Method and system for evaluating carrying capacity of power distribution network

Abstract

The invention discloses a power distribution network bearing capacity evaluation method and system. The method comprises the following steps: acquiring photovoltaic output error data; inputting the photovoltaic output error data into the improved generative adversarial network for continuous training to obtain photovoltaic output scenes, clustering the photovoltaic output scenes to obtain feature scenes and feature scene probability distribution, and constructing a feature scene probability distribution uncertainty set by adopting KL divergence; based on the photovoltaic output scene and the power distribution network constraint condition, establishing a power distribution network bearing capacity evaluation model; and based on the feature scene probability distribution uncertainty set and the power distribution network bearing capacity evaluation model, establishing a power distribution network bearing capacity evaluation distribution robust optimization model, and obtaining a power distribution network bearing capacity evaluation value and a power distribution network optimal decision scheme. Errors caused by a traditional scene generation method are effectively improved; the distribution network zone area bearing capacity evaluation distribution robust optimization model constructed by using KL divergence can flexibly adjust the conservative property of the solution, and also considers the accuracy and robustness at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution network optimization, and in particular to a method and system for evaluating the carrying capacity of a power distribution network. Background Art

[0002] Under the background of "dual carbon", distributed photovoltaics are widely used in power systems as an important form of renewable energy. As an important part of the distribution system, the low-voltage distribution network is a key hub for coordinating the power generation end and the user end. How to manage the distribution network area to improve its carrying capacity and improve the economic efficiency of the distribution network operation is of great research significance. At the same time, the output of distributed photovoltaics has great uncertainty, which brings huge challenges to the stable economic operation of the distribution network area.

[0003] Considering the randomness of distributed photovoltaic output in distribution network substations, the random optimization method or robust optimization method is used to deal with the randomness of distributed photovoltaic output, and the distribution network substation carrying capacity model results are often not accurate or too conservative. Therefore, the distributed robust optimization model is often used for processing. The implementation of distributed robustness requires clustering of a large amount of historical data to obtain representative feature scenarios and their probability distribution values. However, it is difficult to obtain a large amount of historical data on distributed photovoltaic output in distribution network substations, and it is usually generated by technical means.

[0004] Traditional distributed photovoltaic scene generation mostly uses statistical methods. First, a specific mathematical model needs to be established to describe the probability distribution of distributed photovoltaics, and then a scene set is constructed through a random sampling method. The generation model based on statistical methods is highly interpretable, but the statistical assumptions suitable for the probability distribution are specific expressions. In practical applications, it is difficult to determine a mathematical model suitable for multiple types of distributed photovoltaics. In addition, mathematical models often have problems with complex parameter solutions. The scene generation method based on the adversarial generative network is applied to the study of random optimal scheduling, and it is proved that the adversarial generative network can accurately fit the probability distribution of renewable energy samples and generate high-quality scenario sets. However, the adversarial generative network model was originally proposed for image generation. Its network structure does not fully consider the characteristics of time series, and it is difficult to capture the time series dependence of the joint probability distribution of time series data. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a distribution network carrying capacity evaluation method to solve the problem that the traditional generative adversarial network is difficult to fully characterize the time characteristics, day-ahead mode characteristics and seasonality of distributed photovoltaic output, thereby affecting the accurate evaluation of the distribution network area carrying capacity.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a method for evaluating the carrying capacity of a distribution network, comprising:

[0009] Obtaining distributed photovoltaic output data, and obtaining photovoltaic output error data according to the distributed photovoltaic output data;

[0010] The photovoltaic output error data is input into an improved generative adversarial network, the generator and discriminator of the improved generative adversarial network are used to continuously train to obtain photovoltaic output scenarios, the photovoltaic output scenarios are clustered to obtain feature scenarios and feature scenario probability distributions, and then KL divergence is used to construct a feature scenario probability distribution uncertainty set;

[0011] Based on the photovoltaic output scenario and distribution network constraints, a distribution network carrying capacity evaluation model is established with the consumption rate, voltage deviation rate and line loss rate of distributed photovoltaics as evaluation indicators;

[0012] Based on the uncertainty set of the probability distribution of the characteristic scenarios and the distribution network carrying capacity assessment model, a distribution network carrying capacity assessment distributed robust optimization model is established, and the distribution network carrying capacity assessment distributed robust optimization model is solved to obtain the distribution network carrying capacity assessment value and the distribution network optimal decision-making plan.

[0013] As a preferred solution of the method for evaluating the carrying capacity of a distribution network described in the present invention, the generator and discriminator of the improved generative adversarial network include:

[0014] Introducing a time series convolutional network into the generator, and using the time series convolutional network to extract photovoltaic output prediction error data;

[0015] The discriminator of the improved generative adversarial network adopts an Informer architecture, and uses the discriminator to identify the photovoltaic output prediction error data and the photovoltaic output error data;

[0016] The improved generative adversarial network is continuously trained and optimized to obtain photovoltaic output prediction error data, and the photovoltaic output prediction error data is combined with the photovoltaic output prediction data to obtain the photovoltaic output scenario.

[0017] As a preferred solution of the distribution network carrying capacity evaluation method described in the present invention, wherein: using KL divergence to construct a characteristic scenario probability distribution uncertainty set includes:

[0018] The KL divergence is used to constrain the error between the probability distribution of the feature scene and the probability distribution of the real scene, and the uncertainty set of the probability distribution of the feature scene is constructed, which is expressed as:

[0019]

[0020] Among them, Q n is the uncertainty set of the probability distribution of the characteristic scene, N s is the number of feature scenes after clustering, p s is the true probability value under scenario s, is the initial probability of clustering under scene s, and ξ is the KL divergence value between the initial probability distribution and the true scene probability distribution.

[0021] As a preferred solution of the distribution network carrying capacity evaluation method described in the present invention, wherein: using the consumption rate of distributed photovoltaics, voltage deviation rate and line loss rate as evaluation indicators, establishing a distribution network carrying capacity evaluation model includes:

[0022] Based on photovoltaics, the objective function is established with the goal of minimizing the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaics, which is expressed as:

[0023] F=min(-ω1f1+ω2f2+ω3f3)

[0024] Among them, f1 is the absorption rate of distributed photovoltaics, f2 is the voltage deviation rate of the distribution network, f3 is the line loss rate, and ω1, ω2, and ω3 are the weight coefficients of the three evaluation indicators respectively.

[0025] As a preferred solution of the method for evaluating the carrying capacity of a distribution network according to the present invention, the distribution network carrying capacity evaluation distribution robust optimization model includes:

[0026] On the basis of the uncertainty set of probability distribution of characteristic scenarios, according to the types of variables in the distribution network substation carrying capacity assessment model, discrete variables are set as first-stage decision variables, and continuous variables are set as second-stage decision variables, and a distributed robust optimization model for distribution network substation carrying capacity assessment is established.

[0027] As a preferred solution of the distribution network carrying capacity evaluation method described in the present invention, the distribution network carrying capacity evaluation value and the optimal decision strategy include:

[0028] The distribution network carrying capacity assessment robust optimization model is decomposed into a main problem and sub-problems by using the C&CG algorithm;

[0029] According to clustering, characteristic scenarios and probability distribution of characteristic scenarios are obtained, and the optimal solution of the discrete variables in the first stage is obtained by solving the main problem. The optimal solution of the discrete variables in the first stage is the optimal decision-making scheme of the distribution network;

[0030] The sub-problem searches for the uncertainty set of the characteristic scenario probability distribution under the optimal solution of the first-stage discrete variables obtained from the main problem, and obtains the true characteristic scenario probability distribution with the maximum objective function value of the distribution network carrying capacity evaluation model;

[0031] When the C&CG algorithm iteration reaches the convergence condition, an optimal solution is output, and the optimal solution includes the optimal values ​​of the first-stage decision variables and the second-stage decision variables;

[0032] According to the optimal solution, the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaics are calculated to obtain an evaluation value of the distribution network's carrying capacity.

[0033] As a preferred solution of the distribution network carrying capacity evaluation method described in the present invention, the distribution network constraints include distributed photovoltaic flexible operation constraints, OLTC constraints, ESS constraints, reactive power compensation constraints, load reduction constraints and flow constraints.

[0034] In a second aspect, the present invention provides a system for evaluating the carrying capacity of a distribution network, comprising:

[0035] A data acquisition module, used to acquire distributed photovoltaic output data, and acquire photovoltaic output error data according to the distributed photovoltaic output data;

[0036] An improved generative adversarial network training module is used to input the photovoltaic output error data into the improved generative adversarial network, and continuously train and obtain photovoltaic output scenarios using the generator and discriminator of the improved generative adversarial network;

[0037] A feature scene acquisition module is used to cluster the photovoltaic output scenes to obtain feature scenes and feature scene probability distributions, and then use KL divergence to construct a feature scene probability distribution uncertainty set;

[0038] An evaluation module establishment module is used to establish a distribution network carrying capacity evaluation model based on the photovoltaic output scenario and distribution network constraints, taking the consumption rate, voltage deviation rate and line loss rate of distributed photovoltaic as evaluation indicators;

[0039] An evaluation value acquisition module is used to establish a distribution network carrying capacity evaluation distributed robust optimization model based on the uncertainty set of the characteristic scenario probability distribution and the distribution network carrying capacity evaluation model, solve the distribution network carrying capacity evaluation distributed robust optimization model, and obtain the distribution network carrying capacity evaluation value and the distribution network optimal decision-making plan.

[0040] In a third aspect, the present invention provides a computing device, comprising:

[0041] Memory and processor;

[0042] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the distribution network carrying capacity evaluation method are implemented.

[0043] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the distribution network carrying capacity assessment method.

[0044] Compared with the prior art, the present invention has the following beneficial effects: the improved generative adversarial network proposed in the present invention can generate scenarios with consistent time series change characteristics and fluctuations with actual data, effectively improving the errors caused by traditional scenario generation methods; the distribution network substation carrying capacity assessment method proposed in the present invention obtained by active management measures based on OLTC, reactive compensation device, ESS, distributed photovoltaic flexible operation mode and demand-side management measures for load reduction has a better effect on improving the distribution network substation carrying capacity; the distribution network substation carrying capacity assessment robust optimization model constructed based on KL divergence proposed in the present invention can flexibly adjust the conservatism of the solution while taking into account accuracy and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0046] Figure 1 It is a schematic diagram of the overall process of a method for evaluating the carrying capacity of a distribution network according to an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of a solution process of a C&CG algorithm of a distribution network carrying capacity evaluation method according to an embodiment of the present invention;

[0048] Figure 3 A schematic diagram of an improved distribution network area test system for a distribution network carrying capacity evaluation method according to an embodiment of the present invention;

[0049] Figure 4 It is a schematic diagram comparing a scenario generated by an improved adversarial generative network and a historical data scenario of a distribution network carrying capacity assessment method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0053] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0054] At the same time, in the description of the present invention, it should be noted that the orientations or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the system or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0055] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0056] Example 1

[0057] Reference Figure 1-3 , is an embodiment of the present invention, and provides a method for evaluating the carrying capacity of a distribution network, comprising:

[0058] S100: Obtain distributed photovoltaic output data, and obtain photovoltaic output error data according to the distributed photovoltaic output data;

[0059] Specifically, the actual output data of the distributed photovoltaic system is collected and obtained first. This process usually relies on intelligent monitoring equipment installed at each distributed photovoltaic site. These devices can record and transmit the power generation data of the photovoltaic system at different time points in real time; based on the historical photovoltaic output data, the photovoltaic output error data is further calculated. The photovoltaic output error data refers to the deviation value between the actual output and the predicted output or theoretical output, which reflects the uncertainty of the photovoltaic system's power generation capacity. The key to this step is to use statistical methods and data analysis techniques, such as time series analysis and regression analysis, to process the collected photovoltaic output data to extract the error characteristics of the photovoltaic output, which provides an important basis for subsequent photovoltaic output prediction and distribution network carrying capacity assessment.

[0060] S102: Inputting the photovoltaic output error data into the improved generative adversarial network, using the generator and discriminator of the improved generative adversarial network to continuously train and obtain photovoltaic output scenarios, clustering the photovoltaic output scenarios to obtain feature scenarios and feature scenario probability distributions, and then using KL divergence to construct a feature scenario probability distribution uncertainty set;

[0061] Preferably, a time series convolutional network is introduced into the generator, and the photovoltaic output prediction error data is extracted using the time series convolutional network;

[0062] In the embodiment of the present application, the traditional generative adversarial network is improved, and an improved generative adversarial network combining information and time series is proposed;

[0063] Specifically, a generator based on a temporal convolutional network is constructed; considering that distributed photovoltaic output has obvious time correlation, that is, there is a strong correlation between data in adjacent time periods, when the generator generates the dynamic characteristics of the prediction error at time t, it should make full use of the information at the previous t-1 moment and have the ability to capture the gradual characteristics between the dynamic feature moments; for this reason, the present invention selects a time feature extraction model with autoregressive characteristics to construct a generator network.

[0064] The generators of existing time series generation models are usually implemented based on recurrent neural networks, i.e. RNNs. However, the serial structure of RNNs causes the time and memory they consume to be proportional to the length of the sequence, resulting in low computational efficiency of the generator model and the tendency to ignore local temporal features. Unlike sequence models with sequential structures, such as RNNs, temporal convolutional networks, i.e. TCNs, use causal convolution and dilated convolution structures to consider global and local temporal patterns and reduce unnecessary neuron connections to increase training speed.

[0065] Therefore, the present invention introduces TCN into the generator network, aiming to use its special convolutional structure to perform parallel calculations on time series information and efficiently and fully mine time series features;

[0066] The random noise z of length L, the label c of the day-ahead forecast error, and the timestamp st are used as the input of the generator G to generate the photovoltaic output forecast error data Output through the generator G, the generator is expressed as:

[0067]

[0068] Among them, Z and Z t are the static and dynamic characteristics of random noise, and are the static and dynamic feature vectors generated by the generator respectively;

[0069] Preferably, the discriminator of the improved generative adversarial network adopts an Informer architecture, and uses the discriminator to identify photovoltaic output prediction error data and photovoltaic output error data;

[0070] Preferably, the improved generative adversarial network is continuously trained and optimized to obtain photovoltaic output prediction error data, and the photovoltaic output prediction error data is combined with the photovoltaic output prediction data to obtain the photovoltaic output scenario;

[0071] It should be noted that the most widely used model in time series tasks is Transformer, which can extract features of input data in parallel and model long-term and short-term time series features; however, its secondary calculation leads to high computational complexity and large memory usage; in addition, the stacking of multiple encoders and decoders makes the total memory of the model very large, which limits its scalability when receiving long sequence inputs, so Transformer has certain limitations in solving long time series problems. In the study of distributed photovoltaic scenario power generation on the day-ahead time scale, the input of the model is a long time series of photovoltaic forecast errors with a time resolution of 1h and a length of 24. Informer uses distillation operations on the stacked layers of encoders and decoders to reduce spatial complexity, enabling the model to effectively receive and process long sequences.

[0072] Therefore, the present invention constructs an Informer-based discriminator network, which consists of ProbSpare self-attention, Distilling, Multihead attention and Fully connection layers;

[0073] Specifically, firstly, the key features of the long time series are obtained through ProbSpare self-attention; the real photovoltaic output error data H n The generated photovoltaic output prediction error data Multiply by w q 、w k 、w v Weight coefficients, forming a query Q, key K, and value V matrix, L Q =L K =L V =L,d h H n and The dimension of the data, dot product operation on the matrix, get H n and The attention results between the data at each moment of the sequence; Since the results of the self-attention score analysis show a long-tail feature, that is, only a few dot product pairs have significant contributions, in order to distinguish the dot product pairs that can produce significant contributions, it is necessary to adjust the elements q of the Q matrix i Perform sparsity measure calculation and define the i-th sparsity measure of the Q matrix as:

[0074]

[0075] in, is the computed dot product of the ith query and the jth transposed key, and L is the length of the K matrix, which is the same size as the day-ahead PV error sequence.

[0076] Assume that the M(q i ,K) is larger and its attention probability is more diverse. In this case, the calculated dot product pair has a higher probability of making an important contribution to the self-attention result, that is, the characteristics of the PV prediction error data at time i are more related to the data at other times of the sequence;

[0077] Sort the results of the sparsity measure to obtain the one with the largest M(q i ,K) of U queries, U-type queries are called active queries, and the remaining LU-type queries are called lazy queries. Each K only obtains self-attention through U active queries, expressed as:

[0078]

[0079] in, It consists of U active queries, It is the self-attention calculation result of actively querying the corresponding row;

[0080] It should be noted that for lazy queries, the row corresponding to the self-attention score is directly replaced by the mean of the V matrix, and the two parts of the self-attention results are concatenated to obtain the self-attention feature F P .

[0081] Secondly, considering the self-attention result F P The redundant information of the V matrix in the above example is combined to reduce the time dimension of the input sequence through the distillation layer, further capturing F P The most effective attention information F h , expressed as:

[0082] F h =Distilling(F p )=MaxPool(ELU(Conv1d(F p )))

[0083] Among them, Conv1d is the convolution layer, ELU is the activation function, and MaxPool is the maximum pooling layer;

[0084] Finally, using the day-ahead forecast error Errnd sequence in the original feature space as an excitation is conducive to the discriminator to quickly distinguish true and false error features. Therefore, the present invention inputs the Errnd segment of length O into the new ProbSpare self-attention layer and Distilling layer to extract the key feature F in the original feature space. o ; Then calculate F through multi-head attention o and F h The mutual attention results between them are used to exchange information, and then the judgment results of the authenticity of data features and the accuracy of labels are output through the fully connected layer;

[0085] The discriminator is represented as:

[0086]

[0087] in, H n and The general name of f prob 、f distill 、f AT 、f FC Corresponding to the ProbSpare self-attention layer, Distilling layer, Multi-head attention layer and fully connected layer of the discriminator network, M(q i ,K) and It is the result of judging the authenticity and label accuracy of the low-dimensional error feature set, and sou represents the data source.

[0088] Preferably, the KL divergence is used to constrain the error between the probability distribution of the feature scene and the probability distribution of the real scene, and the uncertainty set of the probability distribution of the feature scene is constructed, which is expressed as:

[0089]

[0090] Among them, Q n is the uncertainty set of the probability distribution of the characteristic scene, N s is the number of feature scenes after clustering, p s is the true probability value under scenario s, is the initial probability of clustering under scene s, ξ is the KL divergence value between the initial probability distribution and the real scene probability distribution;

[0091] It should be noted that by inputting the photovoltaic output error data into the improved generative adversarial network, combining it with the time series convolutional network to effectively extract the time series characteristics of the prediction error, and using the discriminator of the Informer architecture to enhance the recognition ability of the error data, the accurate generation of photovoltaic output scenarios is achieved; then, the characteristic scenes and their probability distributions are extracted by the clustering method, and the uncertainty set is constructed using the KL divergence, which effectively quantifies the error between the probability distribution of the characteristic scenes and the probability distribution of the real scenes, thereby significantly improving the accuracy and reliability of the photovoltaic output forecast, and providing a more solid data support for the scheduling and planning of the power system.

[0092] S104: Based on the photovoltaic output scenario and distribution network constraints, a distribution network carrying capacity evaluation model is established with the consumption rate, voltage deviation rate and line loss rate of distributed photovoltaics as evaluation indicators;

[0093] Preferably, an objective function is established based on photovoltaics with the goal of minimizing the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaics, which can be expressed as:

[0094] F=min(-ω1f1+ω2f2+ω3f3)

[0095] Among them, f1 is the absorption rate of distributed photovoltaics, f2 is the voltage deviation rate of the distribution network, f3 is the line loss rate, ω1, ω2, and ω3 are the weight coefficients of the three evaluation indicators respectively;

[0096] In the embodiment of the present application, the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaic are expressed as follows:

[0097]

[0098] Among them, f1 is the absorption rate of distributed photovoltaic, f2 is the voltage deviation rate of the distribution network, f3 is the line loss rate, T is the total number of time periods, is the active power generated by distributed photovoltaic at node i in time period t, Ω P is the set of distributed photovoltaic installation nodes in the distribution network, N bus is the total number of nodes in the distribution network, U0 is the voltage reference value of the distribution network, Ω L is the line set of the distribution network, l max is the maximum value of the square of the line current;

[0099] Preferably, the distribution network constraints include distributed photovoltaic flexible operation constraints, OLTC constraints, ESS constraints, reactive power compensation constraints, load reduction constraints and power flow constraints;

[0100] ① Distributed photovoltaic flexible operation constraints:

[0101] The operation modes of distributed photovoltaics in the distribution network area include black start mode, PV control mode and PQ control mode. According to the current state of the system, working in different operation modes can play a role in adjusting the voltage level and regulating the power flow. The flexible operation constraints of distributed photovoltaics are expressed as:

[0102]

[0103] in, is the active power input from the photovoltaic power plant at node i in time period t, is the maximum limit of the active power input from distributed photovoltaic power station i in time period t, i.e. the predicted power, is the photovoltaic capacity at node i, U base is the square value of the voltage when the photovoltaic is started in black start mode, U i,0 is the square of the expected voltage value on photovoltaic node i. When distributed photovoltaic is in PV operation mode, U can be flexibly selected according to the system status of the distribution network. i,0 The value of U i,t is the square of the voltage value on photovoltaic node i at time period t, U max and U min The node voltages U i,t The maximum and minimum values ​​of is the reactive power injection amount of photovoltaic at photovoltaic node i, and are the upper and lower limits of the photovoltaic reactive power injection amount on photovoltaic node i, g i,t , are binary state auxiliary variables introduced. Since the black start mode is usually used in power supply restoration, the present invention does not consider the black start mode of distributed photovoltaics, that is, g i,t =0, indicating that the distributed photovoltaic is in PV or PQ control mode. When the voltage value on the photovoltaic node i is fixed to U i,0 , at this time the photovoltaic is in PV operation mode, when or When , the reactive power of the PV at node i will be fixed at its upper and lower limits, and the PV is in PQ operation mode.

[0104] 2. OLTC Constraints

[0105] OLTC is a voltage regulating device usually installed between the distribution network station and the upper transmission network. It plays a role in regulating voltage and maintaining system operation stability in the distribution network. The OLTC constraint is expressed as:

[0106]

[0107] Among them, o is the auxiliary node added on the line ij where OLTC is located, U o,t is the square value of the voltage at node o in time period t, r ij is the resistance value on line ij, P ij,t is the active power flowing into line ij during time period t, x ij is the reactance value on line ij, Q ij,t is the reactive power flowing into line ij during time period t, l ij,t is the current value on line ij during time period t, N ij,t is the gear position of OLTC on line ij in time period t, λ ij,n,t To convert a binary number, K ij is the maximum number of OLTC gears on line ij, is the minimum transformation ratio of OLTC on line ij, Δk ij,t is the unit increment of OLTC on line ij in time period t, m ij,t 、x ij,n,t ,y ij,n,t is the intermediate variable, and M is the constant introduced by the Big-M method.

[0108] ③ESS constraints

[0109] The charge and discharge status of ESS can be used to control the active power in the distribution network area and improve the carrying capacity of distributed photovoltaics. According to the working state of ESS, the ESS constraint is expressed as:

[0110]

[0111]

[0112] in, and Indicates the working status of ESS. When ESS is in charging state, When ESS works in the discharge state, When , ESS is in a balanced working state, without charging or discharging. and are the minimum and maximum charging power of ESS on node j, and are the minimum and maximum discharge power of ESS on node j, H j,t is the charge of ESS on node j in time period t, and are the charging and discharging rates of energy storage at point j, and are the minimum and maximum values ​​of the charge respectively.

[0113] ④Reactive power compensation constraints

[0114] The reactive power compensation device adjusts the reactive power distribution of the distribution network system to enable distributed photovoltaics to generate more active power and improve the consumption rate. The reactive power compensation constraint is expressed as:

[0115]

[0116] in, is the reactive power regulation of the static VAR compensator (SVC) at node j in time period t, and are the minimum and maximum values ​​of the SVC adjustable reactive power at node j, is the reactive power regulation of the group switching capacitor (CB) at node j in time period t, is the number of capacitor banks put into operation at node j in time period t, is the unit reactive compensation power of CB at node j, and are the increase and decrease of the number of CB groups put into operation at node j in time period t, B CB,max The maximum number of groups that can be put into operation by CB.

[0117] ⑤ Load reduction constraints

[0118] Load reduction measures are strategies that use the controllable characteristics of the load in the distribution network area to adjust the power in the distribution network area from the user demand side. The load reduction constraint is expressed as:

[0119]

[0120] in, and are the active power and reactive power provided to the user at node i in time period t, and are respectively the active and reactive power required by the user at node i in time period t, and are the active power and reactive power that can be reduced by the user at node i in time period t, and are respectively the active power and reactive power that have been reduced at node i in time period t.

[0121] ⑥ The power flow constraint is expressed as:

[0122]

[0123] Among them, P j,t is the active input at node j in time period t, and are the active and reactive power input to the distribution network gateway at node j in time period t, and The minimum and maximum values ​​of the squared voltage at node i, and are the minimum and maximum values ​​of the active power input at node j, and are the minimum and maximum values ​​of the reactive power input at node j respectively.

[0124] S106: Based on the uncertainty set of probability distribution of characteristic scenarios and the distribution network carrying capacity assessment model, a distribution network carrying capacity assessment distributed robust optimization model is established, the distribution network carrying capacity assessment distributed robust optimization model is solved, and the distribution network carrying capacity assessment value and the distribution network optimal decision-making plan are obtained;

[0125] Preferably, based on the uncertainty set of the probability distribution of characteristic scenarios, according to the types of variables in the distribution network substation carrying capacity assessment model, discrete variables are set as first-stage decision variables, and continuous variables are set as second-stage decision variables, and a distributed robust optimization model for distribution network substation carrying capacity assessment is established;

[0126] In the embodiment of the present application, the distributed robust optimization model for evaluating the carrying capacity of the distribution network area is expressed as:

[0127]

[0128] st Ax≥b,x∈{0,1}

[0129] hair

[0130]

[0131] Among them, N sis the number of feature scenes after K-means clustering, the subscript s represents the sequence number of the feature scene, x is the discrete decision variable of the first stage, and y s is the continuous decision variable in the second stage, Ω x is the constraint set of the variables in the first stage, Ω y is the constraint set of the second stage variables, a, A, b, C, D, e, F, g, J, q, c, d, G, u s is the constant matrix corresponding to the variable constraints.

[0132] Preferably, the C&CG algorithm is used to decompose the distribution network carrying capacity assessment and robust optimization model into a main problem and sub-problems; the solution process of the C&CG algorithm is as follows: Figure 2 As shown, when the difference between the upper and lower bounds is less than the given convergence accuracy ρ, the optimal solution is output;

[0133] Preferably, characteristic scenarios and probability distributions of characteristic scenarios are obtained by clustering, and the optimal solution of the discrete variables in the first stage is obtained by solving the main problem. The optimal solution of the discrete variables in the first stage is the optimal decision-making scheme for the distribution network.

[0134] In the embodiment of the present application, the main problem is to make a decision on the discrete variables of the first stage under the input of the probability distribution of the characteristic scene obtained by clustering, obtain the optimal decision solution, and determine a lower bound L in the iterative process of the C&CG algorithm. B , expressed as:

[0135]

[0136] st Ax≥b,x∈{0,1}

[0137]

[0138] Among them, N s is the number of feature scenes after clustering by K-means, y s is the continuous decision variable of the second stage, the subscript k is the number of iterations when solving using C&CG, a, A, b, C, D, e, F, g, J, q, c, d, G, u s is the constant matrix corresponding to the variable constraints;

[0139] Preferably, the subproblem searches for a characteristic scenario probability distribution uncertainty set under the optimal solution of the first-stage discrete variables obtained from the main problem, and obtains a true characteristic scenario probability distribution with the maximum objective function value of the distribution network carrying capacity assessment model;

[0140] In the embodiment of the present application, the sub-problem model is the optimal decision solution x obtained from the main problem decision. *Under the constraint of the uncertainty set of feature scenes, the worst-case probability distribution value p of the real feature scene is searched. s,k+1 , and determines an upper bound U in the iterative process of the C&CG algorithm B , expressed as:

[0141]

[0142] Among them, p s,k+1 is the probability distribution value of the real feature scene in the worst case, y s is the continuous decision variable in the second stage, and the subscript k is the number of iterations when solving using C&CG;

[0143] Because y s and p s The constraint sets of have no intersection, so they are solved separately, expressed as:

[0144]

[0145] Among them, N s is the number of feature scenes after clustering by K-means, y s is the continuous decision variable of the second stage, the subscript k is the number of iterations when solving using C&CG, and p s,k+1 is the probability distribution value of the real feature scenario in the worst case;

[0146] Preferably, when the C&CG algorithm iteration reaches a convergence condition, an optimal solution is output, and the optimal solution includes the optimal values ​​of the first-stage decision variables and the second-stage decision variables;

[0147] Preferably, according to the optimal solution, the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaics are calculated to obtain the distribution network carrying capacity assessment value;

[0148] It should be noted that the present invention significantly improves the planning and decision-making capabilities of the distribution network in an uncertain environment by constructing a robust optimization model for distribution network carrying capacity assessment based on a set of uncertainty in the probability distribution of characteristic scenarios, and adopts the C&CG algorithm for efficient solution. The present invention innovatively processes discrete variables and continuous variables in stages to ensure the accuracy and robustness of the decision-making scheme; by iteratively solving the main problem and sub-problems, not only the optimal decision-making scheme for the distribution network is obtained, but also the carrying capacity of the distribution network is accurately evaluated, including key indicators such as photovoltaic absorption rate, voltage deviation rate and line loss rate. This technical achievement provides strong theoretical support and practical guidance for the intelligent and reliable operation of the distribution network, effectively responds to the complex challenges brought about by the access of new energy, and promotes the sustainable development of the energy system.

[0149] The above is a schematic scheme of a distribution network carrying capacity evaluation method of this embodiment. It should be noted that the technical scheme of the distribution network carrying capacity evaluation system and the technical scheme of the distribution network carrying capacity evaluation method described above belong to the same concept. For details not described in detail in the technical scheme of the distribution network carrying capacity evaluation system in this embodiment, please refer to the description of the technical scheme of the distribution network carrying capacity evaluation method described above.

[0150] The distribution network carrying capacity evaluation system in this embodiment includes:

[0151] A data acquisition module is used to acquire distributed photovoltaic output data, and acquire photovoltaic output error data according to the distributed photovoltaic output data;

[0152] An improved generative adversarial network training module is used to input photovoltaic output error data into the improved generative adversarial network, and continuously train the generator and discriminator of the improved generative adversarial network to obtain photovoltaic output scenarios;

[0153] The feature scene acquisition module is used to cluster the photovoltaic output scenes to obtain the feature scenes and the probability distribution of the feature scenes, and then use the KL divergence to construct the uncertainty set of the probability distribution of the feature scenes;

[0154] The evaluation module is used to establish a distribution network carrying capacity evaluation model based on the photovoltaic output scenario and distribution network constraints, taking the consumption rate, voltage deviation rate and line loss rate of distributed photovoltaic as evaluation indicators;

[0155] The evaluation value acquisition module is used to establish a distribution network carrying capacity evaluation distributed robust optimization model based on the uncertainty set of characteristic scenario probability distribution and the distribution network carrying capacity evaluation model, solve the distribution network carrying capacity evaluation distributed robust optimization model, and obtain the distribution network carrying capacity evaluation value and the distribution network optimal decision-making plan.

[0156] This embodiment further provides a computing device, which is applicable to the case of evaluating the carrying capacity of a distribution network, and includes:

[0157] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the distribution network carrying capacity evaluation method proposed in the above embodiment.

[0158] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for evaluating the carrying capacity of a distribution network proposed in the above embodiment is implemented.

[0159] The storage medium proposed in this embodiment and the method for implementing the distribution network carrying capacity evaluation proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0160] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0161] Example 2

[0162] Reference Figure 4 Table 1 and Table 1 are an embodiment of the present invention, which provides a distribution network carrying capacity evaluation method. In order to verify its beneficial effects, comparison results of several solutions are provided.

[0163] In order to verify the effectiveness of the distributed photovoltaic output data generated by the improved adversarial generation network proposed in the present invention, the present invention compares and analyzes the test set data in the selected generation scenario and historical scenario through the time series correlation measurement index, such as Figure 4 As shown;

[0164] from Figure 4 It can be seen that the distributed photovoltaic output generated by the improved adversarial generative network proposed in the present invention has basically the same change trend in the time series, and the autocorrelation coefficient and partial correlation coefficient of the generated scenario and the historical scenario at each time lag are very similar, which verifies the effectiveness of the improved adversarial generative network proposed in the present invention in generating scenarios, and the generated distributed photovoltaic output and user demand scenarios can effectively reflect their characteristics of changing over time.

[0165] In order to verify the influence of the distribution network area carrying capacity improvement strategy proposed in the present invention on the distribution network area carrying capacity, the present invention compares and analyzes the strategy results under 6 different management measures, as shown in Table 1;

[0166] Table 1 Comparison of evaluation values ​​of different management measures

[0167]

[0168]

[0169] It can be seen from the table that the comprehensive index of Strategy 6 proposed by the present invention, which comprehensively considers various active management measures and demand-side management measures, is the lowest, so the distribution network station area has the strongest carrying capacity.

[0170] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for evaluating the carrying capacity of a distribution network, characterized in that: include: Obtaining distributed photovoltaic output data, and obtaining photovoltaic output error data according to the distributed photovoltaic output data; The photovoltaic output error data is input into an improved generative adversarial network, the generator and discriminator of the improved generative adversarial network are used to continuously train to obtain photovoltaic output scenarios, the photovoltaic output scenarios are clustered to obtain feature scenarios and feature scenario probability distributions, and then KL divergence is used to construct a feature scenario probability distribution uncertainty set; Based on the photovoltaic output scenario and distribution network constraints, a distribution network carrying capacity evaluation model is established with the consumption rate, voltage deviation rate and line loss rate of distributed photovoltaics as evaluation indicators; Based on the uncertainty set of the probability distribution of the characteristic scenarios and the distribution network carrying capacity assessment model, a distribution network carrying capacity assessment distributed robust optimization model is established, and the distribution network carrying capacity assessment distributed robust optimization model is solved to obtain the distribution network carrying capacity assessment value and the distribution network optimal decision-making plan.

2. The method for evaluating the carrying capacity of a distribution network according to claim 1, characterized in that: The generator and discriminator of the improved generative adversarial network include: Introducing a time series convolutional network into the generator, and using the time series convolutional network to extract photovoltaic output prediction error data; The discriminator of the improved generative adversarial network adopts an Informer architecture, and uses the discriminator to identify the photovoltaic output prediction error data and the photovoltaic output error data; The improved generative adversarial network is continuously trained and optimized to obtain photovoltaic output prediction error data, and the photovoltaic output prediction error data is combined with the photovoltaic output prediction data to obtain the photovoltaic output scenario.

3. The method for evaluating the carrying capacity of a distribution network according to claim 1, characterized in that: The KL divergence is used to construct the uncertainty set of feature scene probability distribution, including: The KL divergence is used to constrain the error between the probability distribution of the feature scene and the probability distribution of the real scene, and the uncertainty set of the probability distribution of the feature scene is constructed, which is expressed as: Among them, Q n is the uncertainty set of the probability distribution of the characteristic scene, N s is the number of feature scenes after clustering, p s is the true probability value under scenario s, is the initial probability of clustering under scene s, and ξ is the KL divergence value between the initial probability distribution and the true scene probability distribution.

4. The method for evaluating the carrying capacity of a distribution network according to claim 3, characterized in that: Taking the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaic as evaluation indicators, the distribution network carrying capacity evaluation model is established, including: Based on photovoltaics, the objective function is established with the goal of minimizing the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaics, which is expressed as: F=min(-ω1f1+ω2f2+ω3f3) Among them, f1 is the absorption rate of distributed photovoltaics, f2 is the voltage deviation rate of the distribution network, f3 is the line loss rate, and ω1, ω2, and ω3 are the weight coefficients of the three evaluation indicators respectively.

5. The method for evaluating the carrying capacity of a distribution network according to claim 4, characterized in that: The distribution network carrying capacity assessment and optimization model includes: On the basis of the uncertainty set of probability distribution of characteristic scenarios, according to the types of variables in the distribution network substation carrying capacity assessment model, discrete variables are set as first-stage decision variables, and continuous variables are set as second-stage decision variables, and a distributed robust optimization model for distribution network substation carrying capacity assessment is established.

6. The method for evaluating the carrying capacity of a distribution network according to claim 5, characterized in that: The distribution network carrying capacity assessment value and optimal decision-making strategy include: The distribution network carrying capacity assessment robust optimization model is decomposed into a main problem and sub-problems by using the C&CG algorithm; According to clustering, characteristic scenarios and probability distribution of characteristic scenarios are obtained, and the optimal solution of the discrete variables in the first stage is obtained by solving the main problem. The optimal solution of the discrete variables in the first stage is the optimal decision-making scheme of the distribution network; The sub-problem searches for the uncertainty set of the characteristic scenario probability distribution under the optimal solution of the first-stage discrete variables obtained from the main problem, and obtains the true characteristic scenario probability distribution with the maximum objective function value of the distribution network carrying capacity evaluation model; When the C&CG algorithm iteration reaches the convergence condition, an optimal solution is output, and the optimal solution includes the optimal values ​​of the first-stage decision variables and the second-stage decision variables; According to the optimal solution, the absorption rate, voltage deviation rate and line loss rate of distributed photovoltaics are calculated to obtain an evaluation value of the distribution network's carrying capacity.

7. The method for evaluating the carrying capacity of a distribution network according to claim 1, characterized in that: The distribution network constraints include distributed photovoltaic flexible operation constraints, OLTC constraints, ESS constraints, reactive power compensation constraints, load reduction constraints and power flow constraints.

8. A system for evaluating the carrying capacity of a distribution network, characterized in that: include, A data acquisition module, used to acquire distributed photovoltaic output data, and acquire photovoltaic output error data according to the distributed photovoltaic output data; An improved generative adversarial network training module is used to input the photovoltaic output error data into the improved generative adversarial network, and continuously train and obtain photovoltaic output scenarios using the generator and discriminator of the improved generative adversarial network; A feature scene acquisition module is used to cluster the photovoltaic output scenes to obtain feature scenes and feature scene probability distributions, and then use KL divergence to construct a feature scene probability distribution uncertainty set; An evaluation module establishment module is used to establish a distribution network carrying capacity evaluation model based on the photovoltaic output scenario and distribution network constraints, taking the consumption rate, voltage deviation rate and line loss rate of distributed photovoltaic as evaluation indicators; An evaluation value acquisition module is used to establish a distribution network carrying capacity evaluation distributed robust optimization model based on the uncertainty set of the characteristic scenario probability distribution and the distribution network carrying capacity evaluation model, solve the distribution network carrying capacity evaluation distributed robust optimization model, and obtain the distribution network carrying capacity evaluation value and the distribution network optimal decision-making plan.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the distribution network carrying capacity assessment method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for evaluating the carrying capacity of a distribution network as claimed in any one of claims 1 to 7.

Images