Active power distribution network balance area division method, equipment and medium

Abstract

The invention relates to an active power distribution network balance area division method, equipment and a medium. The method comprises the following steps: S1, collecting annual daily power data of a to-be-researched power distribution network load and a distributed power supply; s2, an improved WGAN algorithm based on a zoom dot product attention mechanism is adopted to extract a typical output scene of a load and a distributed power supply from the acquired data; s3, collecting electrical distances between feeder line blocks in the to-be-researched power distribution network; s4, calculating the supply and demand imbalance degree, the load similarity and the distance modularity of each feeder line block according to the data obtained in S2 and S3, and forming a fitness function; and S5, based on the fitness function, determining a power distribution network balance region division scheme by using a genetic algorithm. Compared with the prior art, the active power distribution network balance area is divided based on the electrical characteristics, the source load characteristics and the power utilization behavior factors, the self-balance capability of the power distribution network can be effectively exerted, the installation and maintenance cost of the active power distribution network is remarkably reduced, and the consumption rate of the distributed power supply and the source load balance level of the grid power distribution network are improved.

Description

Technical Field

[0001] The present invention relates to the field of distribution network planning, and in particular to a method, equipment and medium for dividing a balancing area of ​​an active distribution network. Background Art

[0002] With the massive access of distributed power sources and flexible loads, the distribution network faces unprecedented challenges. In the ever-changing power market environment, the power and load characteristics of different regions have gradually shown obvious differentiation, resulting in increasingly prominent power balance issues. This phenomenon not only affects the stability of the distribution network, but also puts higher requirements on the economic efficiency of the power market. Therefore, how to achieve a more economical and larger-scale power balance in a complex power system has become an important issue that needs to be solved urgently.

[0003] In order to achieve an effective balance of power, it is necessary to start with the zoning of the distribution network. The current distribution network structure is mostly a traditional centralized power supply model. In the face of emerging distributed power sources, traditional planning methods can no longer meet the needs.

[0004] Therefore, before active object joint planning, the existing distribution network needs to be optimized in terms of supply area to adapt to the dynamic changes in source and load characteristics. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method, equipment and medium for dividing the balancing area of ​​an active distribution network, which can effectively exert the self-balancing ability of the distribution network, significantly reduce the installation and maintenance costs of the active distribution network, and improve the distributed power supply absorption rate and the source-load balance level of the grid distribution network.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to a first aspect of the present invention, there is provided a method for dividing a balancing area of ​​an active power distribution network, comprising:

[0008] S1. Collect the daily power data of the distribution network load and distributed power sources to be studied throughout the year;

[0009] S2, using the improved WGAN algorithm based on the scaled dot product attention mechanism to extract typical output scenarios of loads and distributed generation from the collected data obtained in S1;

[0010] S3, collecting the electrical distances between the feeder blocks in the distribution network to be studied;

[0011] S4, based on the data obtained by S2 and S3, the supply and demand imbalance, load similarity and distance modularity of each feeder block are calculated to form a fitness function;

[0012] S5. Based on the fitness function, a genetic algorithm is used to determine the distribution network balancing area division scheme.

[0013] Preferably, the daily power data of the distribution network load and distributed power sources to be studied throughout the year are collected in S1, specifically: data is collected every fifteen minutes, and the daily power data of the distribution network load and distributed power sources to be studied throughout the year are obtained after aggregation.

[0014] Preferably, the improved WGAN algorithm is used in S2 to extract typical output scenarios of loads and distributed power sources from the collected data obtained in S1, which specifically includes the following sub-steps:

[0015] S21, using the improved WGAN algorithm based on the scaled dot product attention mechanism to generate scenes for the collected data obtained in S1;

[0016] S22, reducing the scene generated by S21 with the goal of retaining the maximum amount of information;

[0017] S23. The probability of each scenario after reduction is used as a weight to obtain a typical output scenario of the load and distributed power source.

[0018] Preferably, the improved WGAN algorithm is used to generate a scene for the collected data obtained in S1 in S21, which specifically includes:

[0019] The Wasserstein distance calculation formula and objective function are as follows:

[0020]

[0021] Where: V(D,G) is the cross entropy function of the two-class classification; E is the expected distribution; G(z) and p G (z) is the generated sample and its probability distribution; x and p data (x) is the real sample and its probability distribution; W(·) is the Wasserstein distance between G(z) and x; Ω(·) is p G (z) and p data The joint distribution set of (x);

[0022] Set weights for matching the input and output power of the generator and discriminator:

[0023]

[0024] Where: x in is the input matrix; K is the key value matrix used for learning; d K is the dimension of K; the softmax(·) function is used to normalize the weights.

[0025] Preferably, in S22, a scene reduction model is used to reduce the scenes generated in S21 with the goal of retaining the maximum amount of information, wherein the mathematical expression of the scene reduction model is:

[0026]

[0027] P peak =|1-ξ max,j / ξ max,i |

[0028] P trou =|1-ξ min,j / ξ min,i |

[0029]

[0030] Where: I R (ξ i ,ξ i ) is the amount of information about scene j in the typical scene set Z' to scene i in the generated scene set Z; N Z' and N Z are the number of scenes of Z' and Z respectively; p i is the probability of scene i appearing, and its sum is 1; C sim is the probability similarity between scenes i and j; P peak , P trou and E tot are the daily peak power information, valley power information and total power information of scenarios i and j respectively; ξ i,t is the power of scene i at time t; ξ max,i , min,i are the daily peak and valley values ​​of scene i respectively.

[0031] Preferably, in S4, based on the data obtained in S2 and S3, the supply and demand imbalance, load similarity and distance modularity of each block are calculated to form a fitness function, which specifically includes:

[0032] Calculate the supply-demand imbalance:

[0033]

[0034] Where: and It is divided into the load of feeder block i and the power value of DG typical scenario at time t; ε is a sufficiently small positive number;

[0035] The expression for calculating load similarity is:

[0036]

[0037]

[0038] Where: P i,t With P j,t are the comprehensive loads of feeder blocks i and j at time t in the peak-valley overlap period of a day; n is the number of peak-valley overlap moments in the daily comprehensive load curve; T is the total time of the daily load curve, which is taken as 24 in this paper; t ij is the peak-valley overlap time of blocks i and j within one day;

[0039] Calculate distance modularity:

[0040]

[0041]

[0042] Where: b is the sum of all edge weights of the grid; N m is the number of blocks in the mth balance zone; e ij is the weight of the edge connecting the center points of feeder blocks i and j; k i is the sum of all edge weights connected to feeder block i; L ij is the electrical distance between feeder blocks i and j.

[0043] Preferably, the fitness function is expressed as follows:

[0044]

[0045]

[0046]

[0047]

[0048]

[0049] Where: F is the fitness function; F 1 is the total degree of polymerization in the equilibrium zone; F 2 is the total coupling degree in the equilibrium region; f m is the correlation degree of the mth equilibrium area after combination; f max is the sum of the correlations between each block before combination; M is the number of balanced areas; N is the total number of blocks.

[0050] Preferably, the method of determining the distribution network balancing area division scheme based on the fitness function and using a genetic algorithm in S5 specifically includes the following sub-steps:

[0051] S51, according to the electrical connection characteristics, using the network adjacency matrix to encode the chromosome, generate multiple individuals, and form an initial population;

[0052] S52, calculating the fitness of each individual;

[0053] S53, calculating the sum of all individual fitness values, and generating a random number according to the ratio of the fitness to the total fitness, and determining the selected individual by the random number;

[0054] S54, pairing the first half and the second half of the population to form a crossover parent generation;

[0055] S55. Randomly select certain positions for each individual to mutate, and check whether the connectivity constraint and line N-1 constraint are met after the mutation. If not, trace the source of the relevant blocks and re-examine the possible connection situation;

[0056] S56, calculating the new fitness value of each individual after the mutation, and sorting them by fitness size;

[0057] S57, determine whether the set maximum number of iterations is reached or the convergence condition is met: if so, sort the individuals in the best individual set according to the fitness value and output the optimal partition result; otherwise, return to S52 and continue the selection, crossover and mutation operations.

[0058] According to a second aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the program, any one of the methods described above is implemented.

[0059] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, any one of the methods described above is implemented.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] The present invention proposes a method for dividing the active distribution network balancing area by considering factors such as the electrical characteristics, source-load characteristics, and electricity consumption behavior in the area. This method can effectively exert the self-balancing ability of the distribution network, meet the zoning principle of "high cohesion and low coupling", significantly reduce the installation and maintenance costs of active objects, and improve the distributed power supply absorption rate and the source-load balance level of the grid distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a flow chart of the method of the present invention;

[0063] Figure 2 It is the flow chart of genetic algorithm;

[0064] Figure 3 This is a schematic diagram of a power supply grid in an embodiment;

[0065] Figure 4Plan solutions for each scenario in the embodiment;

[0066] Figure 5 is the load characteristic of feeder block 1 in the embodiment;

[0067] Figure 6 The time-of-use electricity price for each scenario in the embodiment;

[0068] Figure 7 These are the equilibrium zone operation curves for Case 2 and Case 3 in the examples. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0070] Example

[0071] like Figure 1 As shown, this embodiment provides a method for dividing an active power distribution network balancing area, the method comprising the following steps:

[0072] S1. Collect the daily power curves of the loads at each node of the distribution network to be studied and the distributed generation DG throughout the year.

[0073] In this embodiment, the active power of one point is collected every fifteen minutes within twenty-four hours of a day.

[0074] The basic situation of the power grid in this embodiment is as follows: there are 34 feeder blocks as basic planning units, including 6 types of loads: residential, industrial, commercial, administrative, educational and medical. The upper limit of load capacity that can be reduced is 20% of the node load, and administrative, medical and educational loads do not participate in DR. It is also assumed that the DG output characteristics of each block in the same grid are the same, only the installed capacity is different, and it is only built at the industrial, commercial and residential load nodes. The specific block division and source-load characteristics are as follows: Figure 3 shown.

[0075] S2. Use the improved WGAN algorithm based on the scaled dot product attention mechanism to extract typical output scenarios of loads and distributed power sources from the collected data obtained in S1. The WGAN algorithm is a generative adversarial network based on Wasserstein distance, which specifically includes:

[0076] S21. Use the improved WGAN algorithm to generate scenes based on the data obtained in S1:

[0077] The Wasserstein distance calculation formula and objective function are as follows:

[0078]

[0079] Where: V(D,G) is the cross entropy function of the two-class classification; E is the expected distribution; G(z) and p G (z) is the generated sample and its probability distribution; x and p data (x) is the real sample and its probability distribution; W(·) is the Wasserstein distance between G(z) and x; Ω(·) is p G (z) and p data The joint distribution set of (x);

[0080] The weights are set for the matching degree of the input and output power of the generator and the discriminator. The weight formula is as follows:

[0081]

[0082] Where: x in is the input matrix; K is the key value matrix used for learning; d K is the dimension of K; the softmax(·) function is used to normalize the weights.

[0083] S22, reducing the generated scenes with the goal of retaining the maximum amount of information, the scene reduction model is as follows:

[0084]

[0085]

[0086]

[0087] P trou =|1-ξ min,j / ξ min,i |

[0088]

[0089] Where: I R (ξ i ,ξ i ) is the amount of information about scene j in the typical scene set Z' to scene i in the generated scene set Z; N Z' and N Z are the number of scenes of Z' and Z respectively; p i is the probability of scene i appearing, and its sum is 1; C sim is the probability similarity between scenes i and j; P peak , P trou and E tot are the daily peak power information, valley power information and total power information of scenarios i and j respectively; ξi,t is the power of scene i at time t; ξ max,i , min,i are the daily peak and valley values ​​of scene i respectively.

[0090] S23. The probability of each scenario after reduction is used as a weight to obtain the typical output scenario of the load and DG.

[0091] S3. Collect electrical distance data between feeder blocks.

[0092] S4. Use the load and DG typical scenario curve data obtained in S2 to calculate the supply and demand imbalance and load similarity. The calculation formulas for the imbalance and load similarity are as follows:

[0093]

[0094]

[0095]

[0096] Where: and It is divided into the load of feeder block i and the power value of the DG typical scenario at time t; ε is a sufficiently small positive number to avoid the denominator being 0; P i,t With P j,t are the comprehensive loads of feeder blocks i and j at time t in the peak-valley overlap period of a day; n is the number of peak-valley overlap moments in the daily comprehensive load curve; T is the total time of the daily load curve, which is taken as 24 in this paper; t ij It is the peak-valley overlap time of blocks i and j within one day.

[0097] The distance modularity is calculated using the electrical distance data between feeder blocks collected by S3. The calculation formula is as follows:

[0098]

[0099]

[0100] Where: b is the sum of all edge weights of the grid; N m is the number of blocks in the mth balance zone; e ij is the weight of the edge connecting the center points of feeder blocks i and j; k i is the sum of all edge weights connected to feeder block i; L ij is the electrical distance between feeder blocks i and j.

[0101] Forming the fitness function:

[0102]

[0103]

[0104]

[0105]

[0106]

[0107] Where: F is the fitness function; F 1 is the total degree of polymerization in the equilibrium zone; F 2 is the total coupling degree in the equilibrium region; f m is the correlation degree of the mth equilibrium area after combination; f max is the sum of the correlations between each block before combination; M is the number of balanced areas; N is the total number of blocks.

[0108] S5. According to the obtained fitness function, the genetic algorithm is used to determine the equilibrium area division scheme. The genetic algorithm flow chart is as follows: Figure 2 The specific steps are described as follows:

[0109] S51. According to the characteristics of electrical connection, the chromosomes are encoded using the network adjacency matrix, where 0 indicates no connection and 1 indicates connection. Multiple individuals are generated to form an initial population.

[0110] S52. Calculate the fitness of each individual.

[0111] S53, calculating the sum of all individual fitness values, and generating a random number according to the ratio of the fitness to the total fitness, and determining the selected individual by the random number.

[0112] S54. Pair the first half and the second half of the population to form a crossover parent generation.

[0113] S55. Randomly select certain positions for each individual to mutate. After the mutation, check whether the connectivity constraints and line N-1 constraints are met. If not, trace the relevant blocks and re-examine the possible connections.

[0114] S56. Calculate the new fitness value of each individual after the mutation and sort them by fitness size.

[0115] S57, judging whether the set maximum number of iterations is reached or the convergence condition is met. If yes, sort the individuals in the best individual set according to the fitness value and output the optimal partition result; otherwise, return to S52 and continue the selection, crossover and mutation operations.

[0116] In order to verify the superiority of the method proposed in this paper, the following three scenarios are set up to carry out case simulation.

[0117] Table 1 Three scenarios to be simulated

[0118]

[0119] At the same time, in order to further reflect the impact of each scenario on the active object configuration plan, energy storage is configured for them respectively with the goal of economic optimization. Follow the above steps in sequence to obtain the corresponding planning plan for each scenario as follows Figure 4 As shown in Table 2, the time-of-use electricity price strategy is as follows Figure 6 As shown. Among them, Figure 4 Parts (a), (b) and (c) correspond to scenarios Case 1, Case 2 and Case 3 respectively.

[0120] Table 2 Energy storage planning schemes for each scenario

[0121]

[0122]

[0123] Depend on Figure 4 As can be seen in (a), considering only the electrical characteristics for the balanced area division, the result is an independent feeder block 1, in which there are only two types of loads: residential and industrial. Further extraction of the load characteristics of this feeder block is as follows Figure 5 It can be seen that the area is dominated by industrial loads with a single form. It is impossible to use the timing complementarity between loads to alleviate the peak-to-valley difference. The peak-to-valley difference rate of the entire block reaches 60.87%, which is not conducive to the stable operation of the system.

[0124] The planning effects of each scenario are shown in Table 3 below.

[0125] Table 3 Planning effects of each scenario

[0126] Scenario <![CDATA[C ESS (10,000 yuan / day)]]> <![CDATA[I sell( 10,000 yuan / day)]]> Average wind and solar power curtailment rate (%) Case 1 10.31 79.46 11.26％ Case 2 10.55 77.36 13.72％ Case 3 1.95 85.23 4.71％

[0127] Comparing Case 1 and Case 3, it can be seen that Case 1 only uses active objects to balance the source and load of the grid distribution network. Therefore, in order to achieve a better source and load balance effect, it increases the peak-valley price difference rate of time-of-use electricity prices, that is, the price difference rates of residential electricity prices and industrial electricity prices are 3.56 and 4.18 respectively, which are higher than 2.55 and 3.42 in Case 3, in order to mobilize users' enthusiasm for participating in DR; on the other hand, it increases the configuration scale of ESS, C ESS Reached 103,100 yuan / day. Although this strategy successfully reduced I sell, but the average daily total cost of Case 1 is 897,700 yuan, higher than 871,800 yuan of Case 3, and the average wind and solar curtailment rate is 11.26%, higher than 4.71% of Case 3. This shows that only using active objects to balance the source and load of the grid distribution network will lead to the problem of increased action costs of active objects and insufficient regulation capabilities due to the failure to call the inherent source and load balancing capabilities of the blocks. In addition, because the time-of-use electricity price has a greater impact on the electricity consumption behavior of users in the region, it may result in reduced user satisfaction with electricity consumption.

[0128] Comparing Case 2 and Case 3, it can be seen that although Case 2 first adopts the method of dividing the balancing area to achieve power mutual assistance within the area, since only the electrical characteristics of the block are considered in the zoning index, it is still necessary to set a higher electricity price difference and a larger ESS capacity, especially the ESS configuration. ESS It is 105,500 yuan / day, which is much higher than Case 3's 19,500 yuan / day. Although Case 3 has a smaller ESS scale and a more moderate DR strategy, which reduces the regulation ability of the active object and the DR response, and the user's electricity purchase cost is higher than Case 2, its total cost is still 7,300 yuan / day lower than Case 2. This shows that in the process of dividing the balancing area, the two source-load characteristics of supply and demand imbalance and load similarity are fully considered. The use of power mutual assistance in the area can effectively reduce the action cost and regulation pressure of the active object, and obtain a better wind and solar absorption rate while reducing the change in user electricity consumption behavior, thereby improving the source-load balance level of the grid distribution network.

[0129] To further verify the superiority of the proposed method, the operating status of each equilibrium zone in Case 2 and Case 3 is analyzed as shown in Table 4 and Figure 7 As shown, Figure 7 (a) to (h) correspond to balance areas 1 to 5 in Case 2 and balance areas 1 to 3 in Case 3 in the following table.

[0130] Table 4 Operation status of each balance zone of Case 2 and Case 3

[0131]

[0132] Since Case 2 did not consider the source-load characteristics of the blocks during the partitioning process, the source-load characteristics of each zone still showed a differentiated trend: balanced zone 1 and balanced zone 3 showed power supply characteristics, especially balanced zone 1. The power consumption of the balance area 2 and the balance area 4 is -3.52, and there is a large amount of wind and solar power abandonment during 12:00-17:00. The balance area 2 and the balance area 4 show load characteristics, and only the source-load balance level of the balance area 5 is good. is -1.01. Figure 7It can be seen that the main charging period of ESS is 1:00-10:00, and the discharging period is 12:00-21:00, that is, ESS is mainly used to store excess electricity in balancing areas 1 and 3 during non-peak electricity price periods, and send it out during peak electricity consumption periods in balancing areas 2 and 4 to save electricity purchase costs for users in the area. It can be seen that the method of dividing the balancing area only considering electrical characteristics violates the original intention of setting up the balancing area: using the regulation capacity within the area to balance the source and load, thereby achieving the purpose of promoting DG consumption and reducing the action cost of active objects. It failed to change the current source-load differentiation status within the grid distribution network. Comparing Case 1 and Case 2 in Table 3 can further verify this conclusion. Compared with Case 1, Case 2, which divides the balancing area in advance, has a C ESS The average wind and solar power curtailment rates are higher than those in Case 1, which means that this zoning method not only fails to promote regional source-load balance, but even has a counterproductive effect.

[0133] Compared with Case 2, the equilibrium zones 2 and 3 obtained in Case 3 are The wind and solar curtailment rates are 0.89% and 0.29% respectively. The DR in the balance area 1 is 1.13 and 1.15 respectively. The autonomy and balance of the source and load in the area can be basically achieved only through DR. The wind and solar curtailment rates are 0.89% and 0.29% respectively. The value is -1.16. Since the wind and solar power abandonment phenomenon is more serious during 13:00-18:00 and the electricity price level is high, DR regulation is difficult to play a role. Therefore, ESS is configured to charge during the DG output surplus periods of 1:00-3:00 and 13:00-17:00, and discharge during the DG output insufficient periods of 6:00-8:00 and 21:00-22, so as to further adjust the load curve in the area and reduce the wind and solar power abandonment rate in this area from the original 20.81% to 12.94%.

[0134] In summary, by using the active distribution network balancing area division method based on source-load and electrical characteristics proposed in the present invention, the self-balancing ability of the distribution network can be effectively exerted, the zoning principle of "high cohesion and low coupling" can be met, the installation and maintenance costs of active objects can be significantly reduced, and the DG absorption rate and the source-load balance level of the grid distribution network can be improved.

[0135] The electronic device of the present invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0136] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0137] The processing unit performs the various methods and processes described above, such as methods S1 to S5. For example, in some embodiments, methods S1 to S5 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1 to S5 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S5 in any other appropriate manner (e.g., by means of firmware).

[0138] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip systems (SOCs), load programmable logic devices (CPLDs), and the like.

[0139] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0140] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0141] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for dividing the balancing area of ​​an active distribution network, characterized in that: include: S1. Collect the daily power data of the distribution network load and distributed power sources to be studied throughout the year; S2, using the improved WGAN algorithm based on the scaled dot product attention mechanism to extract typical output scenarios of loads and distributed generation from the collected data obtained in S1; S3, collecting the electrical distances between the feeder blocks in the distribution network to be studied; S4, based on the data obtained in S2 and S3, calculate the supply and demand imbalance, load similarity and distance modularity of each feeder block to form a fitness function; S5. Based on the fitness function, a genetic algorithm is used to determine the distribution network balancing area division scheme.

2. The method according to claim 1, characterized in that In the S1, the daily power data of the distribution network load and distributed power source to be studied throughout the year are collected. Specifically, the data is collected every fifteen minutes, and the daily power data of the distribution network load and distributed power source to be studied throughout the year are obtained after aggregation.

3. The method according to claim 1, characterized in that: In S2, the improved WGAN algorithm is used to extract typical output scenarios of loads and distributed power sources from the collected data obtained in S1, which specifically includes the following sub-steps: S21, using the improved WGAN algorithm based on the scaled dot product attention mechanism to generate scenes for the collected data obtained in S1; S22, reducing the scene generated by S21 with the goal of retaining the maximum amount of information; S23. The probability of each scenario after reduction is used as a weight to obtain a typical output scenario of the load and distributed generation.

4. The method according to claim 3, characterized in that In S21, the improved WGAN algorithm is used to generate a scene for the collected data obtained in S1, which specifically includes: The Wasserstein distance calculation formula and objective function are as follows: Where: V(D,G) is the cross entropy function of the two-class classification; E is the expected distribution; G(z) and p G (z) is the generated sample and its probability distribution; x and p data (x) is the real sample and its probability distribution; W(·) is the Wasserstein distance between G(z) and x; Ω(·) is p G (z) and p data The joint distribution set of (x); Set weights for matching the input and output power of the generator and discriminator: Where: x in is the input matrix; K is the key value matrix used for learning; d K is the dimension of K; the softmax(·) function is used to normalize the weights.

5. The method according to claim 3, characterized in that: In S22, a scene reduction model is used to reduce the scenes generated in S21 with the goal of retaining the maximum amount of information, wherein the mathematical expression of the scene reduction model is: P peak =|1-ξ max,j / x max,i P trou =|1-ξ min,j / x min,i Where: I R (ξ i ,ξ i ) is the amount of information about scene j in the typical scene set Z' to scene i in the generated scene set Z; N Z' and N Z are the number of scenes of Z' and Z respectively; p i is the probability of scene i appearing, and its sum is 1; C sim is the probability similarity between scenes i and j; P peak , P trou and E tot are the daily peak power information, valley power information and total power information of scenarios i and j respectively; ξ i,t is the power of scene i at time t; ξ max,i , min,i are the daily peak and valley values ​​of scene i respectively.

6. The method according to claim 5, characterized in that In S4, the supply-demand imbalance, load similarity and distance modularity of each block are calculated based on the data obtained by S2 and S3 to form a fitness function, which specifically includes: Calculate the supply-demand imbalance: Where: and It is divided into the load of feeder block i and the power value of DG typical scenario at time t; ε is a sufficiently small positive number; The expression for calculating load similarity is: Where: P i,t With P j,t are the comprehensive loads of feeder blocks i and j at time t in the peak-valley overlap period of a day; n is the number of peak-valley overlap moments in the daily comprehensive load curve; T is the total time of the daily load curve, which is taken as 24 in this paper; t ij is the peak-valley overlap time of blocks i and j within one day; Calculate distance modularity: Where: b is the sum of all edge weights of the grid; N m is the number of blocks in the mth balance zone; e ij is the weight of the edge connecting the center points of feeder blocks i and j; k i is the sum of all edge weights connected to feeder block i; L ij is the electrical distance between feeder blocks i and j.

7. The method according to claim 6, characterized in that The fitness function, mathematical expression is: Where: F is the fitness function; F1 is the total aggregation degree of the equilibrium zone; F2 is the total coupling degree of the equilibrium zone; f m is the correlation degree of the mth equilibrium area after combination; f max is the sum of the correlations between each block before combination; M is the number of balanced areas; N is the total number of blocks.

8. The method according to claim 1, characterized in that In S5, based on the fitness function, a genetic algorithm is used to determine a distribution network balancing area division scheme, which specifically includes the following sub-steps: S51, according to the electrical connection characteristics, using the network adjacency matrix to encode the chromosome, generate multiple individuals, and form an initial population; S52, calculating the fitness of each individual; S53, calculating the sum of all individual fitness values, and generating a random number according to the ratio of the fitness to the total fitness, and determining the selected individual by the random number; S54, pairing the first half and the second half of the population to form a crossover parent generation; S55. Randomly select certain positions for each individual to mutate, and check whether the connectivity constraint and line N-1 constraint are met after the mutation. If not, trace the source of the relevant blocks and re-examine the connection status; S56, calculating the new fitness value of each individual after the mutation, and sorting them by fitness size; S57, determine whether the set maximum number of iterations is reached or the convergence condition is met: if so, sort the individuals in the best individual set according to the fitness value and output the optimal partition result; otherwise, return to S52 and continue the selection, crossover and mutation operations.

9. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images