Power distribution system reliability index analysis method considering source load multi-period time sequence scene

Abstract

The invention relates to a power distribution system reliability index analysis method considering a source-load multi-period time sequence scene, and the method comprises the steps: completely analyzing and expressing a space incidence relation between a fault event consequence and a fault element through a fault influence incidence matrix; complex fault event enumeration and repeated fault influence range search in a traditional reliability evaluation process are avoided. In addition, the method utilizes a K-Medoids clustering algorithm to construct a typical operation scene set, depicts the internal relation and continuous dynamic characteristics of the distributed power supply and the load, retains the source-load time sequence association relation of the power distribution system, greatly reduces the calculation overhead, and improves the reliability evaluation efficiency of the power distribution system after the novel source load is accessed. A power failure load multi-time-sequence scene optimal recovery strategy considering capacity constraint and connectivity constraint is provided by establishing a reliability evaluation-oriented time sequence element model, and power distribution covering a source load multi-time-period time sequence scene within a statistical age limit is explicitly and quantitatively evaluated based on four types of fault influence incidence matrixes and a post-fault island recovery matrix.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power systems, and particularly relates to a distribution system reliability index analysis method considering source-load multi-period time sequence scenarios. BACKGROUND

[0002] Currently, a large number of distribution system reliability index calculation methods take the average load demand in the statistical year as the load demand in the entire statistical year, that is, the difference in load levels between different scenarios is ignored, and the system operating state is simplified as a processing mode of a constant value in the statistical year, which has a significant technical bottleneck, is difficult to adapt to the dynamic coupling characteristics of new time sequence elements such as distributed power sources, energy storage devices and flexible loads, and makes it difficult to accurately represent the difference between time sequence scenarios. In view of the problem of distribution system reliability evaluation containing time sequence source and load characteristics, the existing technology mainly forms two different solutions: a method for evaluating reliability based on an element probability model using an analytical method, and a method for evaluating reliability based on a time sequence model using a simulation method.

[0003] The analytical method based on the probability model driven evaluation system realizes quantitative analysis of reliability by constructing a multi-state probability model of the time sequence element. Typical technical paths include: establishing a multi-state transition probability model of distributed power output, combining Markov chain to solve the island formation probability; constructing a multi-state model of power outage sequence and designing a forward propagation algorithm to calculate the state transition probability; using a random function to represent the reliability index, combining the probability distribution reconstruction technology to realize the index probability density analysis; designing a mapping mechanism of point estimation scenario and reliability index, generating an evaluation scenario set through the probability distribution function, etc. This kind of method abstracts the time sequence characteristics through the probability model, but there is a solution bottleneck of complex system state space explosion.

[0004] The simulation method based on the time sequence model driven evaluation system realizes quasi-dynamic simulation by constructing an element time sequence behavior model. The main technical features include: establishing a high-precision time sequence model and state transition mechanism of wind-solar-storage elements, combining quasi-sequential Monte Carlo method to simulate the fault scenario evolution process; constructing a dynamic correlation model of photovoltaic-energy storage system capacity configuration and reliability index, quantifying the capacity influence coefficient through time sequence Monte Carlo simulation; designing a multi-period dynamic island division optimization model, integrating network reconfiguration strategy and AC / DC control logic, and constructing a sequential simulation evaluation framework, etc. Although this kind of method can depict the time sequence dynamic characteristics, it has the inherent contradiction that the calculation efficiency and precision are difficult to balance.

[0005] The two methods have differences in calculation complexity and time sequence representation ability, but have not effectively solved the problem of cooperative optimization of time sequence accuracy and calculation efficiency in distribution system reliability evaluation, and cannot realize the analytical calculation of distribution system reliability index considering source-load multi-period time sequence scenarios.

[0006] For the reliability calculation problem of distribution system considering source-load multi-period time sequence scene, how to balance the modeling accuracy of time sequence components and the fast and accurate reliability evaluation of analytical method, under the premise of retaining the time sequence correlation of each component, typical scene extraction and reduction are carried out, the reliability analytical calculation of distribution system considering the randomness of distributed power output and the time sequence fluctuation of load is realized, which becomes a problem to be solved. SUMMARY

[0007] The present application aims to overcome the deficiencies of the prior art, and proposes a distribution system reliability index analytical method considering source-load multi-period time sequence scene, which can realize one-time analytical calculation of complex distribution system reliability index, and explicitly quantify the reliability level of distribution system covering source-load multi-period time sequence scene within a statistical period.

[0008] The present application solves its technical problems by adopting the following technical solutions: The distribution system reliability index analytical method considering source-load multi-period time sequence scene comprises the following steps: Step 1, based on the spatial correlation between the power supply path of the load node and the key components, the influence of branch fault on the load node in the distribution system is classified, and four types of fault influence correlation matrixes are established, which provide matrix analysis basis for subsequent reliability index analytical calculation.

[0009] Step 2, establish time sequence operation reliability evaluation model of DG, energy storage device, flexible load and other components, and clearly define the island restoration potential of DG and energy storage under different operating states and the response characteristics of flexible load, and construct island restoration matrix calculation method considering capacity constraint and connectivity constraint under various fault scenes.

[0010] Step 3, considering the time sequence difference characteristics of distribution system operation scene caused by DG, energy storage and flexible load, based on K-Medoids clustering algorithm, construct a typical scene set of distribution system, reduce the number of scenes to be evaluated while retaining the time sequence correlation of source and load in the distribution system, and reduce the calculation cost.

[0011] Step 4, based on the typical operation scene set obtained by clustering, combined with the established fault influence correlation matrix and island restoration matrix model, a distribution system reliability level analysis method covering source-load multi-period time sequence scene is proposed, and the distribution system reliability index analytical calculation covering source-load multi-period time sequence scene is carried out.

[0012] Moreover, the specific implementation method of step 1 is to classify the influence of branch fault on load node into four types: Type a: the branch fault causes all power supply paths of the load node to be interrupted, and the load node needs to wait for the completion of fault repair to restore power supply, so the outage time caused by the branch fault is the fault repair time of the branch; Type b: Branch failure causes all power supply paths of the load node to be interrupted, and the load node needs to wait for the fault to be isolated and then be supplied by the main power supply, thus the power outage time caused by the branch failure is the isolation time of the fault; Type c: Branch failure causes all power supply paths of the load node to be interrupted, and the load node needs to wait for the fault to be isolated and then be supplied by the standby power supply, thus the power outage time caused by the branch failure is the isolation and power supply transfer time of the fault; Type d: Branch failure has no effect on the load node; Constructing the fault influence correlation matrix FEIM , FEIM including FEIMA , FEIM B , FEIM C and FEIM D , a the fault influence correlation matrix of the type a is: FEIMA

[0013] b the fault influence correlation matrix of the type b is: FEIMB

[0014] c the fault influence correlation matrix of the type c is: FEIMC

[0015] d the fault influence correlation matrix of the type d is: FEIMD

[0016] In the formula, the matrix represents a dimensional matrix with all elements being “1”, the matrix represents a dimensional matrix with all elements being “0”, and the logical operator “ ” represents taking “exclusive or” of elements in the matrix bit by bit.

[0017] Moreover, the single dispatchable DG island restoration matrix correction model of a certain time t in the step 2 is:

[0018] In the formula, the matrix represents the element in the i th row and the j th column of the i th corrected island restoration matrix, k i j ​​​​​​​, They represent the first k Within an isolated island, schedulable DG is currently... t Every moment, whether contributing meritoriously or not, requires effort. , They represent the first k An unschedulable DG within an isolated island is currently... t Every moment, whether contributing meritoriously or not, requires effort. , They represent the first k An isolated energy storage element in the current t Efforts made at all times, whether they yield merit or not; In obtaining the first k Island recovery matrix IRM k Then, obtain a certain moment. t The whole system island recovery matrix IRM :

[0019] in, N sch This indicates the amount of dispatchable DG and energy storage.

[0020] Furthermore, the specific implementation method of step 3 is as follows: Euclidean distance is selected to characterize the distance between different classes. A contour coefficient is used. s Measuring the number of clusters K Validity:

[0021]

[0022] In the formula, m This represents the total number of sample points. Indicates sample To the i The mean distance of each sample point in the cluster. Indicates sample Mean distance to each sample point in the nearest cluster, number of clusters K The range of values ​​for is .

[0023] Furthermore, step 4 includes the following steps: Step 4.1: Count the number of all scenarios to be clustered and reduced. m Set typical scenario numbers K =2; Step 4.2, Random selection K One scenario was used as the initial clustering result; Step 4.3: Calculate the required reduction based on the distributed power output, energy storage output, and load demand in different operating scenarios. m From a single running scenario to a randomly selected one. K The Euclidean distance between the cluster centers (medoids); Step 4.4: Based on the principle of finding the closest cluster center point for each scene sample, sort the remaining samples... m - K Each scene is assigned to the class represented by the current best center point; Step 4.5: In each class, calculate the sum of distances from each scene to other scenes in that class according to equation (33), and take the point with the smallest sum of distances from other scenes in each class as the center point of that class, that is: (31) In the formula, Indicates the first i Sample points in a cluster Indicates the first j The iteration process of the iteration process i The center point of the cluster.

[0024] Step 4.6: Repeat steps 4.3 to 4.5 until the new center point converges to the previous center point; Step 4.7: Calculate and record the profile coefficients. s And determine the number of clusters. K Has the value reached the upper limit? If the upper limit has not yet been reached, then take... K = K +1, jump to step 4.2; Step 4.8: Compare all K Profile coefficient under the given value s And the profile coefficient s When taking the maximum value K The values ​​and their corresponding clustering results are used as the final cluster number and clustering result.

[0025] The advantages and positive effects of this invention are: This method fully analyzes the spatial correlation between the consequences of fault events and faulty components through a fault impact correlation matrix, avoiding the complex fault event enumeration and repetitive fault impact range search in traditional reliability assessments. Furthermore, this method utilizes the K-Medoids clustering algorithm to construct a set of typical operating scenarios, characterizing the inherent connections and continuous dynamic features of distributed power sources and loads, preserving the temporal correlation between power distribution system sources and loads, significantly reducing computational overhead, and improving the efficiency of power distribution system reliability assessment after the integration of new power sources and loads. By establishing a temporal component model oriented towards reliability assessment, it proposes an optimal recovery strategy for multi-temporal scenarios of power outage loads that considers capacity constraints and connectivity constraints. Based on four types of fault impact correlation matrices and a post-fault island recovery matrix, it explicitly quantifies and assesses the reliability level of power distribution systems covering multi-time-period temporal scenarios of sources and loads within a statistical period, significantly improving the efficiency of power distribution system reliability assessment while ensuring the accuracy of analytical calculations. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the analytical calculation of reliability indicators for power distribution systems considering multi-period time-series source-load scenarios in this invention. Figure 2 This is a schematic diagram illustrating the process of determining the type of impact of a branch fault on a load node according to the present invention. Figure 3 This is a structural diagram of a 137-node system that integrates various novel components according to an embodiment of the present invention; Figure 4 This is a distribution map of the 8760-hour scene data of this invention; Figure 5 This is a graph showing the variation of the contour coefficient s of the present invention with the number of scenes K; Figure 6 This is a schematic diagram of the clustering reduction results for the 8760-hour scenario of this invention. Figure 7 This is a probability distribution diagram for a typical scenario of the present invention; Figure 8 This is a map showing the island recovery range in eight typical scenarios of this invention; Figure 9 This is a schematic diagram illustrating the number of power outages at each load node under each typical scenario of the present invention; Figure 10 This is a schematic diagram illustrating the power outage time of each load node under each typical scenario of the present invention; Figure 11 This is a schematic diagram of the power shortage at each load node under each typical scenario of the present invention; Figure 12 This is a schematic diagram showing the number of power outages, power outage time, and power shortage at each load node in this invention. Detailed Implementation

[0027] The present invention will be further described in detail below with reference to the accompanying drawings.

[0028] A method for analyzing the reliability indicators of power distribution systems considering multi-time-sequence scenarios of source and load, such as... Figure 1 As shown, it includes the following steps: Step 1: Construct the correlation matrix of the impact of four types of faults.

[0029] Because distribution network topologies are characterized by numerous segmented interconnection switches and complex reconfigurable paths, they often have multiple transfer paths in fault scenarios. Therefore, matrix modeling tools can be considered for reliability assessment. Furthermore, matrix computation is a parallel computation method, which can effectively reduce redundant searches and is more suitable for the characteristics of distribution networks with numerous points, wide distribution areas, and complex reconfigurable paths.

[0030] This invention categorizes the impact of branch faults on load nodes into four types, with the characteristics of each type as follows: Type a: A branch fault causes an interruption of all power supply paths to the load node. The load node needs to wait for the fault to be repaired before power can be restored. The power outage time caused by this branch fault is the fault repair time. Type b: A branch fault causes an interruption of all power supply paths to the load node. The load node needs to wait for the fault isolation before power can be restored by the main power supply. The power outage time caused by this branch fault is the fault isolation time. Type c: A branch fault causes an interruption of all power supply paths to the load node. The load node needs to wait for the fault isolation before power can be restored by the backup power supply. The power outage time caused by this branch fault is the fault isolation and power transfer time. In fact, a branch fault may have no impact on the load node, such as faulty branches and nodes located on different feeders. This type of impact can be categorized as Type d.

[0031] Based on the definition of the four types of component failure impact types, this method further defines the four types of failure effect incidence matrices. FEIM ): FEIMA , FEIM B , FEIM C and FEIM D The four types of faults affect the elements in the correlation matrix. A i,j , B i,j , C i,j and D i,j The definition is as follows: (1) (2) (3) (4) Meanwhile, to ensure accurate identification of fault impact types, a fault verification matrix is ​​defined. CHECK as follows: (5) By introducing the basic principles of power distribution system reliability analysis, it can be found that the impact of a component failure on a load node mainly depends on whether the fault location is on the power supply path, the configuration of switching devices (fuses, circuit breakers, sectionalizing switches, etc.), and whether there is a valid tie line as a backup power source. A valid tie line refers to a tie line that exists downstream of the branch and a sectionalizing switch (or fuse) between the node where the tie line is located and the faulty branch.

[0032] The following section will analyze in detail the types of impacts of a fault on load nodes and the load transfer paths based on the location of the fault. Note that if the faulty branch and the load node are located on different feeders, the load node on the faulty branch will definitely be unaffected; that is, the type of impact will always be [missing information]. d Therefore, it is only necessary to analyze the case where the faulty branch and the load node are located on the same feeder.

[0033] The process for determining the impact type of branch faults on load nodes in all scenarios is as follows: Figure 2 As shown. By Figure 2 It can be seen that when searching for branch faults in a loop, each loop process requires determining the upstream and downstream relationship of the power supply path between the faulty branch and the load node, as well as the actual configuration of the sectionalizing switch, fuse, and tie feeder. These conditions can be obtained by analyzing the spatial topology of the power distribution system.

[0034] Meanwhile, to facilitate subsequent matrix operations, four matrices are used to analytically represent the four judgment dimensions, namely the power supply path association matrix. PSPM Segmented switch matrix SE Fuse matrix FUSE and effective contact matrix TIE The elements in each matrix are defined as follows: (6) (7) (8) (9) When studying the branches of the same feeder i Failure occurred to the node j When considering the impact, assume PSPM i,j = e, SE i,j =f, FUSE’ i,j= g, TIE i,j =h The type of fault impact can be determined using characteristic features. Figure 2 Table 1 summarizes the correspondence between fault impact types and characteristic formulas for determining the impact type of branch faults on load nodes.

[0035] Table 1. Correspondence between Fault Impact Types and Characteristic Equations

[0036] Based on the constructed power supply path association matrix PSPM Segmented switch matrix SE Fuse matrix FUSE and effective contact matrix TIE Then, the correlation matrix of the four types of fault impacts can be obtained. FEIMA , FEIM B , FEIM C and FEIM D .

[0037] a Fault-related correlation matrix FEIMA As shown in equation (10): (10) b Fault-related correlation matrix FEIMB As shown in equation (11): (11) c Fault-related correlation matrix FEIMC As shown in equation (12): (12) d Fault-related correlation matrix FEIMD As shown in equation (13): (13) In the formula, the matrix This indicates that all elements are "1". 3D matrix, matrix This indicates that all elements are "0". 3D matrix, logical operator " " indicates that the elements in the matrix are bitwise XORed.

[0038] (1) Second, establish a timing element model for reliability assessment.

[0039] 1) Distributed photovoltaic The photovoltaic active power output exhibits a linear relationship with the solar irradiance. Considering that the solar irradiance follows a Beta distribution over a certain period of time, the photovoltaic active power output also approximately follows a Beta distribution, as shown in equation (14): (14) In the formula, Indicating distributed photovoltaic elements in t Active power output at any given time Indicates the number of distributed photovoltaic elements. This represents the illuminated area of ​​a distributed photovoltaic (PV) element. This represents the rated active power of a photovoltaic element per unit area under standard light intensity and standard ambient temperature. Indicating distributed photovoltaic elements in t The intensity of light received at all times, This represents the standard illuminance of a distributed photovoltaic (PV) element. Indicates the power temperature rise coefficient. and These represent distributed photovoltaic elements in t The ambient temperature at any given time and the standard ambient temperature.

[0040] 2) Distributed wind turbines The active power output of wind turbine generators is mainly affected by wind speed. According to statistical laws, wind speed... The randomness follows a Weibull distribution. Considering the inertia of the mechanical rotating structure of the generator set, a small wind speed cannot drive the blades to rotate; considering the requirements of operational safety and stability, the blades should be braked when the wind speed is large. Within a certain wind speed range, its active power output increases with the increase of wind speed, but cannot exceed its rated active power. The relationship between the active power output of the wind turbine generator set and the wind speed is shown in equation (15): (15) In the formula, Indicating distributed wind turbines t Active power output at any given time This indicates the number of distributed wind turbine components. This indicates the rated active power of a single fan at the rated wind speed. , , and These represent the current wind speed, cut-in wind speed, cut-out wind speed, and rated wind speed of the fan, respectively.

[0041] 3) Energy storage devices From the perspective of reliability improvement, the reasonable configuration of energy storage devices can provide power to load nodes in the form of islanded operation under fault scenarios, thereby reducing the power outage time of the load. Its operating status is mainly described by the State of Charge (SOC), as shown in Equation (16): (16) In the formula, , They represent the first i An energy storage device in t+1 time, t State of charge at time t, Indicates the first i The charging power of an energy storage device Indicates the first i The discharge power of the energy storage device. Considering that the lifespan of the battery energy storage device is related to its state of charge, overcharging or over-discharging will damage the lifespan of the battery energy storage device. Therefore, the state of charge and charge / discharge power of the battery must be within a reasonable range as shown in equations (17) to (19): (17) (18) (19) In the formula, , These represent the upper and lower limits of the state of charge of the energy storage device, respectively. , These represent the upper and lower limits of the charging power of the energy storage device, respectively. , These represent the upper and lower limits of the discharge power of the energy storage device, respectively.

[0042] 4) Flexible load Based on their degree of demand response, loads in new power distribution systems can be divided into two categories: rigid loads and flexible loads. Rigid loads cannot respond to the demand commands of the power distribution network, while flexible loads can respond to the network's needs by performing actions such as load shifting (LS), load transferring (LT), and load curtailment (LC). In fault scenarios where islanded power is insufficient, flexible loads can be shifted, transferred, or curtailed to prioritize the continuous power supply to critical loads.

[0043] The reliability models for the transferable load are shown in equations (20) and (21): (20) (twenty one) In the formula, , The curves after translation are respectively t Time and The load demand at any time, , These are the original curves. t Time and The load demand at any time, Indicates in t Participate in the translation at all times The load ratio at any given time.

[0044] The reliability models for transferable loads are shown in equations (22) and (23): (twenty two) (twenty three) In the formula, , The curves after the transfer are respectively t Time and The load demand at any time, , These are the original curves. t Time and The load demand at any time, Indicates in t Time shift The load ratio at any given time.

[0045] The reliability model for load reduction is shown in equation (24): (twenty four) In the formula, Indicates the curve after reduction t The load demand at any time, Represents the original curve t The load demand at any time, Indicates in t The percentage of load that is constantly being reduced.

[0046] In new power distribution systems, due to the large-scale integration of distributed generation (DG), energy storage, and flexible loads, the system will possess the capability for islanded operation. Loads that were previously unable to be transferred due to faults can be restored to power through islanding. Since islanding operation is typically short-lived, this method assumes that the active power output of the energy storage device remains constant during islanding operation.

[0047] Compare the impact of branch faults on the load before and after DG (energy storage) integration with the four types of faults. b and type cThe fault transfer process always involves fault isolation (and power transfer) followed by restoration of power from the original main power source (or tie line). There is no change before and after the DG (energy storage) is connected to the transfer path, and the type... d This indicates that a branch fault has no impact on the load node, therefore only the original type is present. a The fault requires power to be restored from the island.

[0048] To represent the relationship between load nodes and the islanding recovery range, an islanding recovery matrix is ​​defined. (Island Recovery Matrix) is as follows: (25) In actual distribution networks, the output of distributed generation (DG) and energy storage has an upper limit. During fault transfer, it is often impossible to meet the demand of all loads within the island, requiring load reduction, i.e., islanding recovery matrix. Some elements need to be changed from "1" to "0". Since the load node requirements, distributed power output, and energy storage component output vary in different scenarios, capacity constraints and connectivity constraints need to be considered comprehensively to define the island recovery range in the current scenario.

[0049] Let the corrected island restoration matrix be . Then consider a certain moment under capacity constraints. t The modified model for the islanded recovery matrix of a single schedulable DG is as follows: (26) In the formula, This indicates the corrected island recovery matrix. i Line 1 j Column elements, Indicates load node j The importance of Indicates the current t Time-based load nodes j The active demand, , and They represent the current t Time-based load nodes j The proportion of loads that can be moved, transferred, and reduced. , They represent the current t The active and passive contributions of DG at all times , They represent the current t It can store energy at any time, both active and reactive power. Indicates DG (energy storage) to load nodes j The set of nodes on the power supply path.

[0050] The model's objective function represents restoring as many higher-importance loads as possible. The first and second constraints respectively indicate that, after considering the active abandonment of flexible loads, the active and reactive power demands of all loads within the islanded area cannot exceed the DG and energy storage time limits. t The third constraint states that the modified islanded power restoration range cannot exceed the original islanded power restoration range. The fourth constraint represents the weighting of importance. The fifth constraint states that the islanded power restoration path satisfies the radial connectivity constraint, i.e., if the branch... i Fault node j Power can be restored by DG (distributed generation) and energy storage, then from DG (energy storage) to the node j All nodes along the power supply path can also be restored to power by DG and energy storage.

[0051] The application model, including capacity constraints, at a certain moment. t The islanding recovery matrix of a single schedulable DG can be easily solved using a 0-1 integer optimization model.

[0052] Based on the single schedulable DG islanding recovery matrix optimization model, further research is conducted on a certain time step under capacity constraints. t The island recovery matrix. Assume that dispatchable distributed generation (DG) and energy storage are shared in the distribution system. N sch There are 1,0 ... unschedulable DGs. N unsch If there are 1, then the island recovery matrix needs to be calculated for each island formed by a schedulable DG (energy storage). Let's denote the 1st island as... k The island recovery matrix formed by a schedulable DG (energy storage) is as follows: IRM k Then the optimization model in equation (26) can be extended: (27) In the formula, Indicates the revised version of the first... k In the island recovery matrix, the first... i Line 1 j Column elements, , They represent the first k Within an isolated island, schedulable DG is currently... t Every moment, whether contributing meritoriously or not, requires effort. , They represent the first k An unschedulable DG within an isolated island is currently... t Every moment, whether contributing meritoriously or not, requires effort. , They represent the first k An isolated energy storage element in the currentt The active and reactive power outputs at any given time, and the meanings of the other variables are consistent with those in the optimization model (26).

[0053] The model's objective function represents restoring as many loads of higher importance as possible. The first and second constraints represent the conditions after considering the active discarding of flexible loads. k The active and reactive power demands of all loads within the isolated island must not exceed the DG and energy storage time limits. t The active and reactive power outputs. The third constraint condition represents the modified first... k The power restoration range for each isolated island cannot exceed the range before the revision. k The fourth constraint represents the normalized weighting of importance. The fifth constraint represents the radial connectivity constraint that the island restoration path satisfies. The sixth constraint represents the non-overlapping island ranges, and each load node can only be restored to power by a maximum of one island per failure.

[0054] In obtaining the first k Island recovery matrix IRM k Then, a certain moment can be obtained. t The whole system island recovery matrix IRM : (28) In the formula, N sch This indicates the amount of dispatchable DG and energy storage.

[0055] (3) Third, construct typical operation scenarios of power distribution network based on K-Medoids clustering algorithm.

[0056] Euclidean distance (2-norm) is chosen to characterize the distance between different classes. Silhouette coefficients are used. s (SilhouetteCoefficient) measures the number of clusters. K Validity: (29) (30) In the formula, m This represents the total number of sample points. Indicates sample To the i The mean distance of each sample point in the cluster. Indicates sample Mean distance to each sample point in the nearest cluster, number of clusters K The range of values ​​for is .

[0057] Based on the K-Medoids clustering algorithm described above, a method for constructing a set of typical operating scenarios for power distribution systems oriented towards reliability assessment is established. The specific steps are as follows: (1) Count the number of all scenarios to be clustered and reduced. m Set typical scenario numbers K =2; (2) Random selection K One scenario was used as the initial clustering result; (3) Calculate the required reduction based on the output of distributed power sources, energy storage, and load demand in different operating scenarios. m From a single running scenario to a randomly selected one. K The Euclidean distance between the cluster centers (medoids); (4) Based on the principle that each scene sample is closest to the current cluster center, the remaining samples are sorted out. m - K Each scene is assigned to the class represented by the current best center point; (5) In each class, calculate the sum of distances from each scene to other scenes in that class according to equation (33), and take the point with the smallest sum of distances from other scenes in each class as the center point of that class, that is: (31) In the formula, Indicates the first i Sample points in a cluster Indicates the first j The iteration process of the iteration process i The center point of the cluster.

[0058] (6) Repeat steps (3), (4), and (5) until the new center point converges to the previous center point; (7) Calculate and record the profile coefficients according to equations (29) and (30). s And determine the number of clusters. K Has the value reached the upper limit? If the upper limit has not yet been reached, then take... K = K +1, jump to step (2); (8) Compare all K Profile coefficient under the given value s And the profile coefficient s When taking the maximum value K The values ​​and their corresponding clustering results are used as the final cluster number and clustering result.

[0059] Therefore, it can be m The operational scenarios were reduced to representative ones. KA typical operating scenario. The method for constructing a typical operating scenario set based on the K-Medoids clustering algorithm characterizes the inherent relationship and continuous dynamic characteristics of distributed power sources and loads, and preserves the time-series correlation between power distribution system sources and loads, making it reasonable to carry out reliability calculations based on the typical operating scenario set.

[0060] (4) Fourth, establish an analytical calculation model for reliability indicators of complex power distribution systems based on the fault impact correlation matrix.

[0061] set up t isld Indicates the time when the island was formed, all K In a typical scenario N A matrix composed of the load demand of each node As shown in equation (32): (32) In the formula, the matrix The Middle i Line 1 j Column elements represent the first i In the scenario, the first j The load demand of the nth load node can then be calculated. k In each scenario N The reliability indices of each load node are shown in equations (33) to (35): (33) (34) (35) In the formula, row vector , and They represent N The node at the th k Average number of power outages, average outage duration, and power shortage amount in each scenario. Representing the load demand matrix The k OK, Indicates the first k System-wide island recovery matrix for various scenarios.

[0062] N The expected reliability index values ​​for each load node are shown in equations (36) to (38): (36) (37) (38) In the formula, row vector express NThe average number of power outages per year for each node, row vector express N The average annual power outage time for each node, row vector express N The power shortage of each node, row vector express K The probability of each scenario occurring , They represent the first i In the scenario, the first j The power outage time and power shortage of each node.

[0063] Similarly, the first can be calculated to obtain the second. k The system-level reliability metrics for each scenario are shown in equations (39) to (41): (39) (40) (41) In the formula, , and They represent the first k Average number of system outages, average system outage duration, and expected power shortage in each scenario. n Indicates to be in numerical order N A row vector consisting of the number of users on each load node. N cons This represents the total number of users in the power distribution system. Indicates the first k Transpose the row vector formed by all load demands in each scenario. Indicates taking a vector The 1-norm.

[0064] The expected values ​​of the system-level reliability indicators are shown in equations (42) to (44): (42) (43) (44) In the formula, P k Indicates the first k The probability of each scenario occurring can reveal indicators related to power outage duration. and The reduction is somewhat mitigated after considering isolated operation, because some of the original impacts... a Loads affected by this type of fault that can only wait for power restoration after the fault is repaired can be restored by islanding, reducing the power outage time to the islanding time.t isld .

[0065] (5) Fifth, analytical calculation of reliability indicators of power distribution system covering multi-time-sequence scenarios of source and load. This method analyzes 137 nodes. To prioritize the restoration of power to critical loads and fully leverage the islanding support role of photovoltaic, wind turbine, and energy storage components, the connection location is selected at the node containing Class A loads, and a new smart soft switch SOP is added as a tie device. Figure 3 As shown. Set the island formation time. t isld The lifespan is 0.5 hours, and there are circuit breakers at the outlets of each photovoltaic element, wind turbine element, and energy storage element.

[0066] The branch failure rate and fault repair time, load demand of each node and number of users are shown in Tables 2 and 3.

[0067] Table 2 Component Failure Rate and Failure Repair Time of 137-Node System

[0068]

[0069] Table 3 Load requirements and number of users for each node in the 137-node case study

[0070]

[0071] The maximum output of photovoltaic components, wind turbine components, and energy storage components are shown in Table 4.

[0072] Table 4 Maximum output of photovoltaic components, wind turbine components and energy storage components

[0073] This method's implementation scheme uses typical annual solar irradiance and wind speed data from a region in my country, setting one sampling point every hour. The data is input into a distributed photovoltaic (PV) model and a distributed wind turbine output model, and then normalized to 0-1 to obtain 8760 hours of per-unit output data for PV and wind turbine units. Based on the hourly per-unit curves of PV output, wind turbine output, and load demand, 8760 time-series scenarios to be extracted for load reduction can be generated. These scenarios are then plotted according to their temporal distribution. Figure 4 middle.

[0074] After normalizing the data in all scenarios to 0-1, the K-Medoids clustering algorithm was used to reduce the number of scenarios from 8760. To mitigate the influence of randomly selected initial cluster centers, each scenario was clustered 10 times, and the optimal clustering result was taken. The curve of the calculated silhouette coefficient s as a function of the number of scenarios K is shown below. Figure 5 As shown.

[0075] As shown in the figure, the silhouette coefficient reaches the global optimum when the number of classification scenes K=8, with a corresponding silhouette coefficient s=0.92404. Figure 6 The image shows the scene clustering reduction results at this time, with eight different colors representing the clustering results of eight typical scenes.

[0076] The probability distributions of 8 typical scenarios are as follows Figure 7 As shown.

[0077] At this point, the per-unit values ​​of photovoltaic output, wind turbine output, and load demand for eight typical scenarios are shown in the table. These eight typical scenarios include typical scenarios such as abundant solar and wind power during the day, clear skies with no wind, strong winds at night, and no solar power and no wind at night, which conform to the typical temporal distribution characteristics of solar irradiance, wind speed, and load energy consumption in this region. Among them, the per-unit values ​​of photovoltaic output, wind turbine output, and load demand for typical scenarios are shown in Table 5.

[0078] Table 5. Per-unit values ​​of photovoltaic output, wind turbine output, and load demand in typical scenarios.

[0079] Based on the optimization model of formula (27), the island recovery matrix (IRM) of the entire system under eight typical scenarios can be obtained. The positions of elements with a value of 1 in the island recovery matrix of the entire system under these eight typical scenarios are colored according to different scenarios, such as... Figure 8 As shown, each colored point represents the node that can be powered back by the island when a branch fails. It can be seen that these elements are naturally divided into 4 regions, corresponding to 4 feeders connected to DG or energy storage, and the island restoration matrix in each scenario satisfies the radial connectivity constraint.

[0080] Based on the island recovery matrix (IRM) for each typical scenario, the number of outages, outage time and power shortage of each load node under each typical scenario can be calculated using equations (33) to (35).

[0081] The number of power outages at each load node under each typical scenario is as follows: Figure 9 As shown.

[0082] The number of power outages at each load node is exactly the same in each scenario, and is identical to the number of power outages without DG or energy storage. The power outage time at each load node in each typical scenario is as follows: Figure 10As shown, for load nodes on the feeder that are not connected to DG or energy storage (i.e., nodes 1-38, 75-84, and 85-98), the outage time is completely consistent across different scenarios and is the same as the outage time without DG or energy storage. For load nodes on the feeder that are connected to DG or energy storage, the outage time decreases significantly, but the degree of decrease varies across different scenarios.

[0083] The power shortage at each load node in each typical scenario is as follows: Figure 11 As shown, the power shortage at each load node varies under different scenarios, and even nodes with the same outage time will have different power shortages. In each typical scenario, the maximum power shortage occurs at node 105.

[0084] Based on load reliability indicators for each typical scenario, the number of power outages, outage duration, and power shortage at each load node within a given year are statistically analyzed. Figure 12 As shown. (Through) Figure 12 As can be seen, compared with scenarios without DG and energy storage, the number of power outages at all load nodes is exactly the same, the outage time at some load nodes is the same, and the power shortage at all load nodes has been reduced.

[0085] Based on the load reliability index for each typical scenario, the number of system power outages, outage time, and power shortage amount for each typical scenario were calculated, as shown in Table 6.

[0086] Table 6. Number of system power outages, outage duration, and power shortage in 8 typical scenarios.

[0087] Based on the system reliability indicators for each typical scenario in Table 6, the SAIFI, SAIDI, and EENS indicators of the power distribution system within the statistical period were calculated and are shown in Table 7.

[0088] Table 7 SAIFI, SAIDI and EENS Indicators of Power Distribution System

[0089] In summary, based on the four types of fault impact correlation matrices and the post-fault island recovery matrix, a quantitative assessment of the reliability indicators of the power distribution system covering multiple time-series scenarios of source and load was achieved for typical operating scenarios within the statistical period.

[0090] It should be emphasized that the embodiments described in this invention are illustrative rather than limiting. Therefore, this invention includes, but is not limited to, the embodiments described in the specific implementation. Any other implementations derived by those skilled in the art based on the technical solutions of this invention are also within the scope of protection of this invention.

Claims

1. A method for analyzing the reliability indicators of a power distribution system considering multi-time-sequence scenarios of source and load, characterized in that: Includes the following steps: Step 1: Based on the spatial correlation between the power supply path of the load node and the key components, classify the impact of branch faults on the load node in the power distribution system, establish a correlation matrix of four types of fault impacts, and provide a matrix analysis basis for the subsequent analytical calculation of reliability indicators. Step 2: Establish time-series operational reliability assessment models for components such as DG, energy storage devices, and flexible loads; clarify the islanding recovery potential and flexible load response characteristics of DG and energy storage under different operating states; and construct an islanding recovery matrix calculation method that considers capacity constraints and connectivity constraints under various fault scenarios. Step 3: Considering the time-series differences in the operation scenarios of the power distribution system caused by DG, energy storage and flexible loads, construct a typical scenario set of the power distribution system based on the K-Medoids clustering algorithm. While preserving the source-load time-series correlation of the power distribution system, reduce the number of scenarios to be evaluated and reduce computational overhead. Step 4: Based on the typical operating scenario set obtained by clustering, and combined with the established fault impact correlation matrix and island recovery matrix model, a reliability level analysis method for power distribution systems covering multiple time-series scenarios of source and load is proposed, and the reliability index of power distribution systems covering multiple time-series scenarios of source and load is analyzed and calculated.

2. The method for analyzing the reliability index of a power distribution system considering multi-time-sequence scenarios of source and load as described in claim 1, characterized in that: The specific implementation method of step 1 is as follows: the impact of branch faults on load nodes is summarized into four types: Type a: A branch fault causes all power supply paths to the load node to be interrupted. The load node needs to wait for the fault to be repaired before power can be restored. Therefore, the power outage time caused by the branch fault is the fault repair time of the branch. Type b: A branch fault causes all power supply paths to the load node to be interrupted. The load node needs to wait for the fault to be isolated before the main power supply can restore power. The power outage time caused by this branch fault is the fault isolation time. Type c: A branch fault causes all power supply paths to the load node to be interrupted. The load node needs to wait for the fault to be isolated before the backup power supply can be restored. The power outage time caused by this branch fault is the fault isolation and power supply transfer time. Type d: Branch faults have no impact on load nodes; Constructing a fault impact correlation matrix FEIM , FEIM include FEIMA , FEIM B , FEIM C and FEIM D , a Fault-related correlation matrix FEIMA for: ; b Fault-related correlation matrix FEIMB for: ; c Fault-related correlation matrix FEIMC for: ; d Fault-related correlation matrix FEIMD for: ; In the formula, the matrix This indicates that all elements are "1". 3D matrix, matrix This indicates that all elements are "0". 3D matrix, logical operators " indicates that the elements in the matrix are bitwise XORed.

3. The method for analyzing the reliability index of a power distribution system considering multi-time-sequence scenarios of source and load as described in claim 1, characterized in that: At a certain moment in step 2 t The modified model for the islanded recovery matrix of a single schedulable DG is as follows: ; In the formula, Indicates the revised version of the first... k In the island recovery matrix, the first... i Line 1 j Column elements, , They represent the first k Within an isolated island, schedulable DG is currently... t Every moment, whether contributing meritoriously or not, requires effort. , They represent the first k An unschedulable DG within an isolated island is currently... t Every moment, whether contributing meritoriously or not, requires effort. , They represent the first k An isolated energy storage element in the current t Efforts made at all times, whether they yield merit or not; In obtaining the first k Island recovery matrix IRM k Then, obtain a certain moment. t The whole system island recovery matrix IRM : ; in, N sch This indicates the amount of dispatchable DG and energy storage.

4. The method for analyzing the reliability index of a power distribution system considering multi-time-sequence scenarios of source and load as described in claim 1, characterized in that: The specific implementation method of step 3 is as follows: Euclidean distance is selected to characterize the distance between different classes. Silhouette coefficients are used. s Measuring the number of clusters K Validity: ; ; In the formula, m This represents the total number of sample points. Indicates sample To the i The mean distance of each sample point in the cluster. Indicates sample Mean distance to each sample point in the nearest cluster, number of clusters K The range of values ​​for is .

5. The method for analyzing the reliability index of a power distribution system considering multi-time-sequence scenarios of source and load as described in claim 1, characterized in that: Step 4 includes the following steps: Step 4.1: Count the number of all scenarios to be clustered and reduced. m Set typical scenario numbers K =2; Step 4.2, Random selection K One scenario was used as the initial clustering result; Step 4.3: Calculate the required reduction based on the distributed power output, energy storage output, and load demand in different operating scenarios. m From a single running scenario to a randomly selected one. K The Euclidean distance between the cluster centers (medoids); Step 4.4: Based on the principle of finding the closest cluster center point for each scene sample, sort the remaining samples... m - K Each scene is assigned to the class represented by the current best center point; Step 4.5: In each class, calculate the sum of distances from each scene to other scenes in that class according to equation (33), and take the point with the smallest sum of distances from other scenes in each class as the center point of that class, that is: (31) In the formula, Indicates the first i Sample points in a cluster Indicates the first j The iteration process of the iteration process i The center point of the cluster; Step 4.6: Repeat steps 4.3 to 4.5 until the new center point converges to the previous center point; Step 4.7: Calculate and record the profile coefficients. s And determine the number of clusters. K Has the value reached the upper limit? If the upper limit has not yet been reached, then take... K = K +1, jump to step 4.2; Step 4.8: Compare all K Profile coefficient under the given value s And the profile coefficient s When taking the maximum value K The values ​​and their corresponding clustering results are used as the final cluster number and clustering result.

Images