Active power distribution network planning method for improving utilization rate and absorption capability of distributed energy resources

Abstract

The invention provides an active power distribution network planning method for improving the utilization rate and the absorption capability of distributed energy resources. The method comprises the following steps: S1, establishing an active power distribution ADN planning model based on multi-source collaboration; and S2, solving the model based on an improved crowd search algorithm of SFLA-GA and a CLMCS probabilistic load flow calculation method considering uncertain factors to realize active power distribution network planning. According to the method, a crowd search algorithm is improved based on SFLA-GA, the constructed three-layer planning model containing generalized power supply constant volume and site selection, controllable load contract quantity and overall operation of the power distribution system is solved, and the active power distribution network planning method is obtained, so that the utilization rate and the absorption capacity of distributed energy resources are improved.

Description

Technical Field

[0001] The present invention relates to the field of distribution network planning, and in particular to an active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources. Background Art

[0002] With the development of smart grid and active distribution network (ADN), the integration of distributed energy resources (DER) such as distributed generation (DG), energy storage system (ESS) and controllable load in distribution network is increasing. However, the integration of DER brings new challenges to the planning and operation of distribution network.

[0003] In order to solve the above problems, the present invention proposes an active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources; this method starts from the three levels of environment, operation and economy, and establishes a three-layer planning model including generalized power source capacity selection, controllable load contract quantity and overall operation of the distribution system to improve the utilization rate and absorption capacity of distributed energy resources. Summary of the invention

[0004] The purpose of the present invention is to propose an active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources; this method starts from the three levels of environment, operation and economy, and solves the constructed three-layer planning model including generalized power source capacity selection, controllable load contract quantity and overall operation of distribution system through an improved crowd search algorithm based on SFLA-GA, and obtains an active distribution network planning method to improve the utilization rate and absorption capacity of distributed energy resources.

[0005] To achieve the above object, the technical solution of the present invention is: an active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources, comprising the following steps:

[0006] S1. Establish an active ADN planning model based on multi-source collaboration;

[0007] S2. The improved crowd search algorithm based on SFLA-GA and the CLMCS probabilistic power flow calculation method taking into account uncertain factors are used to optimize the model and realize active distribution network planning.

[0008] Preferably, the ADN planning model based on multi-source collaboration is a three-layer planning model, including upper-level planning, middle-level planning and lower-level planning; the upper-level planning aims to maximize the utilization rate of clean energy, and determines the generalized power source capacity selection and the optimal configuration of the section switches; the middle-level planning aims to maximize user satisfaction, and determines the number and amount of contracted contracts for controllable loads; the lower-level planning aims to minimize network losses, and dispatches and controls the output of various energy sources and the opening and closing of section switches.

[0009] Preferably, the upper-level planning determines the installation location and capacity of the distributed generation device DG and the energy storage device ESS from the perspective of maximizing the utilization rate of clean energy, and its objective function is:

[0010]

[0011] Among them, η multi-s is the clean energy utilization rate, P load is the total load of the distribution network system, n is the total number of nodes in the distribution network system; η DG , η ESS are the access capacity of DG and ESS and P load The ratio of DG,a , P ESS,b are the active powers of the DG and ESS connected to nodes a and b respectively; c and d are the access position sets of the selected DG, ESS and controllable load respectively.

[0012] Preferably, the middle-level planning objective function is as follows:

[0013]

[0014] Where E is user satisfaction, N is the number of controllable loads signed; μ 1,n' , μ 2,n' are the influencing factors of the shiftable load and interruptible load at the node where the n'th controllable load is located; T a,n' , T b,n' They represent the expected scheduling time of the shiftable load and interruptible load at the node where the n'th controllable load is located, T n',all Then it is the total power consumption time of the n'th controllable load.

[0015] Preferably, the lower-level planning objective function is specifically as follows:

[0016]

[0017] Among them, P LOSS is the network loss, P LINE represents the line loss of the distribution network; YT represents the number of time periods in a year; α ESS That is, the state change dimension of the energy storage device, which is 0 or 1; ΔP ESS.i' is the loss of the i'th ESS during state transition, nESS is the total number of ESS; ξ ch.i' , dis.i' , are the charging and discharging efficiency and power of the i'th ESS in the h period respectively.

[0018] Preferably, the specific steps of S2 are as follows:

[0019] S21, initializing various parameters of the improved crowd search algorithm based on SFLA-GA and the CLMCS probability power flow calculation method, and inputting various original parameters of the computing power network system;

[0020] S22. Initialize the optimization candidate solutions represented by each searcher in the crowd. The information carried by each searcher in different search dimensions represents a decision variable, which is expressed as a matrix X:

[0021]

[0022] In the formula, and (i=1,2,…N w ,j=1,2,…N p ,k=1,2,…N e ,l=1,2,…N c q=1,2,…N s , m=1,2,…M) respectively represent the installed capacity of wind turbines and photovoltaic power generation equipment at the i-th, j-th, and k-th RDG access nodes, the contracted amount of controllable load at the l-th node, and the access position of the section switch is q; N w 、N p 、N e 、N c 、N s are the alternative numbers of wind turbines, photovoltaic generators, energy storage devices, controllable load users, and section switches respectively;

[0023] S23, bring the information stored by the searcher into the CLMCS power flow calculation method, calculate and save the probability power flow results of each candidate planning scheme one by one; at the same time, check whether it meets the limit standards of various constraint conditions, and obtain the fitness value of the objective function; if it does not meet the standards, use the crowd search algorithm to re-update the searcher's search information;

[0024] S24, update and store the optimal position and the optimal fitness value of the current searcher, and determine whether the end condition of the optimization search is met. If the condition is met, the loop is terminated and the optimal planning scheme is output. Otherwise, execute step S25;

[0025] S25. Use the improved crowd search algorithm based on SFLA-GA to update the position of the searcher, return to step S23, and re-apply the CLMCS power flow calculation method to perform power flow calculation analysis on the updated crowd individuals until the termination condition of the algorithm is met in step S24.

[0026] Preferably, the optimization candidate solution represented by each searcher in the crowd is initialized, specifically:

[0027] In each dimension, the search direction f randomly generated by the searcher mk0 With search step length β mk0 Multiplying together gives the initial value of the searcher within the range of feasible solutions:

[0028] x mk0 =β mk0 ·f mk0 (twenty two)

[0029] Where: f mk0 It represents the search direction of the mth search individual in the kth dimension; when f mk0 When the value is 1, it means that the search individual moves forward; when f mk0 When it is -1, it means that the search individual moves backwards; when its value is 0, it means that the search individual stays where it is.

[0030] Preferably, the improved crowd search algorithm based on SFLA-GA is used to update the location of the searcher, specifically:

[0031] According to the strategy of simulating frog grouping and periodic mixing in SFLA, the population P is first divided into S groups in ascending order, with T individuals in each group; the L individuals with the lowest ranking in the group and the remaining TL individuals use different position update methods, and the populations of different groups are mixed after the searcher's position is updated;

[0032] The position update method for the last L individuals in the group is: regard the top two individuals in the group as elite parents, and obtain new L individuals through genetic crossover mutation of the elite parents:

[0033]

[0034] Where: represents the individual with rank t in the sth group; r represents an arbitrary constant in the range [0,1];

[0035] The position update method of the remaining TL individuals in the group is: use the crowd search algorithm to determine the search step and search direction of the searcher, and update the searcher's position according to the search step and search direction.

[0036] Preferably, the crowd search algorithm is used to determine the search step length and search direction of the searcher, and the position of the searcher is updated according to the search step length and search direction, specifically:

[0037] Determine the search step size through uncertain reasoning:

[0038]

[0039] Where: α mkis the search step length of the mth search individual in the kth dimension, δ mk 、u mk represent the Gauss membership function parameters and the membership of the k-th dimension objective function m respectively; the Gauss membership function is a fuzzy variable used to construct the intrinsic relationship between the perception and behavior of the searcher. mk It is used to simulate the randomness of the searcher's search behavior and can be expressed by formula (24). Then, the search step length δ is deduced by uncertainty reasoning. mk :

[0041] u mk =rand(u m ,1)(k=1,2,...,D) (24)

[0042] δ mk =ω·|x min -x max |(25)

[0043] ω=(iter max -iter) / iter max (26)

[0044] Where: function rand(u m , 1) represents u m A random real number in the range from 1 to 1, x max and x min They represent the locations of the maximum and minimum values ​​of the objective function in the search population, and ω represents the inertia weight, which is the value of the current iteration number iter and the maximum optimization number iter. max jointly determined;

[0045] Determination of search direction; the direction of crowd search is determined by the self-interest direction Altruistic Direction With pre-movement direction The geometric mean of the three is obtained by randomly weighting them:

[0046]

[0047]

[0048]

[0049]

[0050] In formula (27) to formula (29), and Respectively The best location in represents the position of the mth search individual in the kth dimension in the t'th update, and They represent the best historical position of the mth search individual in the region of the kth dimension and in the population; sign() is the sign function; All represent any constant between 0 and 1; ω represents the inertia weight;

[0051] Through equation (31)-(32), the position of the searcher in the group is updated and the fitness value of the searcher is recalculated:

[0052]

[0053] x mk (t'+1)=x mk (t')+Δx mk (t'+1) (32).

[0054] Preferably, the planning model comprehensively considers multiple costs, including distributed energy construction and operation costs, switch switching costs, network loss costs, power purchase costs, reliability costs and environmental costs, to obtain the total cost of the planning, wherein the mathematical model of the annual cost of the planning is:

[0055] minC total =C build +C OR.DG +C OR.ESS +C B +C E +C LOSS +C R (11)

[0056] In the formula, C total is the total cost, C build is the construction cost of distributed generation and energy storage devices, C OR.DG is the DG operating cost, C OR.ESS is the efficiency of the energy storage device, C B is the electricity purchase cost, C E is the environmental cost, C LOSS is the network loss cost, C R The cost of reliability.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] The present invention solves the constructed three-layer planning model including generalized power source sizing and site selection, controllable load contract quantity and overall operation of the distribution system by improving the crowd search algorithm based on SFLA-GA, and obtains the active distribution network planning method to improve the utilization rate and absorption capacity of distributed energy resources, and realize the optimization of ADN scientifically and effectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is the main feeder system diagram;

[0060] Figure 2 Provide a three-layer planning model for ADN;

[0061] Figure 3 Search algorithm flow chart for the crowd;

[0062] Figure 4 Optimization flow chart for improving crowd search algorithm;

[0063] Figure 5 Optimization flow chart for the ADN three-layer planning model;

[0064] Figure 6 Power distribution system for IEEE-33 nodes;

[0065] Figure 7 Comparison of the number of iterations between the algorithm of the present invention and the original crowd search algorithm. DETAILED DESCRIPTION

[0066] The following is combined with Figure 1-7 , the technical solution of the present invention is specifically described.

[0067] The present invention focuses on the three aspects of environment, operation and economy, and establishes an ADN planning model based on multi-source collaboration from the three levels of capacity selection of generalized power sources, contract quantity of controllable loads, and overall operation of the distribution system.

[0068] The upper-level planning aims to maximize the utilization rate of clean energy, select the capacity and site for generalized power sources, and optimize the configuration of sectional switches to improve the reliability and safety of the distribution network; the middle-level planning aims to maximize user satisfaction and determine the number and amount of contracted controllable loads; the lower-level planning divides the whole year into 8760 time periods, takes the minimum network loss as the goal, and dispatches and controls the output of various energy sources and the opening and closing of sectional switches, and comprehensively considers the construction and operation costs of distributed energy, switch switching costs, network loss costs, power purchase costs, reliability costs, environmental costs, etc., to obtain the total cost of the planning. An improved crowd search algorithm based on SFLA-GA is used to perform relevant simulation calculations using the IEEE33-node distribution system to optimize the ADN scientifically and effectively.

[0069] The present invention proposes an active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources, comprising the following steps:

[0070] S1. Establish an active ADN planning model based on multi-source collaboration;

[0071] S2. The improved crowd search algorithm based on SFLA-GA and the CLMCS probabilistic power flow calculation method taking into account uncertain factors are used to optimize the model and realize active distribution network planning.

[0072] In this embodiment, the ADN planning model based on multi-source collaboration is a three-layer planning model, including upper-level planning, middle-level planning and lower-level planning; the upper-level planning aims to maximize the utilization rate of clean energy, and determines the generalized power supply capacity site selection and the optimal configuration of the section switches; the middle-level planning aims to maximize user satisfaction, and determines the number and amount of contracted controllable loads; the lower-level planning aims to minimize network losses, and dispatches and controls the output of various energy sources and the opening and closing of section switches.

[0073] The upper-level planning determines the installation location and capacity of the distributed generation device DG and the energy storage device ESS from the perspective of maximizing the utilization rate of clean energy. Its objective function is:

[0074]

[0075] Among them, η multi-s is the clean energy utilization rate, P load is the total load of the distribution network system, n is the total number of nodes in the distribution network system; η DG , η ESS are the access capacity of DG and ESS and P load The ratio of DG,a , P ESS,b are the active powers of the DG and ESS connected to nodes a and b respectively; c and d are the access position sets of the selected DG, ESS and controllable load respectively.

[0076] From this, the construction cost of distributed generation and energy storage devices converted to annual cost can be obtained:

[0077]

[0078] Among them, C DG,i , C ESS,i are the unit capacity construction costs of DG and ESS converted to annual terms; r is the annual interest rate; n DG 、n ESS are the number of DG and ESS invested and constructed respectively; Y DG , Y ESS are the service life of DG and ESS respectively.

[0079] like Figure 1 As shown in the figure, it is assumed that the main feeder is divided into n sections by n-1 section switches, L i , P i , l i 、p iThey represent the feeder length and load of the main line or branch line of the i-th section, and the average failure rate and average fault repair time of the main line and branch line are ρ, ε, and ρ, respectively. 1 , ε 1 , the switching operation time of the section switch and the tie switch is set to t 1 ,t 2 , then the power shortage of the ring network system is

[0080]

[0081] Compared with the main feeder without adding section switches, the change value of power shortage in the distribution network can be expressed as

[0082]

[0083] It can be seen that the addition of section switches can greatly reduce the power shortage cost of the active distribution network. The total construction cost can be expressed as follows:

[0084] C S =B S +M S +S S (5)

[0085] Among them, B S 、M S , S S They respectively represent the construction, operation and maintenance, and scrapping costs of the sectional switch.

[0086] With the continuous advancement of electricity marketization, demand-side response has also been gradually added to the optimized operation of ADN. However, the power consumption habits of controllable loads participating in the operation and control of the power grid will inevitably be affected by changes in the operation status of the power grid, and may even have to bear power outage losses caused by forced interruption of loads. Therefore, the present invention refers to the historical data of the power grid load, sets the influencing factors of the shiftable load and the interruptible load, and evaluates the user expectation value for the contracted load, with the goal of maximizing the user expectation value, and determines the contract number and contract amount of the controllable load. The middle-level planning objective function is as follows:

[0087]

[0088] Where E is user satisfaction, N is the number of controllable loads signed; μ 1,n' , μ 2,n' are the influencing factors of the shiftable load and interruptible load at the node where the n'th controllable load is located; T a,n' , T b,n' They represent the expected scheduling time of the shiftable load and interruptible load at the node where the n'th controllable load is located, T n',all Then it is the total power consumption time of the n'th controllable load.

[0089] The lower-level planning model determines the hourly output of distributed energy and controllable load according to the timing characteristics of load and distributed clean energy generation, with the goal of minimizing network losses, and controls the opening and closing of the optimized section switches according to the real-time operation status of the ADN, thereby achieving optimal economy while ensuring the safe and reliable operation of the ADN.

[0090] During the operation of the distribution network, the total network loss must be minimized. The lower-level planning objective function is as follows:

[0091]

[0092] Among them, P LOSS is the network loss, P LINE represents the line loss of the distribution network; YT represents the number of time periods in a year; α ESS That is, the state change dimension of the energy storage device, which is 0 or 1; ΔP ESS.i' is the loss of the i'th ESS during state transition, nESS is the total number of ESS; ξ ch.i' , dis.i' , are the charging and discharging efficiency and power of the i'th ESS in the h period respectively.

[0093] After determining the hourly output of distributed energy and the interruption capacity of controllable loads, and controlling the opening and closing of the sectional switches, a series of economic indicators such as the operating cost of distributed power sources, the benefits of energy storage devices, the cost of electricity purchase, environmental costs, network loss costs, and reliability costs can be obtained, as follows:

[0094] (1) The mathematical expression of DG operating cost is:

[0095]

[0096] Among them, τ DG is the ratio of the actual output of DG to the rated output in the h period, C M,DG , C PSE They are the annual operation and maintenance cost per unit capacity of DG and the government environmental protection subsidy.

[0097] (2) The mathematical expression of the energy storage device benefit is:

[0098]

[0099] Among them, C M.ESS The operating and maintenance costs of the energy storage device; They represent the charging and discharging electricity prices of the energy storage device in a period of h respectively. It should be noted that, in a certain period of time, at least one of the charging and discharging powers is 0.

[0100] (3) The mathematical expression of electricity purchase cost is:

[0101]

[0102] Among them, P load,h represents the total load of the active distribution network in the h period, C POW , C COMP.CL They respectively represent the unit electricity price and the incentive unit price of controllable load.

[0103] (4) Environmental cost C E The NO X , CO 2 、SO 2 It is calculated based on the environmental value standards of gas emissions.

[0104] Table 1 Waste gas emission data of power generation technology and its environmental value standards

[0105]

[0106] (5) Network loss cost C LOSS It can be obtained from formula (7) and the unit electricity price C POW Multiply them together to get .

[0107] (6) Reliability cost C R It consists of two parts: the cost of power shortage and the cost of construction, operation and maintenance, scrapping and switching of section switches.

[0108] The mathematical model of the annual cost of the plan can be obtained from this

[0109] minC total =C build +C OR.DG +C OR.ESS +C B +C E +C LOSS +C R (11)

[0110] (1) Node power balance constraints

[0111]

[0112] Where N represents the number of nodes in the distribution network; P i , Q i Respectively represent the values ​​of active and reactive power injected into node i; e i 、f i Represent the real and imaginary parts of the voltage at node i respectively; G ij , B ijThey represent the conductance and susceptance of the circuit connecting node i and node j respectively.

[0113] (2) Node voltage constraints

[0114] U imin ≤U i ≤U imax (13)

[0115] Where U imax , U imin They represent the upper and lower limits of the voltage allowed at node i respectively.

[0116] (3) Branch current constraints

[0117] 0≤I l ≤I l,max (14)

[0118] In the formula, I l Indicates the current amplitude flowing through branch l, I l,max It indicates the maximum limit that satisfies the thermal stability condition of branch 1.

[0119] (4) DER access capacity upper limit constraints

[0120]

[0121] Where α is the maximum absorption rate of ADN for various DERs.

[0122] (5) Power constraints on the line

[0123] S i ≤S i.max (16)

[0124] In the formula, S i represents the apparent power flowing through line i; S i.max is the maximum value of the apparent power that can flow through the i-th line.

[0125] (5) DG active output constraints

[0126] P DG,min ≤P DG (n)≤P DG,max (17)

[0127] Where P DG (n) represents the active power output by DG in period n; P DG,min , P DG,max They are the minimum and maximum values ​​of DG active output respectively.

[0128] (7) ESS active output constraints

[0129] PESS,min ≤P ESS (n)≤P ESS,max (18)

[0130] Where P ESS (n) represents the charging and discharging power of ESS in period n, P ESS,max , P ESS,min They represent the upper and lower limits of ESS charging and discharging power in time period t respectively.

[0131] (8) Reliability constraints

[0132] R≥R 0 (19)

[0133] In the formula, R 0 It represents the minimum limit of the reliability requirement of the distribution network system, and R represents the reliability value of the current distribution network system.

[0134] (9) Power shortage constraints

[0135] W ENS ≤W max (20)

[0136] W max =8760×(1-ASAI)×P(21)

[0137] Where W ENS , W max They represent the actual power shortage and maximum power shortage of the system respectively; ASAI represents the average power supply availability; and P is the total load power consumption per hour in the distribution network area.

[0138] By organically integrating the above-mentioned three-layer planning model of active distribution network based on multi-source collaborative optimization, we can obtain Figure 2 The overall framework of the three-level planning model of active distribution network is shown as follows:

[0139] The upper-level planning aims to maximize the utilization rate of clean energy, select the capacity and site for the generalized power source, and optimize the configuration of the sectional switches to improve the reliability and safety of the distribution network; the middle-level planning aims to maximize user satisfaction and determine the number and amount of contracted controllable loads; the lower-level planning divides the whole year into 8760 time periods in units of one hour, and takes the minimum network loss as the goal to dispatch and control the output of various energy sources and the opening and closing of sectional switches, and comprehensively considers the construction and operation costs of distributed energy, switch switching costs, network loss costs, power purchase costs, reliability costs, environmental costs, etc., to obtain the total cost of the planning. The upper and middle-level planning models customize the most self-interested plans for the two different stakeholders, energy suppliers and demand-side load users, and pass them to the lower-level planning model to obtain the optimization plan that satisfies the minimum total network loss of the distribution network operation, and pass it to the upper and middle-level planning models again. Through the interaction and mutual influence between the optimal strategies of the planning models at each level, the optimization results that meet the overall optimality are decided.

[0140] The Seeker Optimization Algorithm (SOA) is a new type of group intelligence algorithm that combines human social learning and cognitive learning. The algorithm simulates the self-centered selfish behavior of humans when they are exploring an unknown target, the altruistic behavior that reflects the value of personal contribution to society, and the pre-action behavior based on social experience. On the basis of "historical experience", it completes the judgment and reasoning of the uncertainty of the unknown target, so that each search individual integrates human thinking learning, experience exchange, solidarity and cooperation and other intelligent behaviors when completing the update of the search direction and search step length. While improving the system's optimization speed, it can also ensure the robustness of the system. The following four aspects introduce the specific optimization steps of the crowd search algorithm:

[0141] Step 1: In each dimension, the search direction f randomly generated by the searcher mk0 With search step length β mk0 Multiply them together to obtain the initial value of the searcher within the feasible solution range, completing the initialization of the SOA algorithm:

[0142] x mk0 =β mk0 ·f mk0 (twenty two)

[0143] Where: f mk0 It represents the search direction of the mth search individual in the kth dimension. In particular, when f mk0 When the value is 1, it means that the search individual moves forward; when the value is -1, it means that the search individual moves backward; when the value is 0, it means that the search individual stays where it is.

[0144] Step 2: Determine the search step length through uncertain reasoning

[0145]

[0146] Where: α mk is the search step length of the mth search individual in the kth dimension, δ mk 、u mk They represent the Gauss membership function parameters and the membership of the j-th dimension objective function i. The Gauss membership function is a fuzzy variable used to construct the intrinsic relationship between the perception and behavior of the searcher. mk It is used to simulate the randomness of the searcher's search behavior and can be expressed by formula (24). Then, the search step length δ is deduced by uncertainty reasoning. mk :

[0147] u mk =rand(u m ,1)(k=1,2,...,D) (24)

[0148] δ mk =ω·|x min -x max | (25)

[0149] ω=(iter max -iter) / iter max (26)

[0150] Where: function rand(u m , 1) represents u m A random real number in the range from 1 to 1, x max and x min They represent the positions of the maximum and minimum values ​​of the function in the search population, and ω represents the inertia weight, which is the value of the current iteration number iter and the maximum optimization number iter. max Jointly decided.

[0151] Step 3: Determine the search direction. The search direction of the crowd is determined by the self-interest direction. Altruistic Direction With pre-movement direction The geometric mean of the three is obtained by randomly weighting them.

[0152]

[0153]

[0154]

[0155]

[0156] In formula (27) to formula (29), and Respectively The best location in Indicates the position of the mth search individual in the kth dimension in the t'th update. The arrow is the position variable in the algorithm. and They represent the best historical position of the mth search individual in the region of the kth dimension and in the population; sign() is the sign function; Both represent arbitrary constants between 0 and 1; ω represents the inertia weight.

[0157] Step 4: Update the position of the searcher in the group through equation (31)-equation (32) and recalculate the fitness value of the searcher.

[0158]

[0159] x mk (t'+1)=x mk (t')+Δx mk (t'+1) (32)

[0160] Figure 3 What is shown is the specific optimization process of the crowd search algorithm:

[0161] The present invention takes the crowd search algorithm as the basic algorithm, extracts the core ideas of GA and SFLA, introduces the genetic variation strategy of the elite, the grouping and mixing strategy of the frog group into the crowd search algorithm, thereby greatly improving the optimization efficiency of the algorithm while effectively solving the problem of the algorithm being in the local optimum, so that the overall performance of the SFLA-GA crowd search algorithm is significantly improved, giving full play to its superiority in solving the system optimization research carried out by the present invention.

[0162] The specific algorithm flow of the improved crowd search algorithm based on SFLA-GA is shown in Figure 4 The present invention mainly improves the crowd search algorithm in the following two aspects:

[0163] (1) The algorithm is trapped in the local optimum and the convergence speed is too slow, which will greatly affect the scientificity and accuracy of the optimization model solution. According to the strategy of simulating the grouping action of frogs and regular mixing in SFLA, the population P is first divided into S groups in ascending order, with T individuals in each group; the search step size and search direction are updated, and then the populations of different groups are mixed, so that the information carried by individuals can be better exchanged globally, effectively solving the problem of the algorithm being trapped in the local optimum.

[0164] (2) In order to improve the optimization speed of the algorithm, the present invention adopts different update methods for the L individuals ranked at the bottom of the group, and uses the diversity of update methods to seek a faster and more accurate search for the optimal solution to the problem to be solved. Using the idea of ​​genetic mutation in the genetic algorithm, the top two individuals in the group are regarded as elite parents, and new L individuals are obtained through genetic crossover mutation of the elite parents:

[0165]

[0166] Where: represents the individual with ranking t in the sth group; r represents an arbitrary constant in the range of [0,1].

[0167] In addition, the remaining TL individuals in the group use equations (31) and (32) to update their positions.

[0168] like Figure 5 As shown, the solution of the three-layer planning model of the active distribution network constructed by the present invention mainly consists of two parts: an improved crowd search algorithm based on SFLA-GA and a CLMCS probabilistic power flow calculation method taking into account uncertain factors. The specific solution steps of the algorithm are as follows:

[0169] Step 1: Initialize the parameters of the improved crowd search algorithm and the CLMCS probabilistic power flow calculation method, and input the original parameters of the example network system.

[0170] Step 2: Initialize the optimization candidate solutions represented by each searcher in the crowd. The information carried by each searcher in different search dimensions represents a decision variable, which can be represented by matrix X:

[0171]

[0172] In the formula, and (i=1,2,…N w ,j=1,2,…N p ,k=1,2,…N e ,l=1,2,…N c q=1,2,…N s , m=1,2,…M) respectively represent the installed capacity of wind turbines and photovoltaic power generation equipment at the i-th, j-th, and k-th RDG access nodes, the contracted amount of controllable load at the l-th node, and the access position of the section switch is q; N w 、N p 、N e 、N c 、N sThey are the alternative numbers of wind turbines, photovoltaic generators, energy storage devices, controllable load users and section switches.

[0173] Step 3: Bring the information stored by the searcher into the CLMCS power flow calculation method, calculate and save the probability power flow results of each candidate planning scheme one by one, and check whether it meets the limit standards of various constraints. If it does not meet the standards, the crowd search algorithm is used to re-update the searcher's search information.

[0174] Step 4: Update and store the optimal position and optimal fitness value of the current searcher, and determine whether the end condition of the search is met. If the condition is met, the loop is terminated and the optimal planning solution is output.

[0175] Step 5: Use the improved crowd search algorithm to update the searcher's search direction and search step, and return to step 3 to re-apply the CLMCS method to perform power flow calculation and analysis on the updated crowd individuals until the termination condition of the algorithm is met in step 4.

[0176] In order to verify the effectiveness of the above three-layer planning model of active distribution network considering distribution network segment optimization and multi-source collaborative optimization for optimizing power grid planning and design scheme, the present invention adopts Figure 6 The IEEE-33 node power distribution system shown in the figure is used as an example for simulation analysis. The voltage level of the system is 12.66 kV. Among them, node 1 is a substation node, which is set as a balancing node.

[0177] The parameters of the improved crowd search algorithm are set as follows: the initial number of people is 100, divided into 10 teams, each with 10 people; the maximum number of iterations is set to 200; the number of worst-improved individuals in the team P = 2. The construction costs of WG, PV, and ESS are 5.35 million / MW, 4.65 million / MW, and 5.85 million / MW, respectively, and the operation and maintenance costs are 0.3 yuan / (kW·h), 0.2 yuan / (kW·h), and 0.3 yuan / (kW·h), respectively; the government environmental protection subsidy for RDG is 0.2 yuan / (kW·h), and the unit compensation price of controllable load is 2 yuan / (kW·h); the investment and construction cost of the section switch is 25,000 yuan / unit, the average failure rate of the main feeder and branch line is set to 0.1 times / (km·year) and 0.01 times / (km·year), respectively, and the average repair time is 4h / time and 10h / time, respectively; the switching operation time of the section switch and the connecting switch is 20min / time and 30min / time, respectively

[72] The power supply reliability of the system should be higher than 99.98%. The system planning period is 20 years, the annual interest rate is 10%, the return on investment of the power industry is 10%, and the unit electricity price is 0.5 yuan / (kW·h). The maximum access capacity of DG in the active distribution network is 30% of the total active load of the system, and the installation capacity of ESS and the contracted control capacity of controllable load cannot exceed 10% of the total active load.

[73] .

[0178] Table 2 shows the optimization configuration results of various energy sources including DG including wind power generation, photovoltaic power generation, ESS, controllable load, etc. after Matlab simulation using the three-layer planning model described in the present invention. The total access capacity of DG is 1110kW, the total installed capacity of ESS is 360kW, and the total contracted capacity of controllable load is 370kW.

[0179] Table 2 Optimal configuration of various energy sources

[0180]

[0181] Table 3 shows the changes in various indicators after considering the optimization of distribution network segmentation. The data is compared and analyzed. Although the installation of segment switches in ADN increases the investment, maintenance and scrapping costs of equipment, it increases the reliability index of power shortage throughout the year by 0.0149%, reduces the annual power shortage by 52.43% compared with the case where segment switches are not installed, and reduces the reliability cost by 34.30%. In summary, considering the optimal configuration of segment switches in ADN planning can not only improve the power supply reliability of the power grid, but also maximize the investment returns of enterprises.

[0182] Table 3 Data comparison before and after distribution network segment optimization

[0183]

[0184] In order to study the advantages of the three-layer model of active distribution network constructed by the present invention considering the segmented optimization of distribution network and multi-source collaborative optimization in distribution network planning, the simulation results are compared with the optimization results of the ADN planning model with only DG grid connection and the distribution network planning model without any DER grid connection, as shown in Table 4. From the perspective of planning cost, although the distribution network without DER access does not need to consider the investment, construction and operation costs of energy, the capital investment caused by network loss, power purchase, pollutant emissions, power shortage, etc. is much higher than the planning model of the present invention. Compared with the planning results of the present invention, the comprehensive investment cost is reduced from the original 10.5029 million to 7.9594 million, a decrease of 24.22%; if only considering the access of distributed power sources in the active distribution network, although it can reduce the network loss cost and pollution discharge cost to a certain extent, and reduce the total investment cost of the operator, it is still insufficient compared with the planning model of the present invention. Because the planning model of the present invention takes into account the optimal configuration of segmented switches, the reliability cost of the power grid is reduced by 40.77% and 47.15% respectively compared with the other two planning models, which fully verifies the necessity of segmented optimization of the distribution network. In addition, the planning model of the present invention comprehensively considers the coordinated optimization of multi-level energy. Although it increases the capital investment in the construction and operation of various energy sources, other costs have decreased to varying degrees. At the same time, due to the coordination and cooperation of various energy sources, the acceptance rate of the power grid for DG has been improved. Finally, the pollutant gas emissions and average active power losses caused by the planning model of the present invention are also lower than those of the other two planning models. The above data analysis results show that the planning model of the present invention has carried out overall optimization planning for the active distribution network from the three levels of economy, operation, and environment, verifying its advantages in reducing the operating costs of the power grid, improving the reliability of power grid operation, and saving energy and reducing emissions.

[0185] Table 4 Comparison of the results of the three-layer planning model of active distribution network and other planning models

[0186]

[0187] In order to verify the scientificity and effectiveness of the improved crowd search algorithm proposed in this invention in solving the ADN planning problem based on multi-source collaborative optimization, the iteration times and iteration results of the genetic algorithm, hybrid frog leaping algorithm and original crowd search algorithm are compared with the algorithm of this invention. Figure 7 The comparison shown.

[0188] The comparison results show that when the hybrid frog leaping algorithm is used to solve the planning model established by the present invention, after 5 iterations, the algorithm falls into a local optimum, and the total planning cost obtained is 9.1153 million yuan; the number of iterations solved by the genetic algorithm is 72 times, and the solution result is 8.7178 million yuan. The number of iterations, convergence speed and search accuracy of the algorithm are inferior to those of the crowd search algorithm. In addition, the original crowd search algorithm is compared with the improved crowd search algorithm of the present invention. The total planning cost obtained by the original crowd search algorithm is 8.1982 million yuan, the number of iterations is 63 times, and the total planning cost obtained by the improved algorithm proposed by the present invention is 7.9594 million yuan, and the number of iterations is 9 times. It can be seen that the algorithm of the present invention greatly improves the convergence speed of the optimization algorithm and obtains optimization results that are more in line with the economic benefits of the enterprise.

[0189] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. An active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources, characterized in that: The following steps are involved: S1. Establish an active ADN planning model based on multi-source collaboration; S2, the improved crowd search algorithm based on SFLA-GA and the CLMCS probabilistic power flow calculation method taking into account uncertain factors are used to solve the model to realize active distribution network planning.

2. According to claim 1, a method for active distribution network planning for improving the utilization rate and absorption capacity of distributed energy resources is characterized in that: The ADN planning model based on multi-source collaboration is a three-layer planning model, including upper-layer planning, middle-layer planning and lower-layer planning; the upper-layer planning aims to maximize the utilization rate of clean energy and determines the generalized power supply capacity site selection and the optimal configuration of segmented switches; The middle-level planning aims to maximize user satisfaction and determines the number and amount of contracts for controllable loads; the lower-level planning aims to minimize network losses and dispatches and controls the output of various energy sources and the opening and closing of section switches.

3. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 2 is characterized in that: The upper-level planning determines the installation location and capacity of the distributed generation device DG and the energy storage device ESS from the perspective of maximizing the utilization rate of clean energy. Its objective function is: Among them, η multi-s is the clean energy utilization rate, P load is the total load of the distribution network system, n is the total number of nodes in the distribution network system; η DG , η ESS are the access capacity of DG and ESS and P load The ratio of DG,a , P ESS,b are the active powers of the DG and ESS connected to nodes a and b respectively; c and d are the access position sets of the selected DG, ESS and controllable load respectively.

4. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 2 is characterized in that: The middle-level planning objective function is as follows: Where E is user satisfaction, N is the number of controllable loads signed; μ 1,n' , μ 2,n' are the influencing factors of the shiftable load and interruptible load at the node where the n'th controllable load is located; T a,n' , T b,n' They represent the expected scheduling time of the shiftable load and interruptible load at the node where the n'th controllable load is located, T n',all Then it is the total power consumption time of the n'th controllable load.

5. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 2 is characterized in that: The lower-level planning objective function is as follows: Among them, P LOSS is the network loss, P LINE represents the line loss of the distribution network; YT represents the number of time periods in a year; α ESS That is, the state change dimension of the energy storage device, which is 0 or 1; ΔP ESS.i' is the loss of the i'th ESS during state transition, n ESS is the total number of ESS; ξ ch.i' , dis.i' , are the charging and discharging efficiency and power of the i'th ESS in the h period respectively.

6. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 1 is characterized in that: The specific steps of S2 are as follows: S21, initializing various parameters of the improved crowd search algorithm based on SFLA-GA and the CLMCS probability power flow calculation method, and inputting various original parameters of the computing power network system; S22. Initialize the optimization candidate solutions represented by each searcher in the crowd. The information carried by each searcher in different search dimensions represents a decision variable, which is expressed as a matrix X: In the formula, and Respectively represent the installed capacity of wind turbines and photovoltaic power generation equipment at the i-th, j-th, and k-th RDG access nodes, the contracted amount of controllable load at the l-th node, and the access position of the section switch is q; N w 、N p 、N e 、N c 、N s are the alternative numbers of wind turbines, photovoltaic generators, energy storage devices, controllable load users, and section switches respectively; S23, bringing the information stored by the searcher into the CLMCS power flow calculation method, calculating and saving the probability power flow results of each candidate planning scheme one by one, and checking whether it meets the limit standards of various constraint conditions, and obtaining the fitness value of the objective function; If the criteria are not met, the crowd search algorithm is used to re-update the searcher’s search information; S24, update and store the optimal position and the optimal fitness value of the current searcher, and determine whether the end condition of the optimization search is met. If the condition is met, the loop is terminated and the optimal planning scheme is output. Otherwise, execute step S25; S25. Use the improved crowd search algorithm based on SFLA-GA to update the position of the searcher, return to step S23, re-apply the CLMCS power flow calculation method, and perform power flow calculation analysis on the updated crowd individuals until the termination condition of the algorithm is met in step S24.

7. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 6 is characterized in that: The optimization candidate solution represented by each searcher in the crowd is initialized as follows: In each dimension, the search direction f randomly generated by the searcher mk0 With search step length β mk0 Multiplying together gives the initial value of the searcher within the range of feasible solutions: x mk0 =b mk0 ·f mk0 (22) Where: f mk0 It represents the search direction of the mth search individual in the kth dimension; when f mk0 When the value is 1, it means that the search individual moves forward; when f mk0 When it is -1, it means that the search individual moves backwards; when its value is 0, it means that the search individual stays where it is.

8. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 7 is characterized in that: The improved crowd search algorithm based on SFLA-GA is used to update the location of the searcher, specifically: According to the strategy of simulating frog grouping and periodic mixing in SFLA, the population P is first divided into S groups in ascending order, with T individuals in each group; the L individuals with the lowest ranking in the group and the remaining TL individuals use different position update methods, and the populations of different groups are mixed after the searcher's position is updated; The position update method for the last L individuals in the group is: regard the top two individuals in the group as elite parents, and obtain new L individuals through genetic crossover mutation of the elite parents: Where: represents the individual with rank t in the sth group; r represents an arbitrary constant in the range [0,1]; The position update method of the remaining TL individuals in the group is: use the crowd search algorithm to determine the search step and search direction of the searcher, and update the searcher's position according to the search step and search direction.

9. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 8 is characterized in that: The crowd search algorithm is used to determine the search step length and search direction of the searcher, and the position of the searcher is updated according to the search step length and search direction, specifically: Determine the search step size through uncertain reasoning: Where: α mk is the search step length of the mth search individual in the kth dimension, δ mk 、u mk represent the Gauss membership function parameters and the membership of the k-th dimension objective function m respectively; the Gauss membership function is a fuzzy variable used to construct the intrinsic relationship between the perception and behavior of the searcher. mk It is used to simulate the randomness of the searcher's search behavior and can be expressed by formula (24). Then, the search step length δ is deduced by uncertainty reasoning. mk : u mk =rand(u m ,1)(k=1,2,...,D) (24) d mk =ω·|x min -x max |(25) ω=(path max -iter) / iter max (26) Where: function rand(u m , 1) represents u m A random real number in the range from 1 to 1, x max and x min They represent the locations of the maximum and minimum values ​​of the objective function in the search population, and ω represents the inertia weight, which is the value of the current iteration number iter and the maximum optimization number iter. max jointly determined; Determination of search direction; the direction of crowd search is determined by the self-interest direction Altruistic Direction With pre-movement direction The geometric mean of the three is obtained by randomly weighting them: In formula (27) to formula (29), and Respectively The best location in represents the position of the mth search individual in the kth dimension in the t'th update, and They represent the best historical position of the mth search individual in the region of the kth dimension and in the population; sign() is the sign function; All represent any constant between 0 and 1; ω represents the inertia weight; Through equation (31)-(32), the position of the searcher in the group is updated and the fitness value of the searcher is recalculated: x mk (t'+1)=x mk (t')+Δx mk (t'+1) (32)。 10. The active distribution network planning method for improving the utilization rate and absorption capacity of distributed energy resources according to claim 1 is characterized in that: The planning model comprehensively considers multiple costs, including distributed energy construction and operation costs, switch switching costs, network loss costs, power purchase costs, reliability costs and environmental costs, and obtains the total cost of the planning. The mathematical model of the annual cost of the planning is: minC total =C build +C OR.DG +C OR.ESS +C B +C E +C LOSS +C R (11) In the formula, C total is the total cost, C build is the construction cost of distributed generation and energy storage devices, C OR.DG is the DG operating cost, C OR.ESS is the efficiency of the energy storage device, C B is the electricity purchase cost, C E is the environmental cost, C LOSS is the network loss cost, C R The cost of reliability.

Images