CN109829560B - A distribution network renewable energy generation cluster access planning method - Google Patents

Abstract

The invention relates to a method for access planning of renewable energy power generation clusters in a distribution network. An upper-level planning model and a lower-level dispatching model are adopted. The upper-level planning model aims at maximizing the income of investors in renewable energy power generation; the objective function of the lower-level dispatching model includes power The balance index, the adjustment cost of the power distribution company and the active power reduction of the renewable energy generation, the adjustment measures include the action of the tie line switch, the tap action of the on-load tap changer, the active power reduction and reactive power compensation of the renewable energy generation, Aiming at the timing correlation between load and renewable energy resources, the C‑Vine Copula model is used for modeling, combined with the Latin hypercube sampling method to generate typical planning scenarios considering the correlation between load and resources.

Description

A distribution network renewable energy generation cluster access planning method

Technical Field

The invention relates to a distribution network renewable energy power generation cluster access planning method considering load and resource time sequence correlation.

Background Art

The access of renewable energy generation to the distribution network has a positive effect on energy conservation and carbon emission reduction. However, when the power generation of renewable energy accounts for 20%-30% of the total power generation of various power sources, the intermittent and random nature of the output of renewable energy generation will lead to problems such as system overvoltage and power reverse transmission. Therefore, in order to improve the distribution network's acceptance capacity for renewable energy generation, it is necessary to consider the impact of renewable energy generation on system operation during the planning stage.

For the renewable energy power generation planning problem of distribution network, most of the existing research is to select the site and determine the capacity of renewable energy power generation under a single substation. This kind of method assumes that the distribution networks under different substations are independent of each other, does not consider the connection and influence between the operation of multiple substations in the system, and ignores the complementary support capacity of power between multiple substations. Since this method uses the same objective function and constraints when planning for each substation, the output characteristics of multiple substations may be similar after the renewable energy power generation planning. If multiple substations in the same area have power reverse transmission at the same time and transmit it to the upper level power grid, it will affect the normal operation of the high-voltage distribution network. Active distribution network can use advanced automation, communication and power electronics and other new technologies to realize active management of renewable energy power generation and other equipment connected to the distribution network. The distribution network is widely equipped with tie switches and section switches. The active distribution network can realize dynamic reconstruction of the network by controlling the opening and closing of switches, which is conducive to reducing network losses and balancing loads. Therefore, when studying the access and absorption of renewable energy power generation in the distribution network, it is necessary to consider the connection between the output power of multiple substations after the access of renewable energy power generation and the impact of network reconstruction on the change of power supply support capacity between multiple substations, carry out cluster access planning of renewable energy power generation in multiple substations of the distribution network, and determine the cluster access capacity of renewable energy power generation under each substation.

In the planning stage, the access of intermittent renewable energy generation to the distribution network increases the uncertainty of distribution network operation and the difficulty of scenario selection in the distribution network planning process. Using a small number of representative scenarios to accurately characterize the random characteristics of loads and resources can effectively reduce the amount of calculation and improve the accuracy of planning results. The renewable energy generation planning methods that consider uncertainty mainly include planning based on multi-scenario technology, planning based on opportunity constraint theory, and planning based on fuzzy theory. Among them, the scenario analysis method enumerates the possible values of uncertain factors according to rules and combines them into a series of planning scenarios. Each scenario has a corresponding probability, thereby converting the uncertainty problem into a deterministic problem, reducing the difficulty of modeling and solving. In addition to uncertainty, there is a certain correlation between loads and resources. However, the current method can only represent the linear correlation between variables, and the modeling of variables with nonlinear correlation is not accurate enough. The Copula function does not require variables to have the same marginal distribution, and can describe the nonlinearity, asymmetry, tail correlation and other characteristics between variables. The PairCopula method is a branch of the Copula function, which can characterize the correlation between multidimensional variables, has a flexible structure, and can better capture the correlation between any two variables. However, this method is currently rarely used in power system planning.

In summary, the defects and shortcomings of the prior art methods can be summarized as follows:

(1) Regarding the planning of renewable energy generation in distribution networks, existing studies have mostly focused on the site selection and capacity determination of renewable energy generation under a single substation. This type of method assumes that the distribution networks under different substations are independent of each other and does not consider the connection and impact between the operations of multiple substations in the system.

(2) In the process of renewable energy power generation planning, the impact of various regulation measures in the active distribution network, especially network reconstruction, on the changes in the power supply support capacity between multiple substations has not been fully considered.

(3) Existing technical methods can only describe the linear correlation between loads and resources, and are not accurate enough in modeling variables with asymmetric and nonlinear correlations.

Summary of the invention

In view of the above problems, the present invention proposes a two-layer planning method for renewable energy generation cluster access that considers the time correlation of load and renewable energy resources. The technical solution is as follows:

A distribution network renewable energy generation cluster access planning method adopts an upper-level planning model and a lower-level scheduling model. The upper-level planning model aims to maximize the benefits of renewable energy generation investors, and the planning scenario used considers the nonlinear and asymmetric correlation between loads and resources. The objective function of the lower-level scheduling model includes a power balance index, the regulation cost of the distribution company, and the active power reduction of renewable energy generation. The regulation measures include the action of the tie-line switch, the tap action of the on-load tap-changing transformer, the active power reduction and reactive power compensation of renewable energy generation. In view of the temporal correlation between loads and renewable energy resources, a C-Vine Copula model is used for modeling, and a typical planning scenario considering the correlation between loads and resources is generated in combination with a Latin hypercube sampling method.

Among them, the lower-level scheduling model includes:

(1) Objective function 1: operating costs of distribution companies

f ₁ = C _loss + C _reg

Where, C _loss and C _reg are the network loss cost and regulation cost of the distribution company respectively. The regulation cost includes the tap adjustment cost of the on-load tap-changing transformer and the tie switch operation cost. c _l is the power price. _Pij and _Qij are the active and reactive power flowing from the upstream node i to the node j. _Vi is the voltage value of the node i. i→j means that the node i and the node j are connected. _Rij is the line resistance value between the node i and the node j. N is the set of nodes in the distribution network. c _tap is the cost of each tap adjustment. and is the tap position at time t and time t-1, c _swi is the cost of a single action of the tie line switch, and is the switch status of the tie line at time t and time t-1;

(2) Objective Function 2: Block Power Balance Index

Taking a high voltage/medium voltage substation and the network connected to it as a block, active power balance index and reactive power balance index are proposed, which are defined as follows:

In the formula, N _block is the total number of clusters, P _block,i is the active power demand or active power output of the i-th block, and Q _block,i is the reactive power demand or reactive power output of the i-th cluster;

The smaller the active power balance index or reactive power balance index is, the smaller the active power or reactive power exchanged between the block and the outside world is, and the more balanced the active power or reactive power within the block is. By operating the tie switch, the power balance index is optimized to reduce the power flowing through the upper substation caused by power imbalance and improve the power balance of the system.

The power balance index objective function is:

f ₂ =ω ₁ f _{P_Bal} +ω ₂ f _{Q_Bal}

In the formula, ω ₁ and ω ₂ are the weights of the active balance index and the reactive balance index, which can be determined according to the importance of the index and need to satisfy ω ₁ +ω ₂ =1;

(3) Objective function 3: Reduction of renewable energy

Take the reduction of renewable energy as one of the lower-level dispatch targets:

In the formula, and are the active power reduction of the i-th _PV or i-th _WTG respectively;

Normalize the three objective functions:

In the formula, is the normalized objective function, _fimin is the minimum value of the i-th objective function, and _fimax is the maximum value of the i-th objective function;

The overall objective function of the lower-level scheduling model is:

Where λ ₁ , λ ₂ , and λ ₃ are the objective functions after planning. The weight coefficient can be determined comprehensively according to the importance of each target in the scheduling process and the actual operation status, and must satisfy λ ₁ +λ ₂ +λ ₃ =1;

The constraints of the lower-level scheduling model include:

(1) Power flow equation constraints

Among them, P _j =P _Lj -P _total,PV,j -P _total,WTG,j +P _cut,PV,j +P _cut,WTG,j , Q _j =Q _Lj -Q _PV,j -Q _{WTG, j} ;

Where R _ij and X _ij represent the resistance and reactance of the line between node i and node j, respectively; P _j and Q _j represent the active and reactive power of the net load at node j, _PLj and Q _Lj represent the active and reactive power of the load at node j, P _total,PV,j and P _cut,PV,j represent the active power and active power reduction of the photovoltaic power plant at node j, respectively; P _total,WTG,j and P _cut,WTG,j represent the active power and active power reduction of the wind turbine at node j, respectively;

(2) System security constraints

In the formula, and are the upper and lower limits of the voltage at node j respectively;

(3) Distributed PV operation constraints

Q _{PV, j} = (P _{total, PV, j} - P _{cut, PV, j} ) tanθ

In the formula, θ = cos ^-1 PF _min represents the minimum power factor PF _min limit of the photovoltaic output power;

(4) Fan operation constraints

Q _WTG,j = (P _total,WTG,j -P _cut,WTG,j )tanθ

In the formula, θ = cos ^-1 PF _min represents the minimum power factor PF _min limit of the fan output power;

(5) Constraints on on-load tap-changing transformers

U _i = k _ij,t U _j

k _ij,t =1+K _ij,t Δk _ij

In the formula, _Ui and _Uj are the voltages on the high-voltage side and low-voltage side of the transformer, respectively, and _Kij and are the lower and upper limits of the transformer tap position, respectively; _Kij,t is the transformer tap position at time t; _Δkij is the transformer adjacent tap position adjustment ratio; and _Kij,t is the transformer high and low voltage side voltage ratio at time t;

(6) Tie-line switch constraints

The tie line switch state should ensure that the load on the tie line is continuously powered and does not operate in a closed loop. Therefore, for a tie line with N tie switches, only one tie switch should be disconnected.

In the formula, is the opening and closing state of the tie line switch on line ij at time t, if closed, it is 1, if open, it is 0; O is the set of branches along the line that constitute a ring network;

(7) 220kV substation power constraints

In order to ensure the safe operation of the system and avoid reverse transmission of power to the transmission network, it is required that the 220kV substation power is not reversed:

0≤P _sub,220kV ≤P _rated,220kV

g) Power constraints of distribution substations below 220 kV

Distribution companies are authorized to curtail the active output of renewable energy generation to limit the reverse power to less than or equal to 60% of the rated capacity of the substation:

-0.6×P _{rated,＜220kV} ≤P _sub ≤P _{rated,＜220kV} .

The upper-level planning model includes:

Determine the cluster access capacity of photovoltaic and wind turbines under each 35kV substation:

maxF _upper =max(C _cell -C _inv -C _main )

The electricity sales revenue of renewable energy power generation users is calculated based on the electricity sales of photovoltaic and wind turbines:

In the formula, r is the discount rate, Ny, N _PV and N _WTG are the planning period, the number of photovoltaic power plants and the number of wind turbines, respectively, c _sell,PV and c _sell,WTG are the on-grid electricity prices of photovoltaic power plants and wind turbines, respectively. and are the actual grid-connected power of the ith _PV or ith _WTG in the yth year and the sth scenario;

The construction cost is calculated based on the installed capacity of wind turbines and photovoltaics:

Where c _ins,PV and c _ins,WTG are the construction costs per unit capacity of photovoltaic and wind turbines, respectively. and The installed capacity of photovoltaic and wind turbines respectively

Based on the installed capacity of photovoltaic power and the total power generation of wind turbines, the operation and maintenance cost of renewable energy power generation is calculated:

In the formula, c _om,PV is the annual operation and maintenance cost of photovoltaic unit installed capacity, c _om,WTG is the operation and maintenance cost of wind turbine unit power generation, is the actual power generation of the i-th _WTG wind turbine in the y-th year and the s-th scenario;

Renewable energy generation planning is subject to geographical factors of the installation site and total investment cost factors, and needs to meet installation capacity restrictions:

and

In the formula, and They represent the upper limit of the installation capacity of the i-th _PV photovoltaic installation point or the i-th _WTG wind turbine installation point respectively.

The steps for generating a planning scenario are as follows:

(1) Read the historical data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load _Xori = (x1 _,ori , _x2,ori , _x3,ori , x4 _,ori , _x5,ori , _x6,ori ), and standardize each type of data to obtain:

Wherein, x _i,ori ,i＝1,...,6 represents the original data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load, x _i,ori,max ,i＝1,...,6 represents the peak value of the corresponding resource or the equipment installation capacity of the corresponding type of load, and x _i ,i＝1,...,6 is the original normalized data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load obtained;

(2) Using the cumulative distribution function u _i =F _i ( _xi ), u _i ∈[0,1], i=1,...,6, the original data X is converted into uniformly distributed data U=(u ₁ ,u ₂ ,u ₃ ,u ₄ ,u ₅ ,u ₆ ) on [0,1];

(3) For uniformly distributed data U = (u ₁ ,u ₂ ,u ₃ ,u ₄ ,u ₅ ,u ₆ ), use the maximum likelihood function and Anderson-Darling method to perform parameter estimation and goodness-of-fit test, and find the parameters and structure of the C-Vine Copula function corresponding to the data U;

(4) Use the Latin hypercube sampling method to generate independent uniformly distributed variables (w ₁ ,w ₂ ,w ₃ ,w ₄ ,w ₅ ,w ₆ ) on [0,1]. According to the obtained C-Vine Copula structure, the following conditional distribution formula can be obtained:

Where Z = (z ₁ , z ₂ , z ₃ , z ₄ , z ₅ , z ₆ ) is the scene corresponding to the six types of data in the uniform domain;

(5) Use the inverse function of the marginal distribution function of the original data u _i ∈[0,1], i = 1,...,6 to obtain the typical normalized scenarios of the six types of data in the actual domain;

(6) The typical normalized scenarios of the four types of loads are multiplied by the installed capacity of the four types of load equipment corresponding to each substation in the planning area and the sum is obtained to obtain the typical scenario of the total load under each substation. The typical scenario of the resources is obtained by multiplying the normalized scenario of the resources by the resource peak value.

The invention is directed to a distribution network renewable energy generation cluster access planning method considering load and resource timing correlation, and has the following advantages over the prior art:

(1) The present invention carries out renewable energy power generation planning for a distribution network including multiple substations, taking into account the power complementary support capacity between multiple substations and the impact of network reconstruction on planning. The method determines the total installed capacity of photovoltaic and wind turbines under each substation, and guides the determination of the specific installation location and capacity of photovoltaic and wind turbines under each substation in subsequent planning work.

(2) In order to improve the absorption capacity of renewable energy power generation and improve the power balance of the system, the present invention proposes a power balance index. The active power balance index can be optimized through network reconstruction, which is conducive to reducing the impact of reverse power on the upper power grid and increasing the installed capacity of renewable energy power generation.

(3) The present invention models the correlation between loads and resources based on the C-Vine Copula method, which can accurately characterize the nonlinear and asymmetric correlation between variables. Then, the Latin hypercube sampling method is used to generate a smaller number of typical scenarios for renewable energy generation planning, which improves the calculation speed while ensuring the calculation accuracy.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a flow chart of the two-level planning model

Figure 2 is a schematic diagram of the C-Vine Copula structure

Figure 3 is the electrical wiring diagram of the planned embodiment

FIG. 4 is a C-Vine Copula result between load and resources in the embodiment

DETAILED DESCRIPTION

The present invention is described below in conjunction with the accompanying drawings and attached tables.

The planning method for renewable energy generation cluster access includes an upper-level planning model and a lower-level scheduling model. The upper-level planning model aims to maximize the benefits of renewable energy generation investors and determines the cluster access capacity of photovoltaic and wind turbines under each 35kV substation:

maxF _upper =max(C _cell -C _inv -C _main )

The electricity sales revenue of renewable energy power generation users is calculated based on the electricity sales of photovoltaic and wind turbines:

In the formula, r is the discount rate, Ny, N _PV and N _WTG are the planning period, the number of photovoltaic power plants and the number of wind turbines, respectively, c _sell,PV and c _sell,WTG are the on-grid electricity prices of photovoltaic power plants and wind turbines, respectively. and are the actual grid-connected power of the i-th _PV or i-th _WTG in the y-th scenario.

The construction cost is calculated based on the installed capacity of wind turbines and photovoltaics:

Where c _ins,PV and c _ins,WTG are the construction costs per unit capacity of photovoltaic and wind turbines, respectively. and The installed capacity of photovoltaic and wind turbines respectively

Based on the installed capacity of photovoltaic power and the total power generation of wind turbines, the operation and maintenance cost of renewable energy power generation is calculated:

In the formula, c _om,PV is the annual operation and maintenance cost of photovoltaic unit installed capacity, c _om,WTG is the operation and maintenance cost of wind turbine unit power generation, is the actual power generation of the i-th _WTG wind turbine in the s-th scenario in the y-th year.

Renewable energy generation planning is subject to geographical factors of the installation site, total investment costs and other factors, and needs to meet installation capacity restrictions:

and

In the formula, and They represent the upper limit of the installation capacity of the i-th _PV photovoltaic installation point or the i-th _WTG wind turbine installation point respectively.

The objective function of the lower dispatch model includes the power balance index, the regulation cost of the distribution company and the active power reduction of renewable energy generation. The regulation measures include the action of the tie line switch, the tap action of the on-load tap-changing transformer, the active power reduction and reactive power compensation of renewable energy generation.

(1) Objective function 1: operating costs of distribution companies

f ₁ = C _loss + C _reg

In the formula, C _loss and C _reg are the network loss cost and regulation cost of the distribution company respectively. The regulation cost includes the tap adjustment cost of the on-load tap-changing transformer and the tie switch operation cost. _{c l} is the active power price, P _ij and _Qij are the active and reactive power flowing from the upstream node i to the node j, _Vi is the voltage value of the node i, i→j means that the node i and the node j are connected, _Rij is the line resistance value between the node i and the node j, and N is the set of distribution network nodes. c _tap is the cost of each tap adjustment, and is the tap position at time t and time t-1, c _swi is the cost of a single action of the tie line switch, and is the switch status of the tie line at time t and time t-1.

(2) Objective Function 2: Block Power Balance Index

The present invention regards a high-voltage/medium-voltage substation and the network connected to it as a block. The high-penetration distributed renewable energy generation connected to the distribution network will generate power backflow, increase system network losses, and reduce the service life of the substation. In order to improve the block's ability to absorb renewable energy generation, reduce power backflow, and improve the power complementarity between distribution networks within the cluster, the present invention proposes active balance index and reactive balance index, which are defined as follows:

In the formula, N _block is the total number of clusters, P _block,i is the active power demand or active power output of the i-th block, and Q _block,i is the reactive power demand or reactive power output of the i-th cluster.

The smaller the active balance index or reactive balance index, the smaller the active power or reactive power exchanged between the block and the outside world, and the more balanced the active or reactive power within the block. Through the operation of the tie switch, the power balance index can be optimized, reducing the power flowing through the upper substation due to power imbalance and improving the power balance of the system.

The power balance index objective function is:

f ₂ =ω ₁ f _{P_Bal} +ω ₂ f _{Q_Bal}

In the formula, the ratio of ω ₁ to ω ₂ is the weight of the active balance index and the reactive balance index, which can be determined according to the importance of the index and needs to satisfy ω ₁ +ω ₂ =1.

(3) Objective function 3: Reduction of renewable energy

In order to improve the absorption level of distributed power sources, reduce the amount of wind and solar power abandonment, and improve the utilization efficiency of renewable energy, the present invention takes the reduction amount of renewable energy as one of the lower-level scheduling targets.

In the formula, and are the active power reduction of the i-th _PV or i-th _WTG respectively.

Since the three objective functions have different dimensions, they need to be normalized.

In the formula, is the normalized objective function, _fimin is the minimum value of the i-th objective function, and _fimax is the maximum value of the i-th objective function.

The objective function of the lower-level scheduling model is:

Where λ ₁ , λ ₂ , and λ ₃ are the objective functions after planning. The weight coefficient can be determined comprehensively according to the importance of each target in the scheduling process and the actual operation status, and needs to satisfy λ ₁ +λ ₂ +λ ₃ =1.

The objective functions of the lower-level scheduling model include:

(1) Power flow equation constraints

Among them, P _j =P _Lj -P _total,PV,j -P _total,WTG,j +P _cut,PV,j +P _cut,WTG,j , Q _j =Q _Lj -Q _PV,j -Q _{WTG, j} .

where R _ij and _Xij represent the resistance and reactance of the line between node i and node j, respectively; P _j and Q _j are the active and reactive power of the net load at node j; _PLj and Q _Lj are the active and reactive power of the load at node j; P _total,PV,j and P _cut,PV,j are the active power and active power reduction of the photovoltaic power plant at node j, respectively; P _total,WTG,j and P _cut,WTG,j are the active power and active power reduction of the wind turbine at node j, respectively.

(2) System security constraints

In the formula, and are the upper and lower limits of the voltage at node j respectively.

(3) Distributed PV operation constraints

Q _{PV, j} = (P _{total, PV, j} - P _{cut, PV, j} ) tanθ

Wherein, θ=cos ^-1 PF _min represents the minimum power factor PF _min limit of the photovoltaic output power.

(4) Fan operation constraints

Q _WTG,j = (P _total,WTG,j -P _cut,WTG,j )tanθ

Wherein, θ=cos ^-1 PF _min represents the minimum power factor PF _min limit of the wind turbine output power.

(5) Constraints on on-load tap-changing transformers

U _i = k _ij,t U _j

k _ij,t =1+K _ij,t Δk _ij

In the formula, _Ui and _Uj are the voltages on the high-voltage side and low-voltage side of the transformer, respectively, and _Kij and are the lower and upper limits of the transformer tap position respectively, _Kij,t is the tap position of the transformer at time t, _Δkij is the adjustment ratio of adjacent tap positions of the transformer, and _Kij,t is the voltage ratio of the high and low voltage sides of the transformer at time t.

(6) Tie-line switch constraints

The tie line switch state should ensure that the load on the tie line is continuously powered and does not operate in a closed loop. Therefore, for a tie line with N tie switches, only one tie switch should be disconnected for operation.

In the formula, is the opening and closing state of the tie line switch on line ij at time t, if closed, it is 1, if open, it is 0; O is the set of branches along the line that constitute a ring network.

(7) 220kV substation power constraints

In order to ensure the safe operation of the system and avoid reverse transmission of power to the transmission network, it is required that the 220kV substation power is not reversed:

0≤P _sub,220kV ≤P _rated,220kV

g) Power constraints of distribution substations below 220 kV

Due to the access of high penetration rate of distributed renewable energy in distribution network, reverse power will lead to increased network loss and line overcurrent. Therefore, distribution companies have the right to reduce the active output of renewable energy power generation to limit the reverse power to less than or equal to 60% of the rated capacity of substation.

-0.6×P _{rated,＜220kV} ≤P _sub ≤P _{rated,＜220kV}

The upper-level planning model is solved using a genetic algorithm, and the obtained wind turbine and photovoltaic installation capacity is passed to the lower-level scheduling model. The lower-level scheduling model is a mixed-integer nonlinear programming problem. Through linearization and cone relaxation, the original NP-hard nonconvex nonlinear problem is transformed into a mixed-integer second-order cone programming model. In order to ensure the accuracy of cone relaxation, cut constraints are added for solution until the cone relaxation error is reduced to a predetermined range, and finally the scheduling result is passed to the upper-level planning model. The upper and lower models are alternately iterated and solved until the calculation termination condition is triggered, and the REG planning result is output. The algorithm flow of the two-level programming model is shown in Figure 1.

The present invention divides the historical data obtained from the planning area into wind speed, light, industrial load, agricultural load, commercial load and residential load data, and normalizes them according to the resource peaks corresponding to the above six types of data and the installed capacity of various types of load equipment, and then uses the C-Vine Copula method to model the correlation of the above six types of normalized data, and obtains a C-Vine Copula structure that considers nonlinear and asymmetric correlations between variables. Then, the Latent Hypercube Sampling Method is used to generate independent uniformly distributed samples, and the obtained C-Vine Copula structure is combined to generate a typical normalized scenario that considers multivariate correlations. Finally, the normalized scenarios of the four types of loads obtained are multiplied and summed with the installed capacity of the four types of load equipment corresponding to each substation, that is, the typical scenario of the total load under each substation can be obtained, and the resource normalized scenario is multiplied by the resource peak to obtain the typical scenario of the resource.

Copula function is a powerful tool for studying the correlation between random variables. It connects the joint distribution of multivariate random variables with each univariate marginal distribution, and can describe the nonlinearity, asymmetry, tail correlation and other characteristics between variables. According to Sklar's theorem, let F be an n-dimensional joint probability distribution function with marginal distribution functions F ₁ (x ₁ ),...,F _n (x _n ), then there exists an n-dimensional Copula function C such that have

F(x ₁ ,x ₂ ,···,x _n )=C(F ₁ (x ₁ ),F ₂ (x ₂ ),···,F _n (x _n ))

If F ₁ (x ₁ ),...,F _n (x _n ) are all continuous distribution functions, then C is the unique Copula function corresponding to F.

Let _ui = F _i ( _xi ), _ui∈ [0,1], i = 1,..., n obey uniform distribution, then

C(u ₁ ,...,u _n )=P(U ₁ ≤u ₁ ,...,U _n ≤u _n )=F(F ₁ ^-1 (u ₁ ),...,F _n ^{- 1} (u _n ))

Where F _i ^-1 (u _i ) is the inverse function of the marginal distribution function.

The probability density function of the Copula function is defined as follows:

In the formula, _fi ( _xi ) is the probability density function, f( _x1 , _x2 ,..., _xn ) is the joint probability density function, and c( _F1 ( _x1 ), _F2 ( _x2 ),..., _Fn ( _xn )) is the Copula probability density function.

The Pair Copula structure decomposes the multivariate Copula function into multiple pairs of binary Copula functions. It has a flexible structure and can better capture the dependency relationship between any two variables.

The joint probability density function of random variables X = (X ₁ ,...,X _n ) can be decomposed into

f(x ₁ ,x ₂ ,···,x _n )=f _n (x _n )·f(x _n-1 |x _n )·f(x _n-2 |x _n-1 ,x _n ). ..f(x ₁ |x ₂ ,...,x _n )

The above formula can be decomposed into the product of the appropriate Pair Copula function and the conditional probability density function:

Where v is a d-dimensional vector, _vj is any element in v, and v _-j represents the vector v after removing _vj . Therefore, the multivariate probability density function can be represented by multiple Pair Copula functions.

The pair copula structure involves the marginal conditional distribution function F(x|v) of the variables:

Where _Ci,j|k is the binary Copula distribution function. When v only includes a single variable,

There are two main types of Pair Copula structures: D-vine and Canonical Vine (C-Vine). The present invention uses the C-Vine form. The structure of C-Vine can be expressed as

In the formula, j represents the number of layers of C-Vine, i represents the edge of each layer, and in the jth layer, there is always a node connected to n-j edges. The n-dimensional C-Vine Copula structure is shown in Figure 2.

The maximum likelihood method is used to estimate the parameters of the Pair Copula. Assuming there are n-dimensional variables and each variable has T observations, each variable can be expressed as follows:

_xi ＝(xi _,1 ,...,xi _,T )

For a binary Copula density function c _{j,j+i|1,...,j-1} , the parameter Θ of C-Vine is obtained by finding the logarithmic maximum likelihood function:

Because there are many types of binary Copula functions that can be used to fit the correlation between the original data, it is very important to select the most appropriate Copula function in the C-Vine Copula function structure. Therefore, a goodness of fit test is needed to test whether the selected Copula function type can accurately characterize the correlation between the variables and select the most appropriate Copula function. The Anderson Darling method is used for goodness of fit test.

Let X and Y represent two random variables, their respective marginal distribution functions are U = _FX (x) = P(X≤x) and V = _FX (y) = P(Y≤y), and their joint distribution function is FX _,Y (x,y) = P(X≤x,Y≤y). Assuming that both F _X and F _Y are continuous functions, there exists a unique Copula function C:[0,1] ² →[0,1]:

F _{X ,Y (} x,y)＝C(F

When U＝u, the conditional distribution function between U and V is:

Where D ₁ represents the partial derivative of C(u,v) with respect to u.

The random variables Z ₁ = U = _FX (x) and Z ₂ = C(V|U) = C(F _Y (y) | _FX (x)) are independent and uniformly distributed on [0,1]. Therefore, the random variable S(X,Y) = [Φ ^-1 ( _FX (X))] ² + [Φ ^-1 C(F _Y (Y) | _FX (X))] ² is χ ² distributed with 2 degrees of freedom. If (X ₁ ,Y ₁ ), ..., (X _n ,Y _n ) is a random sample from the population (X,Y), then S(X ₁ ,Y ₁ ), ..., S(X _n ,Y _n ) is a random sample from a χ ² distribution with 2 degrees of freedom. Therefore, the test hypothesis is

H ₀ :(X,Y) exists a Copula function C(u,v)

Among them, the marginal distribution functions F _X and F _Y are known. By calculating S(X ₁ ,Y ₁ ),...,S(X _n ,Y _n ), the test hypothesis H ₀ can be transformed into the test auxiliary hypothesis:

yes distributed

if If H 0 is established, then H 0 is established. If H 0 is rejected, then H ₀ is established. Then reject H ₀ .

Because the Anderson Darling test has relatively good properties for a large number of different situations, the present invention uses the Anderson Darling method to The test statistic of the Anderson Darling method is:

Where S _j =S(X _j ,Y _j ), j=1,...,n, and S ₍₁₎ ≤…≤S _(n) . _{F 0} =χ ² distribution with 2 degrees of freedom.

However, in practical applications, the marginal distribution functions of F _X and F _Y are generally unknown, so the empirical distribution function is used instead of the marginal distribution function:

and

use Instead of S(X _j ,Y _j ):

in addition,

It should be noted that if the marginal distribution function of the variable is unknown, using the empirical distribution function instead will affect the critical value of the goodness of fit test. The present invention uses the Bootstrap method to determine the critical value with a confidence level of 1-α, and the steps are as follows:

(1) Estimate the parameter θ of the Copula function C(u,v;θ) from the initial observations (x ₁ ,y ₁ ),...,(x _n , _yn )

(2) From the Copula function Generate n independent observations

(3) By i＝1,...,n Estimate the estimated value of the parameter θ of the Copula function C(u,v;θ) Depend on and calculate Calculate the value AD ^* of the Anderson Darling test statistic using the above calculated values;

(4) Repeat steps (2) and (3) N times to obtain the value of the test statistic AD ^*(1) , ..., AD ^*(N) . The value of the test statistic corresponding to the 1-α quantile is the required critical value.

If the value of the test statistic calculated from the initial observations (x ₁ ,y ₁ ), ...,(x _n ,y _n ) is greater than the obtained critical value, the null hypothesis is rejected and the Copula function is considered unsuitable to describe the dependency structure of the observations; otherwise, the null hypothesis is accepted.

The specific steps of the scenario generation method based on the C-Vine Copula method considering the temporal correlation between loads and resources are as follows:

(1) Read the historical data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load _Xori = ( _x1,ori , _x2,ori , _x3,ori , x4 _,ori , _x5,ori , _x6,ori ), and standardize each type of data to obtain:

In the formula, x _i,ori ,i＝1,...,6 represents the original data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load, x _i,ori,max ,i＝1,...,6 represents the peak value of the corresponding resource or the equipment installation capacity of the corresponding type of load, and x _i ,i＝1,...,6 is the original standardized data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load.

(2) Using the cumulative distribution function _ui = _Fi ( _xi ), _ui∈ [0,1], i = 1, ..., 6, the original data X is converted into uniformly distributed data U = ( _u1 , _u2 , _u3 , _u4 , _u5 , _u6 ) on [0,1].

(3) For the uniformly distributed data U = (u ₁ ,u ₂ ,u ₃ ,u ₄ ,u ₅ ,u ₆ ), the maximum likelihood function and Anderson Darling method are used to perform parameter estimation and goodness of fit test, and the parameters and structure of the C-Vine Copula function corresponding to the data U are obtained.

(4) Use the Latin hypercube sampling method to generate independent uniformly distributed variables (w ₁ ,w ₂ ,w ₃ ,w ₄ ,w ₅ ,w ₆ ) on [0,1]. Based on the obtained C-Vine Copula structure, the following conditional distribution formula can be obtained:

In the formula, Z = (z ₁ , z ₂ , z ₃ , z ₄ , z ₅ , z ₆ ) is the scene corresponding to the six types of data in the uniform domain.

(5) Using the inverse function of the marginal distribution function of the original data x _i =F _i ^-1 (z _i ), _ui ∈ [0,1], i = 1, ..., 6, we obtain the typical normalized scenarios of the six types of data in the actual domain.

(6) The typical normalized scenarios of the four types of loads are multiplied by the installed capacity of the four types of load equipment corresponding to each substation in the planning area and the sum is obtained to obtain the typical scenario of the total load under each substation. The typical scenario of the resources is obtained by multiplying the normalized scenario of the resources by the resource peak value.

The present invention describes embodiments based on FIGS. 3 to 4 and Tables 1 to 5.

A part of the medium and high voltage distribution network in a certain place in China is selected as an embodiment, and Figure 3 is an electrical wiring diagram of the embodiment. The embodiment includes 8 substations, including 1 220kV/110kV substation, 2 110kV/35kV substations, and 5 35kV/10kV substations. In addition, the embodiment also includes 8 lines, 2 of which are 110kV lines and 6 are 35kV lines, and the 35kV lines are all equipped with disconnect switches.

Without considering the specific grid parameters under the 35kV substation, the load under each 35kV substation is equivalent to its low-voltage side. Obtain the historical data of light intensity, wind speed, and load for 8760 hours in the planning area in the past year, and divide the load into industrial load, agricultural load, commercial load, and residential load, and count the peak value of resources and the equipment installation capacity corresponding to each type of load historical data. The capacity of each substation and the installation capacity of various types of load equipment connected to it are shown in Appendix 1. The present invention carries out cluster access planning of photovoltaics and wind turbines for each 35kV substation, with a planning period of 15 years and an annual load growth rate set at 3%. The capacity of a single wind turbine is 2MW. The planning economic parameters and scheduling parameters are shown in Appendix 2.

The C-Vine Copula structure among multiple variables obtained by the scenario generation method proposed in the present invention is shown in Figure 3. The typical planning scenario is generated by combining the Latin hypercube sampling method, and the average load and peak load under each substation are obtained, see Appendix 3.

The planning results obtained using the two-level planning method proposed in the present invention are shown in Appendix 4, and other results are shown in Appendix 5.

Table 1 shows the transformer capacity of each 35kV substation in the planning area and the installed capacity of various load equipment

Table 2 is the parameters of the planned implementation example.

Table 3 shows the average load and peak load of each 35kV substation

Table 4 shows the planning results of wind turbines and photovoltaics for each substation

Table 5 shows other results of planning calculations.

Claims (1)

1. A distribution network renewable energy generation cluster access planning method, using an upper-level planning model and a lower-level scheduling model. The upper-level planning model aims to maximize the benefits of renewable energy generation investors; the objective function of the lower-level scheduling model includes the operating cost of the distribution company, the block power balance index and the amount of renewable energy reduction, and the adjustment measures include the action of the tie-line switch, the tap action of the on-load tap-changing transformer, the active reduction and reactive power compensation of renewable energy generation; in view of the time series correlation between loads and renewable energy resources, the C-Vine Copula model is used for modeling, and the Latin hypercube sampling method is combined to generate a typical planning scenario that considers the correlation between loads and resources, wherein, The lower-level scheduling model includes: (1) Objective function 1: operating costs of distribution companies f ₁ = C _loss + C _reg Where, C _loss and C _reg are the network loss cost and regulation cost of the distribution company respectively. The regulation cost includes the tap adjustment cost of the on-load tap-changing transformer and the tie switch operation cost. c _l is the power price. _Pij and _Qij are the active and reactive power flowing from the upstream node i to the node j. _Vi is the voltage value of the node i. i→j means that the node i and the node j are connected. _Rij is the line resistance value between the node i and the node j. N is the set of distribution network nodes. c _tap is the cost of each tap adjustment. and is the tap position at time t and time t-1, c _swi is the cost of a single action of the tie line switch, and is the tie line switch status at time t and time t-1; (2) Objective Function 2: Block Power Balance Index Taking a high voltage/medium voltage substation and the network connected to it as a block, active power balance index and reactive power balance index are proposed, which are defined as follows: In the formula, N _block is the total number of clusters, P _block,i is the active power demand or active power output of the i-th block, and Q _block,i is the reactive power demand or reactive power output of the i-th cluster; The smaller the active power balance index or reactive power balance index is, the smaller the active power or reactive power exchanged between the block and the outside world is, and the more balanced the active power or reactive power within the block is. By operating the tie switch, the power balance index is optimized to reduce the power flowing through the upper substation caused by power imbalance and improve the power balance of the system. The power balance index objective function is: f ₂ =ω ₁ f _{P_Bal} +ω ₂ f _{Q_Bal} In the formula, ω ₁ and ω ₂ are the weights of the active balance index and the reactive balance index, which are determined according to the importance of the index and need to satisfy ω ₁ +ω ₂ =1; (3) Objective function 3: Reduction of renewable energy Take the reduction of renewable energy as one of the lower-level dispatch targets: In the formula, and are the active power reduction of the i-th _PV or i-th _WTG respectively; Normalize the three objective functions: Where, _fi ^* is the normalized objective function, _fi ^* ∈[0,1], _fimin is the minimum value of the i-th objective function, and _fimax is the maximum value of the i-th objective function; The overall objective function of the lower-level scheduling model is: Where λ ₁ , λ ₂ , and λ ₃ are the planned objective functions f ₁ ^* , The weight coefficient can be determined comprehensively according to the importance of each target in the scheduling process and the actual operation status, and must satisfy λ ₁ +λ ₂ +λ ₃ =1; The constraints of the lower-level scheduling model include: (1) Power flow equation constraints Among them, P _j =P _Lj -P _total,PV,j -P _total,WTG,j +P _cut,PV,j +P _cut,WTG,j , Q _j =Q _Lj -Q _PV,j -Q _{WTG, j} ; Where R _ij and X _ij represent the resistance and reactance of the line between node i and node j, respectively; P _j and Q _j represent the active and reactive power of the net load at node j, _PLj and Q _Lj represent the active and reactive power of the load at node j, P _total,PV,j and P _cut,PV,j represent the active power and active power reduction of the photovoltaic power plant at node j, respectively; P _total,WTG,j and P _cut,WTG,j represent the active power and active power reduction of the wind turbine at node j, respectively; (2) System security constraints In the formula, and are the upper and lower limits of the voltage at node j respectively; (3) Distributed PV operation constraints Q _{PV, j} = (P _{total, PV, j} - P _{cut, PV, j} ) tanθ In the formula, θ = cos ^-1 PF _min represents the minimum power factor PF _min limit of the photovoltaic output power; (4) Fan operation constraints Q _WTG,j = (P _total,WTG,j -P _cut,WTG,j )tanθ In the formula, θ = cos ^-1 PF _min represents the minimum power factor PF _min limit of the fan output power; (5) Constraints on on-load tap-changing transformers U _i = k _ij,t U _j k _ij,t =1+K _ij,t Δk _ij Where _Ui and _Uj are the voltages on the high-voltage side and low-voltage side of the transformer, respectively, and _Kij and are the lower and upper limits of the transformer tap position, respectively; _Kij,t is the transformer tap position at time t; _Δkij is the transformer adjacent tap position adjustment ratio; and _Kij,t is the transformer high and low voltage side voltage ratio at time t; (6) Tie-line switch constraints The tie line switch state should ensure that the load on the tie line is continuously powered and does not operate in a closed loop. Therefore, for a tie line with N tie switches, only one tie switch should be disconnected for operation. In the formula, is the opening and closing state of the tie line switch on line ij at time t, if closed, it is 1, if open, it is 0; O is the set of branches along the line that constitute a ring network; (7) 220kV substation power constraints In order to ensure the safe operation of the system and avoid reverse transmission of power to the transmission network, it is required that the 220kV substation power is not reversed: 0≤P _sub,220kV ≤P _rated,220kV g) Power constraints of distribution substations below 220 kV Distribution companies are authorized to curtail the active output of renewable energy generation to limit the reverse power to less than or equal to 60% of the rated capacity of the substation: -0.6×P _{rated,＜220kV} ≤P _sub ≤P _{rated,＜220kV} ; The upper-level planning model includes: Determine the cluster access capacity of photovoltaic and wind turbines under each 35kV substation: maxF _upper =max(C _cell -C _inv -C _main ) The electricity sales revenue of renewable energy power generation users is calculated based on the electricity sales of photovoltaic and wind turbines: In the formula, r is the discount rate, Ny, N _PV and N _WTG are the planning period, the number of photovoltaic power plants and the number of wind turbines, respectively, c _sell,PV and c _sell,WTG are the on-grid electricity prices of photovoltaic power plants and wind turbines, respectively. and are the actual grid-connected power of the ith _PV or ith _WTG in the yth year and the sth scenario; The construction cost is calculated based on the installed capacity of wind turbines and photovoltaics: Where c _ins,PV and c _ins,WTG are the construction costs per unit capacity of photovoltaic and wind turbines, respectively. and The installed capacity of photovoltaic and wind turbines respectively Based on the installed capacity of photovoltaic power and the total power generation of wind turbines, the operation and maintenance cost of renewable energy power generation is calculated: In the formula, c _om,PV is the annual operation and maintenance cost of photovoltaic unit installed capacity, c _om,WTG is the operation and maintenance cost of wind turbine unit power generation, is the actual power generation of the i-th _WTG wind turbine in the y-th year and the s-th scenario; Renewable energy generation planning is subject to geographical factors of the installation site and total investment cost factors, and needs to meet installation capacity restrictions: In the formula, and They represent the upper limit of the installed capacity of the ith _PV photovoltaic installation point or the ith _WTG wind turbine installation point respectively; The steps for generating a planning scenario are as follows: (1) Read the historical data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load _Xori = ( _x1,ori , _x2,ori , _x3,ori , x4 _,ori , _x5,ori , _x6,ori ), and standardize each type of data to obtain: Wherein, x _i,ori ,i＝1,...,6 represents the original data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load, x _i,ori,max ,i＝1,...,6 represents the peak value of the corresponding resource or the equipment installation capacity of the corresponding type of load, and x _i ,i＝1,...,6 is the original normalized data of wind resources, light resources, industrial load, agricultural load, commercial load and residential load obtained; (2) Using the cumulative distribution function u _i =F _i ( _xi ), u _i ∈[0,1], i=1,...,6, the original data X is converted into uniformly distributed data U=(u ₁ ,u ₂ ,u ₃ ,u ₄ ,u ₅ ,u ₆ ) on [0,1]; (3) For uniformly distributed data U = (u ₁ ,u ₂ ,u ₃ ,u ₄ ,u ₅ ,u ₆ ), use the maximum likelihood function and Anderson-Darling method to perform parameter estimation and goodness-of-fit test, and find the parameters and structure of the C-Vine Copula function corresponding to the data U; (4) Use the Latin hypercube sampling method to generate independent uniformly distributed variables (w ₁ ,w ₂ ,w ₃ ,w ₄ ,w ₅ ,w ₆ ) on [0, and obtain the following conditional distribution formula based on the obtained C-Vine Copula structure: Where Z = (z ₁ , z ₂ , z ₃ , z ₄ , z ₅ , z ₆ ) is the scene corresponding to the six types of data in the uniform domain; (5) Using the inverse function of the marginal distribution function of the original data x _i =F _i ^-1 (z _i ), _ui ∈ [0,1], i = 1, ..., 6, we can obtain the typical normalized scenarios of the six types of data in the actual domain; (6) Multiply the obtained typical normalized scenarios of the four types of loads by the installed capacity of the corresponding four types of load equipment under each substation in the planning area and sum them up to obtain the typical scenario of the total load under each substation. Multiply the normalized scenario of resources by the resource peak value to obtain the typical scenario of resources.