Optimal configuration method and device for photovoltaic power station to participate in power grid voltage regulation
By optimizing the installed capacity of photovoltaic power plants using the Catboost model and heuristics, the problem of low configuration efficiency of photovoltaic power plants is solved, the stability of the power grid and the accuracy of calculation are improved, and efficient configuration of photovoltaic power plants is achieved.
Patent Information
- Application Number
- CN202411168313.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-11-22
AI Technical Summary
The low configuration efficiency of photovoltaic power plants affects the stability of the power grid, and the power flow calculation process is complex, resulting in reduced configuration solution efficiency.
A data-driven model for power flow calculation is constructed using the Catboost model. Combined with a two-layer optimization configuration model for photovoltaic power plants, a heuristic algorithm is used to optimize the installed capacity of photovoltaic power plants. By acquiring the operating data of distribution network nodes, voltage deviation and network loss are calculated to construct the optimal configuration scheme for photovoltaic power plants.
It improves the configuration rate and calculation accuracy of photovoltaic power plant installed capacity, reduces network losses, and enhances the power grid's power supply stability and computational adaptability.
Smart Images

Figure CN119010188A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of power grid configuration, and in particular to a method and device for optimizing the configuration of a photovoltaic power station participating in power grid voltage regulation. Background Art
[0002] New energy power generation represented by wind power and photovoltaics has become an important part of achieving energy structure adjustment, but new energy power generation has the characteristics of randomness, volatility, indirectness and weak regulation. Large-scale grid connection will affect the stability of the power grid.
[0003] The access of photovoltaic power stations to the grid will change the flow direction and increase the difficulty of grid planning. The topological structure of the grid, line parameters, access points and access capacity of photovoltaic power stations determine the impact of photovoltaic access on grid loss, power quality and reliability. Moreover, with the increase of photovoltaic power station access, the distribution network structure becomes complicated, which slows down the flow calculation process and reduces the efficiency of photovoltaic power station configuration solution. Therefore, reasonable planning of the access location and access capacity of photovoltaic power stations, as well as effective coordination of the active and reactive ratio of photovoltaic power output, can make the photovoltaic power output play a good supporting role in the grid voltage, especially the terminal voltage. Summary of the invention
[0004] The present application provides a method and device for optimizing the configuration of a photovoltaic power station participating in grid voltage regulation, so as to solve the problem of low configuration efficiency of photovoltaic power stations in the prior art.
[0005] In a first aspect, the present application provides a method for optimizing the configuration of a photovoltaic power station participating in grid voltage regulation, including:
[0006] Obtaining the operating data of each node of the distribution network, and inputting the operating data of each node of the distribution network into a power flow calculation data-driven model, and outputting the voltage deviation and annual network loss of each node, wherein the power flow calculation data-driven model is constructed based on the Catboost model;
[0007] Constructing a two-layer optimization configuration model for a photovoltaic power station, the two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model, the upper model takes minimizing the annual operating cost of the distribution network as an optimization goal, and the lower model takes minimizing the annual network loss and node voltage deviation of the distribution network as an optimization goal;
[0008] Based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the double-layer optimization configuration model of the photovoltaic power station, the optimal configuration scheme of the photovoltaic power station is obtained, and the optimal configuration scheme is the installed capacity of the photovoltaic power station.
[0009] In a second aspect, the present application provides an optimization configuration device for a photovoltaic power station to participate in grid voltage regulation, including:
[0010] A data acquisition module is used to acquire the operating data of each node of the distribution network, and input the operating data of each node of the distribution network into a power flow calculation data-driven model, and output the voltage deviation and annual network loss of each node. The power flow calculation data-driven model is constructed based on the Catboost model;
[0011] A model building module is used to build a two-layer optimization configuration model for a photovoltaic power station, wherein the two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model, wherein the upper model takes the minimization of the annual operating cost of the distribution network as an optimization target, and the lower model takes the minimization of the annual network loss and node voltage deviation of the distribution network as an optimization target;
[0012] The scheme determination module is used to obtain the optimal configuration scheme of the photovoltaic power station based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the double-layer optimization configuration model of the photovoltaic power station, and the optimal configuration scheme is the installed capacity of the photovoltaic power station.
[0013] The present application provides a method and device for optimizing the configuration of a photovoltaic power station participating in grid voltage regulation. The method obtains the operating data of each node in the distribution network, inputs the operating data of each node in the distribution network into a flow calculation data-driven model, outputs the voltage deviation and annual network loss of each node, and the flow calculation data-driven model is constructed based on a Catboost model; a two-layer optimization configuration model for a photovoltaic power station is constructed, and the two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model. The upper model takes the minimization of the annual operating cost of the distribution network as an optimization goal, and the lower model takes the minimization of the annual network loss of the distribution network and the node voltage deviation as an optimization goal; based on a heuristic algorithm, the voltage deviation and annual network loss of each node in the distribution network, and the two-layer optimization configuration model of the photovoltaic power station, an optimal configuration scheme of the photovoltaic power station is obtained, and the optimal configuration scheme is the installed capacity of the photovoltaic power station. This application uses a power flow calculation data-driven model constructed by the Catboost model to calculate the voltage deviation of each node in the distribution network and the annual network loss, which can increase the calculation accuracy and improve the adaptability on the time scale; by constructing a two-layer optimization configuration model for photovoltaic power stations and using the heuristic method to solve the two-layer optimization configuration model for photovoltaic power stations, the installed capacity of the photovoltaic power station is obtained, which can improve the configuration rate of the installed capacity of the photovoltaic power station. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0015] Figure 1It is a flow chart of the implementation of the optimization configuration method for photovoltaic power stations participating in grid voltage regulation provided in the embodiment of the present application;
[0016] Figure 2 It is a structural diagram of the CatBoost framework provided in the embodiment of the present application;
[0017] Figure 3 It is a schematic diagram of segmentation and traversal of classification nodes provided in an embodiment of the present application;
[0018] Figure 4 It is a structural schematic diagram of an optimized configuration device for a photovoltaic power station participating in grid voltage regulation provided in an embodiment of the present application. DETAILED DESCRIPTION
[0019] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.
[0020] In order to make the purpose, technical solutions and advantages of the present application clearer, specific embodiments will be described below in conjunction with the accompanying drawings.
[0021] Figure 1 The implementation flow chart of the optimization configuration method for photovoltaic power stations participating in grid voltage regulation provided in the embodiment of the present application is described in detail as follows:
[0022] In step 101, the operating data of each node in the distribution network is obtained, and the operating data of each node in the distribution network is input into a power flow calculation data-driven model, and the voltage deviation and annual network loss of each node are output. The power flow calculation data-driven model is constructed based on the Catboost model.
[0023] In the embodiment of the present application, the operation data of each node of the distribution network is obtained, wherein the operation data includes the topological structure between each node of the distribution network, the line parameters of the distribution network, the load data of each node in the distribution network, and the meteorological data corresponding to each node in the distribution network, and the meteorological data may include light intensity and ambient temperature. The purpose of obtaining the topological structure is to ensure that the installed capacity of the photovoltaic power station obtained subsequently meets the demand of the distribution network.
[0024] Then, the obtained operation data of each node of the distribution network is input into the trained power flow calculation data-driven model to obtain the voltage deviation of each node and the annual network loss of the distribution network. Among them, the voltage deviation of each node is the difference between the actual operating voltage of each node and the voltage reference value.
[0025] In a possible implementation, after obtaining the operating data of each node in the distribution network, the method may further include:
[0026] Using the distribution network operation data, the static voltage stability index and loss sensitivity of each node in the distribution network are calculated;
[0027] Use the entropy weight method to assign weights to each node;
[0028] Based on the weighted method, the static voltage stability index, loss sensitivity and weight of each node, the comprehensive evaluation index of each node is calculated;
[0029] Select the node corresponding to the minimum comprehensive evaluation index from all comprehensive evaluation indexes as the target node;
[0030] Accordingly, the optimal configuration scheme of the photovoltaic power station can include:
[0031] The optimal configuration scheme of the photovoltaic power station connected to the target node of the distribution network is obtained.
[0032] In this embodiment, the static voltage stability index (SVSI) and loss sensitivity factors (LSFs) are used as indicators, and a comprehensive evaluation index for the site selection of a photovoltaic power station is calculated by a weighted method.
[0033] Among them, the static voltage stability index is used to judge the impact of node access to photovoltaic power on voltage support. When the SVSI value is closer to 1, the static voltage stability of the system is worse. Therefore, deploying photovoltaic power at nodes with higher SVSI is more likely to reduce node voltage deviation.
[0034] From the perspective of grid losses, it is easier to reduce system losses after network nodes with high loss sensitivity are connected to photovoltaics.
[0035] The specific process of calculating the comprehensive evaluation index is as follows:
[0036] Input the operating data of each node in the distribution network into the first formula to obtain the static grid stability index and loss sensitivity of each node, namely:
[0037]
[0038] Among them, SVSI is the static voltage stability index, LSFs is the loss sensitivity, P b,load is the active load of node b, Q b,load is the reactive load of node b, P ab,L is the active power loss of branch ab, X ab is the branch reactance of branch ab, R abis the branch resistance of branch ab, U a is the node voltage of node a, U b is the node voltage of node b.
[0039] After calculating the static voltage stability index SVSI and loss sensitivity LSFs of each node in the distribution network, the entropy weight method is first used to weight the static voltage stability index SVSI and loss sensitivity LSFs of each node in the distribution network. Then, based on the topological structure of the distribution network, the weighted method is used to input the static voltage stability index SVSI and loss sensitivity LSFs of each node and the corresponding weight coefficient into the second formula to obtain the comprehensive evaluation index of each node, where the second formula is:
[0040]
[0041] Among them, CEISS is a comprehensive evaluation index, η1 is the first weight coefficient, and η2 is the second weight coefficient.
[0042] Finally, from all the comprehensive evaluation indicators, the node corresponding to the minimum comprehensive evaluation index is selected as the node location where the distribution network is connected to the photovoltaic power station.
[0043] The embodiments of the present application can fully consider the adaptability of each node of the distribution network to access photovoltaic power when selecting the site for the photovoltaic power station (i.e., the target node connected to the distribution network), and on this basis, optimize the capacity of the access point of the photovoltaic power station.
[0044] In a possible implementation, before inputting the operation data of each node of the distribution network into the power flow calculation data-driven model and outputting the voltage deviation of each node and the annual network loss, the method may further include:
[0045] Preprocess the operating data of each node in the distribution network for state estimation;
[0046] Accordingly, the operation data of each node of the distribution network is input into the power flow calculation data-driven model, and the voltage deviation and annual network loss of each node are output, which may include:
[0047] The preprocessed operating data of each node in the distribution network is input into the power flow calculation data-driven model, and the voltage deviation of each node and the annual network loss are output.
[0048] Optionally, state estimation is performed based on the acquired operating data of each node in the distribution network, that is, the collected original data is calculated according to the best estimation criterion based on the accuracy of the measurement and Kirchhoff's law to obtain the best estimate value closest to the actual state of the system. In this embodiment, the best estimate criterion can be selected as the least squares criterion, and the original acquired operating data is calculated to obtain the best estimate value closest to the actual state of the system, thereby improving data accuracy.
[0049] Correspondingly, the operation data of each node of the distribution network after state estimation preprocessing is input into the power flow calculation data-driven model, and the voltage deviation of each node and the annual network loss are output.
[0050] In one possible implementation, the construction and training process of the power flow calculation data-driven model is as follows:
[0051] Obtain the historical operation data of each node in the distribution network, and obtain the corresponding historical voltage deviation and historical annual network loss of each node;
[0052] Build a machine learning model based on the Catboost model;
[0053] Taking the historical operation data of each node in the distribution network as input and the corresponding historical voltage deviation and historical annual network loss of each node as output, the machine learning model is trained to obtain a data-driven model for power flow calculation.
[0054] Among them, the power flow calculation data-driven model is trained offline.
[0055] Additionally, for the Catboost model:
[0056] Generally, the hyperparameters in the Catboost model can be determined by trial and error, but the trial and error method can only obtain a relatively good Catboost model, and there is a large degree of randomness and contingency, and the performance of the Catboost algorithm cannot be fully utilized. In this regard, this embodiment uses Bayesian theory to optimize the Catboost machine learning model.
[0057] CatBoost is a GBDT framework based on oblivious trees, which has fewer parameters, supports categorical variables and has high accuracy. It mainly solves the pain point of efficiently and reasonably processing categorical features. In addition, CatBoost also solves the problems of gradient bias and prediction shift, thereby reducing the occurrence of overfitting and improving the accuracy and generalization ability of the algorithm. CatBoost framework reference Figure 2 , including algorithm principles and engineering optimization.
[0058] by Figure 2 As shown, the algorithm principles include the second-order Taylor expansion of the objective function, tree-based regularization, node splitting criterion, column sampling and learning rate, and sparse-aware algorithm. Engineering optimization includes parallel column block design, cache access, and out-of-core block calculation.
[0059] Among them, the algorithm principle:
[0060] 1) Second-order Taylor expansion of the objective function
[0061] Like GBDT, Catboost is an additive model that optimizes only the sub-model in the current step in each iteration. In the mth step:
[0062] F m (x i )=F m-1 (x i )+f m (x i )
[0063] Among them, F m (x i ) is the first m sub-models that have been fixed after training, F m-1 (x i ) is the first m-1 sub-models that have been fixed after training, f m (x i ) is the sub-model of the current step, x i is the i-th sample value.
[0064] The objective function is empirical risk and structural risk (regular term):
[0065]
[0066] Among them, the regularization term Ω(f j ) is the complexity of the sub-model and has nothing to do with the second-order Taylor expansion.
[0067] Taylor's formula is a method of approximating a function f(x) with an order derivative at x=x0 using an n-order polynomial about Δx=x-x0. Catboost uses a second-order expansion to approximate the loss function, namely:
[0068]
[0069] F m-1 (x i ) is regarded as x0, f m (x i ) is regarded as Δx, As about Function, we can get:
[0070]
[0071] The first m-1 sub-models have been determined, so except for the equation about f m The (x) parts are all constants and do not affect f m (x) is the optimal solution. The objective function can be transformed into:
[0072]
[0073] Among them, L is the loss function, which measures the prediction performance. m-1 When (x) is determined, a g can be easily calculated for each sample point i i and h i .
[0074] 2) Tree-based regularization
[0075] The base classifiers supported by Catboost include decision trees and linear models. To prevent overfitting, Catboost sets the tree-based complexity as a regularization term:
[0076]
[0077] Among them, T is the number of leaf nodes of tree f, ω is the vector composed of the output regression values of all leaf nodes, ||ω|| 2 is the square of the L2 norm (modulus) of the vector, and γ and λ are both hyperparameters.
[0078] Then the objective function can be obtained from formula (1) and formula (2), as shown below:
[0079]
[0080] Next, a mathematical process is used to combine the regularization term and the empirical risk term. The empirical risk term is summed at the sample level and converted to the sum at the leaf node level.
[0081] Define the sample set on node j as I(j) = {x i |q(x i )=j}, where q(x i ) is the index function that maps samples to leaf nodes, and the regression value on leaf node j is ω i =f m (x i ),i∈I(j).
[0082]
[0083] Further simplify the expression, let ∑ i∈I(j) g i=G j ,∑ i∈I(j) h i =H j , note that G j and H j They are all functions of j:
[0084]
[0085] At this point, if the structure of a tree has been determined, then the samples (x i ,y i ,g i ,h i ) is also determined, that is, the GHT is determined, and the regression value output by each leaf node should minimize the above formula, from the quadratic function extreme point:
[0086]
[0087] After outputting the regression value according to this rule, the objective function value, that is, the score of the tree, is as follows. The smaller the value, the better the tree structure. Observe the following formula, the score of the tree can also be understood as the sum of the scores of all leaf nodes:
[0088]
[0089] 3) Node Splitting Criteria
[0090] Like the decision tree model, Catboost's sub-model tree relies on the greedy criterion of node recursive splitting to achieve tree generation. In addition, Catboost also supports approximate algorithms to solve the problem of excessive data volume exceeding memory or parallel computing requirements.
[0091] (1) Greedy Principle
[0092] The basic idea is the same as CART. After sorting the eigenvalues, traverse the partition points, take the optimal split benefit as the split benefit of the feature, select the feature with the optimal split benefit as the partition feature of the current node, and perform binary partitioning according to its optimal partition point to obtain the left and right subtrees.
[0093] (2) Approximate algorithm
[0094] Catboost also provides an approximate version of the above greedy criterion. In short, the feature quantile is used as the candidate point for partitioning, which reduces the set of candidate points for partitioning from traversing all samples to traversing between several quantiles.
[0095] Specifically, there are two optional strategies for selecting feature quantiles: global and local. The global strategy selects from the feature values of all samples, and is performed once before the root node is split; the local strategy selects from the feature values of the samples contained in the node to be split, and is performed before each node is split. Usually, since the global strategy can only be split once, its division granularity needs to be finer.
[0096] (3) Weighted quantile
[0097] From formula (2), we can see that setting the partial derivative to 0 easily yields This objective function is understood as h i is the weight, is the quadratic loss function of the label:
[0098]
[0099] Therefore, when the approximate algorithm takes the quantile, Catboost actually takes the second-order derivative h i , is the weighted quantile (Weighted Quantile Sketch).
[0100] 4) Column sampling and learning rate
[0101] Catboost also introduces two features: column sampling and learning rate.
[0102] Column sampling is a method used in random forests. The set of selected features for each node split is not all the remaining features, but a subset of the remaining features. This is to better combat overfitting and reduce computational overhead.
[0103] The learning rate, also known as step size or shrinkage, is multiplied by the coefficient before each sub-model (i.e., on the regression value of each leaf node) to weaken the influence of each tree and make the iteration more stable. The default setting of Catboost is 0.3.
[0104] 5) Sparse perception
[0105] The missing value coping strategy is something that the algorithm needs to consider. Feature sparsity also needs to be considered, such as when a large number of 0s appear in some features or when they are simply one-hot encoded. Catboost uses a sparse-aware strategy to handle both problems at the same time: in general, missing values and sparse 0 values are treated as missing values, and then these missing values are "bound" together, so that the traversal of the split nodes will skip the entirety of the missing values, which greatly improves the computational efficiency.
[0106] The split nodes are still obtained by traversal. There are two directions for NA, based on which the non-missing values are split and traversed. Alternatively, NA is divided into a fixed direction, and the non-missing values are split and traversed in ascending and descending order.
[0107] Reference Figure 3 If a feature value is 1, 2, 5, and a large number of NA, Catboost will traverse the above 6 cases (3 non-missing value split points x two directions of missing values), and the maximum split benefit is the split benefit on this feature. At the same time, NA will be divided into the right node.
[0108] In addition, during the project optimization:
[0109] 1) Parallel column block design
[0110] Catboost sorts each column of features in advance, stores them in the cache in the form of blocks, and uses the index to store the feature values and gradient statistics g i ,h i Correspondingly, the sorted blocks are called repeatedly each time a node splits, and different features are distributed in independent blocks, so distributed or multi-threaded computing can be performed.
[0111] 2) Cache access
[0112] After sorting the eigenvalues, we can get the gradient g by indexing i ,h i , which will lead to inconsistent memory space access, thus reducing the cache hit rate and affecting the efficiency of the algorithm. To solve this problem, Catboost allocates a separate continuous cache area for each thread to store gradient information.
[0113] 3) Out-of-core block computing
[0114] When the amount of data is too large, it cannot be loaded into the memory at the same time. Catboost divides the data into multiple blocks and stores them on the hard disk. It uses an independent thread to read the data from the disk into the memory, so that calculation and data reading can be performed simultaneously. In order to further improve the performance of disk reading data, Catboost also uses two methods: one is to compress the block and exchange the decompression cost for the disk reading cost; the other is to store the block in multiple disks, which helps to improve the disk throughput.
[0115] The objective function of the Catboost model is:
[0116]
[0117] in, is the model prediction value corresponding to the b-th sample data, A is the number of decision trees, f a (x b ) is the ath function value corresponding to the bth sample data in the function space F, x b is the bth sample data input, is the leaf node weight of the bth sample data falling into the corresponding leaf node.
[0118] Before performing capacity optimization configuration, the embodiment of the present application uses a machine learning (i.e., Catboost model) method to optimize the annual network loss of the distribution network and the voltage deviation of each access node, thereby improving the speed of subsequent capacity optimization configuration. In addition, the annual network loss and the voltage deviation of each node are calculated using a machine learning method for the distribution network, and a large amount of distribution network data is used to train the flow calculation data-driven model to increase the calculation accuracy and improve the adaptability of this embodiment on a time scale. The site selection is also carried out through distribution network flow analysis to reduce network losses and improve power supply stability.
[0119] In step 102, a two-layer optimization configuration model for a photovoltaic power station is constructed. The two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model. The upper model takes minimizing the annual operating cost of the distribution network as an optimization goal, and the lower model takes minimizing the annual network loss and node voltage deviation of the distribution network as an optimization goal.
[0120] In the embodiment of the present application, under the constraints of the equipment allowing the configuration of photovoltaic capacity constraints, distributed photovoltaic active output constraints, distribution network flow constraints, distribution network node voltage constraints, distribution network node current constraints and photovoltaic power capacity constraints, a photovoltaic power station double-layer optimization configuration model is constructed. Among them, the photovoltaic power station double-layer optimization configuration model includes an upper model and a lower model. That is:
[0121] The upper model is constructed under the constraints of the equipment's allowed photovoltaic capacity constraints and distributed photovoltaic active output constraints, with the minimum annual operating cost of the distribution network as the optimization goal, and the decision variable is the installed capacity of the photovoltaic power station.
[0122] The objective function of the upper model is:
[0123]
[0124] Among them, F is the annual operating cost of the distribution network, C in To convert to the annual fixed investment cost of the photovoltaic power station, C op is the annual operation and maintenance cost, C G is the annual power supply cost of the substation, C en is the environmental cost, r is the discount rate, y is the economic service life of the wind turbine and photovoltaic system, is the unit capacity investment cost of the photovoltaic power station, Si,PV is the rated capacity of the PV power station installed at the candidate node i, N PV is the number of nodes to be installed in the photovoltaic system, i is the node, N is the annual operating hours, j is the operating time, is the operation and maintenance cost per unit power generation of the photovoltaic system installed at the candidate node i, c g is the unit power supply cost of the substation, S G is the total installed capacity of the substation, a is the carbon emissions per unit of power generation of the upper grid, c is c is the unit price of carbon emissions.
[0125] The device allows the configuration of photovoltaic capacity constraints as follows:
[0126]
[0127] Among them, S i is the configured capacity of node i, S i,max is the maximum configurable capacity of the device at node i, S L,max is the maximum load value, S SE,max is the maximum photovoltaic penetration rate allowed by the system, i is the node, N c is the total number of nodes.
[0128] The active output constraints of distributed photovoltaics are:
[0129]
[0130] Among them, P PV,i,t is the distributed photovoltaic active power output, P PV,i,max It is the maximum value of distributed photovoltaic active output.
[0131] The lower model is constructed under the constraints of distribution network flow constraints, distribution network node voltage constraints, distribution network node current constraints and photovoltaic power capacity constraints, with the optimization goal of minimizing the annual network loss and node voltage deviation of the distribution network. The decision variables are photovoltaic output, branch power, branch current, etc.
[0132] The objective function of the lower model is:
[0133]
[0134] Among them, f1 is the first sub-objective function, f2 is the second sub-objective function, w1 is the third weight coefficient, w2 is the fourth weight coefficient, ΔU i is the voltage deviation of node i, ∑ t P loss,t is the annual network loss of the distribution network, P loss,t is the network loss of each line in the distribution network in a certain period of time t, and n is the total number of nodes in the distribution network.
[0135] The power flow constraint of the distribution network is:
[0136]
[0137] in, is the photovoltaic active power of the distribution network bus k at time t, is the active load of the distribution network bus k at time t, P kl,t is the active power from distribution network bus k to distribution network bus l at time t, P jk,t is the active power from distribution network bus j to distribution network bus k at time t, Q jk,t is the reactive power from distribution network bus j to distribution network bus k at time t, V i,t is the node voltage of node i at time t, r jk is the line resistance from distribution network bus j to distribution network bus k, I jk,t is the line current from distribution network bus j to distribution network bus k at time t, j, k and l are the distribution network bus numbers, L(k) and J(k) are the parent node and child node, and t is the time.
[0138] The voltage constraint of the distribution network node is:
[0139]
[0140] Among them, V k,t is the bus voltage of the distribution network bus k at time t, V j,k is the line voltage from distribution network bus j to distribution network bus k, x jk is the line inductance from distribution network bus j to distribution network bus k, V min is the lower limit of the distribution network voltage, V max The upper voltage limit of the distribution network.
[0141] The current constraint of the distribution network node is:
[0142]
[0143] in, It is the upper limit of the line current between distribution network bus j and distribution network bus k.
[0144] The photovoltaic power capacity constraint is:
[0145]
[0146] in, is the upper limit of the PV active power of the distribution network bus k at time t.
[0147] The embodiment of the present application constructs a two-layer optimization configuration model for a photovoltaic power station, thereby providing a judgment basis for subsequently determining the optimal configuration scheme for the installed capacity of the photovoltaic power station, and can make the subsequently obtained installed capacity of the photovoltaic power station more in line with actual needs.
[0148] In step 103, based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the double-layer optimization configuration model of the photovoltaic power station, the optimal configuration scheme of the photovoltaic power station is obtained, and the optimal configuration scheme is the installed capacity of the photovoltaic power station.
[0149] Among them, heuristic algorithms are proposed relative to optimization algorithms. The optimal algorithm for a problem finds the optimal solution for each instance of the problem. Heuristic algorithms can be defined as follows: an algorithm based on intuition or experience, which gives a feasible solution for each instance of the combinatorial optimization problem to be solved at an acceptable cost (referring to computing time and space), and the degree of deviation between the feasible solution and the optimal solution is generally unpredictable. At this stage, heuristic algorithms are mainly based on natural body algorithms, including ant colony algorithms, simulated annealing methods, neural networks, etc.
[0150] In the embodiment of the present application, the optimal configuration scheme of the installed capacity of the photovoltaic power station is calculated by using a heuristic algorithm combined with the annual network loss of the distribution network and the voltage deviation of the cut-off point output by the power flow calculation data-driven model.
[0151] The embodiments of the present application can effectively improve the reliability of the planning scheme for the installed capacity of photovoltaic power stations while accelerating the site selection and capacity optimization process, reduce system network losses and voltage fluctuations, and effectively ensure the safe and reliable operation of the distribution network.
[0152] In a possible implementation, based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the two-layer optimization configuration model of the photovoltaic power station, the optimal configuration scheme of the photovoltaic power station is obtained, which may include:
[0153] The voltage deviation of each node in the distribution network and the annual network loss are input into the double-layer optimization configuration model of the photovoltaic power station. The heuristic algorithm is used to solve the double-layer optimization configuration model of the photovoltaic power station to obtain the installed capacity of the photovoltaic power station.
[0154] Optionally, the annual network loss of the distribution network and the voltage deviation of each node output by the flow calculation data-driven model are input into the lower model of the two-layer optimization configuration model of the photovoltaic power station, and then the upper model and the lower model of the two-layer optimization configuration model of the photovoltaic power station are solved using a heuristic algorithm to output the installed capacity of the photovoltaic power station that best meets the current actual needs.
[0155] The present application provides an optimization configuration method for photovoltaic power stations participating in grid voltage regulation, by obtaining the operating data of each node in the distribution network, and inputting the operating data of each node in the distribution network into a power flow calculation data-driven model, outputting the voltage deviation and annual network loss of each node, the power flow calculation data-driven model is constructed based on a Catboost model; constructing a two-layer optimization configuration model for photovoltaic power stations, the two-layer optimization configuration model for photovoltaic power stations includes an upper model and a lower model, the upper model takes the minimization of the annual operating cost of the distribution network as an optimization goal, and the lower model takes the minimization of the annual network loss of the distribution network and the node voltage deviation as an optimization goal; based on a heuristic algorithm, the voltage deviation and annual network loss of each node in the distribution network and the two-layer optimization configuration model of the photovoltaic power station, an optimal configuration scheme of the photovoltaic power station is obtained, and the optimal configuration scheme is the installed capacity of the photovoltaic power station. This application uses a power flow calculation data-driven model constructed by the Catboost model to calculate the voltage deviation of each node in the distribution network and the annual network loss, which can increase the calculation accuracy and improve the adaptability on the time scale; by constructing a two-layer optimization configuration model for photovoltaic power stations and using the heuristic method to solve the two-layer optimization configuration model for photovoltaic power stations, the installed capacity of the photovoltaic power station is obtained, which can improve the configuration rate of the installed capacity of the photovoltaic power station.
[0156] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
[0157] The following is an embodiment of the device of the present application. For details not described in detail, please refer to the corresponding method embodiment described above.
[0158] Figure 4 The schematic diagram of the structure of the optimized configuration device for the photovoltaic power station participating in the grid voltage regulation provided by the embodiment of the present application is shown. For the convenience of explanation, only the part related to the embodiment of the present application is shown, which is described in detail as follows:
[0159] like Figure 4 As shown, the photovoltaic power station participates in the grid voltage regulation optimization configuration device 4 includes:
[0160] The data acquisition module 41 is used to acquire the operating data of each node of the distribution network, and input the operating data of each node of the distribution network into the power flow calculation data driving model, and output the voltage deviation and annual network loss of each node. The power flow calculation data driving model is constructed based on the Catboost model;
[0161] A model building module 42 is used to build a two-layer optimization configuration model for a photovoltaic power station. The two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model. The upper model takes the minimization of the annual operation cost of the distribution network as an optimization target, and the lower model takes the minimization of the annual network loss and node voltage deviation of the distribution network as an optimization target;
[0162] The scheme determination module 43 is used to obtain the optimal configuration scheme of the photovoltaic power station based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the double-layer optimization configuration model of the photovoltaic power station. The optimal configuration scheme is the installed capacity of the photovoltaic power station.
[0163] The present application provides an optimization configuration device for a photovoltaic power station to participate in grid voltage regulation. The device obtains the operation data of each node in the distribution network, inputs the operation data of each node in the distribution network into a power flow calculation data-driven model, outputs the voltage deviation of each node and the annual network loss, and the power flow calculation data-driven model is constructed based on a Catboost model; a two-layer optimization configuration model for a photovoltaic power station is constructed, and the two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model. The upper model takes the minimization of the annual operation cost of the distribution network as an optimization target, and the lower model takes the minimization of the annual network loss of the distribution network and the node voltage deviation as an optimization target; based on a heuristic algorithm, the voltage deviation of each node in the distribution network and the annual network loss, and the two-layer optimization configuration model for a photovoltaic power station, an optimal configuration scheme for the photovoltaic power station is obtained, and the optimal configuration scheme is the installed capacity of the photovoltaic power station. This application uses a power flow calculation data-driven model constructed by the Catboost model to calculate the voltage deviation of each node in the distribution network and the annual network loss, which can increase the calculation accuracy and improve the adaptability on the time scale; by constructing a two-layer optimization configuration model for photovoltaic power stations and using the heuristic method to solve the two-layer optimization configuration model for photovoltaic power stations, the installed capacity of the photovoltaic power station is obtained, which can improve the configuration rate of the installed capacity of the photovoltaic power station.
[0164] In a possible implementation, the device may further include a node selection module, which may be used to:
[0165] Using the distribution network operation data, the static voltage stability index and loss sensitivity of each node in the distribution network are calculated;
[0166] Use the entropy weight method to assign weights to each node;
[0167] Based on the weighted method, the static voltage stability index, loss sensitivity and weight of each node, the comprehensive evaluation index of each node is calculated;
[0168] Select the node corresponding to the minimum comprehensive evaluation index from all comprehensive evaluation indexes as the target node;
[0169] Accordingly, the solution determination module can be used to:
[0170] The optimal configuration scheme of the photovoltaic power station connected to the target node of the distribution network is obtained.
[0171] In a possible implementation, the distribution network operation data may include active load, reactive load, branch reactance, branch resistance, node voltage, and active power loss.
[0172] In a possible implementation, the node selection module may also be used to:
[0173] The static grid stability index and loss sensitivity of each node are calculated by the first formula, which is:
[0174]
[0175] Among them, SVSI is the static voltage stability index, LSFs is the loss sensitivity, P b,load is the active load of node b, Q b,load is the reactive load of node b, P ab,L is the active power loss of branch ab, X ab is the branch reactance of branch ab, R ab is the branch resistance of branch ab, U a is the node voltage of node a, U b is the node voltage of node b.
[0176] In a possible implementation, the node selection module may also be used to:
[0177] The comprehensive evaluation index of each node is calculated by the second formula, which is:
[0178]
[0179] Among them, CEISS is a comprehensive evaluation index, η1 is the first weight coefficient, and η2 is the second weight coefficient.
[0180] In one possible implementation, the objective function of the upper model is:
[0181]
[0182] Among them, F is the annual operating cost of the distribution network, C in To convert to the annual fixed investment cost of the photovoltaic power station, C op is the annual operation and maintenance cost, C G is the annual power supply cost of the substation, S en is the environmental cost, r is the discount rate, y is the economic service life of the wind turbine and photovoltaic system, is the unit capacity investment cost of the photovoltaic power station, S i,PV is the rated capacity of the PV power station installed at the candidate node i, N PV is the number of nodes to be installed in the photovoltaic system, i is the node, N is the annual operating hours, j is the operating time, is the operation and maintenance cost per unit power generation of the photovoltaic system installed at the candidate node i, c g is the unit power supply cost of the substation, S G is the total installed capacity of the substation, a is the carbon emissions per unit of power generation of the upper grid, c is c is the unit price of carbon emissions.
[0183] In one possible implementation, the objective function of the lower model is:
[0184]
[0185] Among them, f1 is the first sub-objective function, f2 is the second sub-objective function, w1 is the third weight coefficient, w2 is the fourth weight coefficient, ΔU i is the voltage deviation of node i, ∑ t P loss,t is the annual network loss of the distribution network, P loss,t is the network loss of each line in the distribution network in a certain period of time t, and n is the total number of nodes in the distribution network.
[0186] In a possible implementation, the constraints of the two-layer optimization configuration model of the photovoltaic power station may include constraints on the photovoltaic capacity allowed to be configured by the equipment, constraints on the active output of distributed photovoltaics, constraints on the power flow of the distribution network, constraints on the voltage of the distribution network nodes, constraints on the current of the distribution network nodes, and constraints on the capacity of the photovoltaic power source;
[0187] The device allows the configuration of photovoltaic capacity constraints as follows:
[0188]
[0189] Among them, S i is the configured capacity of node i, S i,max is the maximum configurable capacity of the device at node i, S L,max is the maximum load value, S SE,max is the maximum photovoltaic penetration rate allowed by the system, i is the node, N c is the total number of nodes;
[0190] The active output constraints of distributed photovoltaics are:
[0191]
[0192] Among them, P PV,i,t is the distributed photovoltaic active power output, P PV,i,max The maximum active output of distributed photovoltaics;
[0193] The power flow constraint of the distribution network is:
[0194]
[0195] in, is the photovoltaic active power of the distribution network bus k at time t, is the active load of the distribution network bus k at time t, P kl,t is the active power from distribution network bus k to distribution network bus l at time t, P jk,t is the active power from distribution network bus j to distribution network bus k at time t, Q jk,t is the reactive power from distribution network bus j to distribution network bus k at time t, V i,t is the node voltage of node i at time t, r jk is the line resistance from distribution network bus j to distribution network bus k, I jk,t is the line current from distribution network bus j to distribution network bus k at time t, j, k and l are the numbers of distribution network bus, L(k) and J(k) are the parent node and child node, t is the time;
[0196] The voltage constraint of the distribution network node is:
[0197]
[0198] Among them, V k,t is the bus voltage of the distribution network bus k at time t, V j,k is the line voltage from distribution network bus j to distribution network bus k, x jk is the line inductance from distribution network bus j to distribution network bus k, V min is the lower limit of the distribution network voltage, V max is the upper limit of the distribution network voltage;
[0199] The current constraint of the distribution network node is:
[0200]
[0201] in, is the upper limit of the line current between distribution network bus j and distribution network bus k;
[0202] The photovoltaic power capacity constraint is:
[0203]
[0204] in, is the upper limit of the PV active power of the distribution network bus k at time t.
[0205] In a possible implementation, the solution determination module may be used to:
[0206] The voltage deviation of each node in the distribution network and the annual network loss are input into the double-layer optimization configuration model of the photovoltaic power station. The heuristic algorithm is used to solve the double-layer optimization configuration model of the photovoltaic power station to obtain the installed capacity of the photovoltaic power station.
[0207] In a possible implementation, the device may further include a preprocessing module, which may be used to:
[0208] Preprocess the operating data of each node in the distribution network for state estimation;
[0209] Accordingly, the data acquisition module can be used to:
[0210] The preprocessed operating data of each node in the distribution network is input into the power flow calculation data-driven model, and the voltage deviation of each node and the annual network loss are output.
[0211] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0212] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0213] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned embodiment of the optimization configuration method for each photovoltaic power station to participate in grid voltage regulation can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium.
[0214] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation, characterized in that: include: Obtaining the operating data of each node of the distribution network, and inputting the operating data of each node of the distribution network into a power flow calculation data-driven model, and outputting the voltage deviation and annual network loss of each node, wherein the power flow calculation data-driven model is constructed based on the Catboost model; Constructing a two-layer optimization configuration model for a photovoltaic power station, the two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model, the upper model takes minimizing the annual operating cost of the distribution network as an optimization goal, and the lower model takes minimizing the annual network loss and node voltage deviation of the distribution network as an optimization goal; Based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the double-layer optimization configuration model of the photovoltaic power station, the optimal configuration scheme of the photovoltaic power station is obtained, and the optimal configuration scheme is the installed capacity of the photovoltaic power station.
2. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 1, characterized in that: After obtaining the operation data of each node in the distribution network, the method further includes: Calculating the static voltage stability index and loss sensitivity of each node of the distribution network using the distribution network operation data; Use the entropy weight method to assign weights to each node; Based on the weighted method, the static voltage stability index, loss sensitivity and weight of each node, the comprehensive evaluation index of each node is calculated; Selecting a node corresponding to a minimum comprehensive evaluation index from all comprehensive evaluation indexes as the target node; Accordingly, obtaining the optimal configuration scheme of the photovoltaic power station includes: An optimal configuration scheme of the photovoltaic power station to which the target node of the distribution network is connected is obtained.
3. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 2, characterized in that: The distribution network operation data includes active load, reactive load, branch reactance, branch resistance, node voltage and active loss. The static voltage stability index and loss sensitivity of each node of the distribution network are calculated by using the distribution network operation data, including: The static grid stability index and loss sensitivity of each node are calculated by the first formula, and the first formula is: Wherein, SVSI is the static voltage stability index, LSFs is the loss sensitivity, P b,load is the active load of node b, Q b,load is the reactive load of node b, P ab,L is the active power loss of branch ab, X ab is the branch reactance of branch ab, R ab is the branch resistance of branch ab, U a is the node voltage of node a, U b is the node voltage of node b.
4. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 2, characterized in that: The weighted method, the static voltage stability index, loss sensitivity and weight of each node are used to calculate the comprehensive evaluation index of each node, including: The comprehensive evaluation index of each node is calculated by the second formula, and the second formula is: Among them, CEISS is the comprehensive evaluation index, η1 is the first weight coefficient, and η2 is the second weight coefficient.
5. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 1, characterized in that: The objective function of the upper model is: Among them, F is the annual operating cost of the distribution network, C in To convert to the annual fixed investment cost of the photovoltaic power station, C op is the annual operation and maintenance cost, C G is the annual power supply cost of the substation, C en is the environmental cost, r is the discount rate, y is the economic service life of the wind turbine and photovoltaic system, is the unit capacity investment cost of the photovoltaic power station, S i,PV is the rated capacity of the PV power station installed at the candidate node i, N PV is the number of nodes to be installed in the photovoltaic system, i is the node, N is the annual operating hours, j is the operating time, is the operation and maintenance cost per unit power generation of the photovoltaic system installed at the candidate node i, c g is the unit power supply cost of the substation, S G is the total installed capacity of the substation, a is the carbon emissions per unit of power generation of the upper grid, c is c is the unit price of carbon emissions.
6. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 1, characterized in that: The objective function of the lower model is: Among them, f1 is the first sub-objective function, f2 is the second sub-objective function, w1 is the third weight coefficient, w2 is the fourth weight coefficient, ΔU i is the voltage deviation of node i, ∑ t P loss,t is the annual network loss of the distribution network, P loss,t is the network loss of each line in the distribution network in a certain period of time t, and n is the total number of nodes in the distribution network.
7. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 1, characterized in that: The constraints of the photovoltaic power station double-layer optimization configuration model include equipment allowed photovoltaic capacity constraints, distributed photovoltaic active output constraints, distribution network flow constraints, distribution network node voltage constraints, distribution network node current constraints and photovoltaic power capacity constraints; The device allows to configure the PV capacity constraints as: Among them, S i is the configured capacity of node i, S i,max is the maximum configurable capacity of the device at node i, S L,max is the maximum load value, S SE,max is the maximum photovoltaic penetration rate allowed by the system, i is the node, N c is the total number of nodes; The distributed photovoltaic active output constraint is: Among them, P PV,i,t is the distributed photovoltaic active power output, P PV,i,max The maximum active output of distributed photovoltaics; The power flow constraint of the distribution network is: in, is the photovoltaic active power of the distribution network bus k at time t, is the active load of the distribution network bus k at time t, P kl,t is the active power from distribution network bus k to distribution network bus l at time t, P jk,t is the active power from distribution network bus j to distribution network bus k at time t, Q jk,t is the reactive power from distribution network bus j to distribution network bus k at time t, V i,t is the node voltage of node i at time t, r jk is the line resistance from distribution network bus j to distribution network bus k, I jk,t is the line current from distribution network bus j to distribution network bus k at time t, j, k and l are the numbers of distribution network bus, L(k) and J(k) are the parent node and child node, t is the time; The distribution network node voltage constraint is: Among them, V k,t is the bus voltage of the distribution network bus k at time t, V j,k is the line voltage from distribution network bus j to distribution network bus k, x jk is the line inductance from distribution network bus j to distribution network bus k, V min is the lower limit of the distribution network voltage, V max is the upper limit of the distribution network voltage; The distribution network node current constraint is: in, is the upper limit of the line current between distribution network bus j and distribution network bus k; The photovoltaic power capacity constraint is: in, is the upper limit of the PV active power of the distribution network bus k at time t.
8. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 1, characterized in that: The optimal configuration scheme of the photovoltaic power station is obtained based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the double-layer optimization configuration model of the photovoltaic power station, including: The voltage deviation of each node of the distribution network and the annual network loss are input into the double-layer optimization configuration model of the photovoltaic power station, and the double-layer optimization configuration model of the photovoltaic power station is solved by using the heuristic algorithm to obtain the installed capacity of the photovoltaic power station.
9. The method for optimizing the configuration of photovoltaic power stations participating in grid voltage regulation according to claim 1, characterized in that: Before inputting the operation data of each node of the distribution network into the power flow calculation data driven model and outputting the voltage deviation of each node and the annual network loss, the method further includes: Preprocessing the operating data of each node of the distribution network for state estimation; Accordingly, the operation data of each node of the distribution network is input into the power flow calculation data driven model, and the voltage deviation and annual network loss of each node are output, including: The pre-processed operation data of each node of the distribution network is input into the power flow calculation data driven model, and the voltage deviation of each node and the annual network loss are output.
10. An optimization configuration device for photovoltaic power station participating in grid voltage regulation, characterized in that: include: A data acquisition module is used to acquire the operating data of each node of the distribution network, and input the operating data of each node of the distribution network into a power flow calculation data-driven model, and output the voltage deviation and annual network loss of each node. The power flow calculation data-driven model is constructed based on the Catboost model; A model building module is used to build a two-layer optimization configuration model for a photovoltaic power station, wherein the two-layer optimization configuration model for a photovoltaic power station includes an upper model and a lower model, wherein the upper model takes the minimization of the annual operating cost of the distribution network as an optimization target, and the lower model takes the minimization of the annual network loss and node voltage deviation of the distribution network as an optimization target; The scheme determination module is used to obtain the optimal configuration scheme of the photovoltaic power station based on the heuristic algorithm, the voltage deviation of each node of the distribution network and the annual network loss and the double-layer optimization configuration model of the photovoltaic power station, and the optimal configuration scheme is the installed capacity of the photovoltaic power station.
Citation Information
Cited By
Parameter optimization method and system for flexible interconnection phase shift device
CN120073744A
Chain type micro-grid double-layer optimization configuration method considering line N-1 fault mutual aid
CN122418719A
A double-layer optimization configuration method of chain microgrid considering line N-1 fault mutual aid
CN122418719B
