute value of is the auxiliary variable of the static VAR compensator at node i; is the maximum proportion of load that can be transferred to node i, is the initial load demand of node i at time t, is the load transfer requirement of node i at time t in scenario s, is the load reduction of node i at time t in scenario s, α curt is the maximum proportion of interruptible load on node i, is the final load of node i at time t under scenario s.

4. A multi-resource coordinated operation method for improving distributed photovoltaic carrying capacity according to claim 1, characterized in that: The S1 includes: S11, obtain the grid node load data set X = {x1, x2, ..., x n }, initialize the settings of the improved fuzzy C-means clustering algorithm, randomly select C = {c1, c2, ..., c k } grid load data as the initial clustering centroids, k is the total number of specified clustering centroids, the maximum number of iterations of the clustering algorithm is set, and the initial iteration γ = 1; S12. Use Pearson correlation coefficient and Euclidean distance to calculate local density, and use local density to optimize the distance measure between data points and cluster centroids; In the above formula, For data point x i and cluster centroids The distance measure between For data point x i and cluster centroids The Euclidean distance between For data point x i The local density, N a (X i ) is the value corresponding to the data point X i The set of the most recent data points, d(X i , X j ) is the data point x i and x j The optimal distance between j , v) is the data point x under other metrics i and x j The distance between them is used to measure the similarity between data points. For data point x i and x j The Euclidean distance between i , X j ) is the data point x i and x j The Pearson correlation coefficient between them, n is the number of sample data, x is is the load value of the ith data point at the sth observation, x js is the load value of the jth data point at the Sth observation; S13, using the distance measurement between the data point and the cluster centroid, calculating the membership degree between each data point and the cluster centroid; In the above formula, For data point x i and cluster centroids The degree of membership between For data point x i To other cluster centroids The distance, n is the fuzziness parameter; S14. Use the calculated result of the membership degree to update the position of each cluster centroid to obtain a new cluster centroid: In the above formula, is the updated cluster centroid; S15, judging whether the membership error between two iterations reaches the stopping condition or the maximum number of iterations. If so, the iteration stops and the clustering result is output. If not, set γ=γ+1 and return to S12 to continue the iteration. The stop conditions are: In the above formula, ∈ is the membership error between two iterations, and τ is the error threshold.

5. A multi-resource coordinated operation system for improving distributed photovoltaic carrying capacity, characterized in that: The system includes a load characteristic clustering module, a model building module, and a model solving module; The load characteristic clustering module is used to collect power grid load data, perform cluster analysis on the power grid load data using an improved fuzzy C-means clustering algorithm, and identify the power load characteristics of the distribution network; The model building module is used to build an active-reactive power hybrid control model of the power grid system based on the identified power load characteristics of the distribution network, wherein the active power control is achieved by optimizing the use of transferable and interruptible flexible loads, and the reactive power control is achieved by optimizing the layout of the static VAR compensator; The model solving module is used to solve the active-reactive power hybrid control model of the power grid system to obtain an active power and reactive power coordinated control strategy for improving the distribution network's distributed photovoltaic carrying capacity. The distributed photovoltaic carrying capacity is the maximum capacity of distributed photovoltaics that the power grid can accommodate under the conditions that any device is continuously not overloaded and any node voltage, short-circuit current, and harmonics do not exceed the standard.

6. A multi-resource coordinated operation system for improving distributed photovoltaic carrying capacity according to claim 5, characterized in that: The model building module includes an objective function building unit; The objective function construction unit is used to construct the following objective function with the goal of maximizing the access capacity of photovoltaic resources and minimizing the total expected cost of the power grid system: Min{-in HC ×HC+in BC ×EC}; In the above formula, w HC is the weight coefficient of photovoltaic access capacity, HC is the photovoltaic access capacity, w EC is the weight coefficient of the total expected cost of the system, EC is the total expected cost of the system; is the photovoltaic capacity installed at node i, ρ s is the occurrence probability of scenario s, is the variable operating cost of the static VAR compensator, is a binary decision variable marking whether to install a static VAR compensator at node i, is the reactive power support of the static VAR compensator at node i at time t under scenario s, C INC Paying the cost of interruptible load incentives, is the load power cut on node i at time t in scenario s, is the cost of node i purchasing electricity from the market at time t under scenario s, is the active power purchased by node i from the market at time t in scenario s.

7. A multi-resource coordinated operation system for improving distributed photovoltaic carrying capacity according to claim 5, characterized in that: The model construction module also includes an active-reactive power constraint construction unit, a voltage balance constraint construction unit, a linearization constraint construction unit, a photovoltaic constraint construction unit, a static VAR compensator constraint construction unit, and a flexible load constraint construction unit; The active-reactive power constraint construction unit is used to construct the following active-reactive power constraint: The voltage balance constraint construction unit is used to construct the following voltage balance constraint: The linearization constraint construction unit is used to construct the following linearization constraints: 0≤ΔP i，i′f，t，s ≤ΔS i，i′ ; 0≤ΔQ i，i′f，t，s ≤ΔS i，i′ ; The photovoltaic constraint construction unit is used to construct the following photovoltaic constraints: The static VAR compensator constraint construction unit is used to construct the following static VAR compensator constraints: The flexible load constraint construction unit is used to construct the following flexible load constraint: In the above formula, is the active power purchased by node i from the market at time t in scenario s, is the photovoltaic output power at node i at time t under scenario s, is the active power flow in the downstream direction, is the active power flow in the upstream direction, R i，i′ is the resistance between nodes i and i′, I2 i，i′t，s is the square of the current between nodes i and i′ at time t in scenario s, is the reactive power purchased by node i from the market at time t under scenario s, is the reactive power support of the static VAR compensator on node i at time t under scenario s, is the reactive power flow in the downstream direction, is the reactive power flow in the upstream direction, X i，i′ is the reactance between nodes i and i′, is the reactive power of the load on node i at time t, V Rated is the rated voltage, is the maximum current between nodes i and i′; V2 i，t，s V2 is the square of the voltage on node i at time t in scenario s, i′t，s is the square of the voltage on node i′ at time t in scenario s, is the square of the impedance between nodes i and i′; is the square of the rated voltage on node i, f is the block, ΔS i，i′ is the upper limit of the block linearization of active power and reactive power between nodes i and i′, ΔP i，i′f，t，s is the active power of the fth block between nodes i and i′ at time t in scenario s, ΔQ i，i′f，t，s is the reactive power of the fth block between nodes i and i′ at time t in scenario s, where F is the number of linearized blocks; is the photovoltaic capacity installed at node i, is the photovoltaic output factor at time t under scenario s; is the capacity of the static VAR compensator installed on node i, is the maximum capacity of the static VAR compensator at node i, is a binary decision variable marking whether to install a static VAR compensator at node i, is the total number of static VAR compensators allowed to be installed on node i, The absolute value of is the auxiliary variable of the static VAR compensator at node i; is the maximum proportion of load that can be transferred to node i, is the initial load demand of node i at time t, is the load transfer requirement of node i at time t under scenario s, is the load reduction of node i at time t in scenario s, α curt is the maximum proportion of interruptible load on node i, is the final load of node i at time t under scenario s.

8. A multi-resource coordinated operation system for improving distributed photovoltaic carrying capacity according to claim 5, characterized in that: The load characteristic clustering module includes a cluster initialization unit, a distance measurement optimization unit, a membership calculation unit, a cluster centroid update unit, and a loop iteration unit; The cluster initialization unit is used to obtain a grid node load data set X={x1, x2, ..., x n }, initialize the settings of the improved fuzzy C-means clustering algorithm, randomly select C = {c1, c2, ..., c k } grid load data as the initial clustering centroids, k is the total number of specified clustering centroids, the maximum number of iterations of the clustering algorithm is set, and the initial iteration γ = 1; The distance measure optimization unit is used to calculate the local density using the Pearson correlation coefficient and the Euclidean distance, and optimize the distance measure between the data point and the cluster centroid through the local density; In the above formula, For data point x i and cluster centroids The distance measure between For data point x i and cluster centroids The Euclidean distance between For data point X i The local density, N a (X i ) is the value corresponding to the data point X i The set of the most recent data points, d(X i , X j ) is the data point x i and x j The optimal distance between j , v) is the data point x under other metrics i and x j The distance between them is used to measure the similarity between data points. For data point x i and x j The Euclidean distance between i , X j ) is the data point x i and x j The Pearson correlation coefficient between them, n is the number of sample data, x is is the load value of the ith data point at the sth observation, x js is the load value of the jth data point at the Sth observation; The membership calculation unit is used to calculate the membership between each data point and the cluster centroid using the distance measurement between the data point and the cluster centroid; In the above formula, For data point x i and cluster centroids The degree of membership between For data point x i To other cluster centroids The distance, m is the fuzziness parameter; The cluster centroid updating unit is used to update the position of each cluster centroid using the calculation result of the membership degree to obtain a new cluster centroid: In the above formula, is the updated cluster centroid; The loop iteration unit is used to determine whether the membership error between two iterations reaches the stopping condition or the maximum number of iterations. If the stopping condition or the maximum number of iterations is reached, the iteration stops and the clustering result is output. If the stopping condition or the maximum number of iterations is not reached, γ=γ+1 is set, and the distance measurement optimization unit is returned to continue the iteration. The stop conditions are: In the above formula, ∈ is the membership error between two iterations, and τ is the error threshold.

9. A multi-resource coordinated operation device for improving the carrying capacity of distributed photovoltaics, characterized in that: including a processor and a memory; The memory is used to store computer program code and transmit the computer program code to the processor; The processor is used to execute a multi-resource coordinated operation method for improving distributed photovoltaic carrying capacity as described in any one of claims 1-4 according to the instructions in the computer program code.

10. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a multi-resource coordinated operation method for improving distributed photovoltaic carrying capacity described in any one of claims 1-4 are implemented.

Images