Two-layer optimization dispatching method based on dynamic reconstruction and reactive power voltage adjustment based on improved whale optimization algorithm

Abstract

The double-layer optimization dispatching method of dynamic reconstruction and reactive voltage adjustment based on the improved whale optimization algorithm includes the following steps: establishing a distribution network node system; constructing a double-layer optimization dispatching model of active distribution network with dynamic network reconstruction and reactive voltage adjustment: the upper optimization model takes the lowest comprehensive operation cost of the distribution network as the goal, adopts opportunity constraint programming to deal with the randomness of wind and solar output, and takes the dynamic reconstruction of opportunity constraints as a means to achieve the economic operation of the distribution network on a long-term scale; the lower optimization model takes the minimum total deviation of the system node as the goal, takes static reactive compensation, micro gas turbine and energy storage output as the control objects, and realizes the optimization of the system voltage quality; adopts a hybrid solution strategy combining the improved whale optimization algorithm with the interior point method to solve the double-layer optimization dispatching model of the active distribution network. The present invention can greatly reduce network loss and improve voltage quality, taking into account the economic operation of the distribution network and the system voltage stability.

Description

Technical Field

[0001] The present invention belongs to the technical field of active distribution network optimization dispatching, and in particular to a double-layer optimization dispatching method of dynamic reconstruction and reactive voltage adjustment based on an improved whale optimization algorithm. Background Art

[0002] In recent years, in order to promote energy conservation and emission reduction and solve the energy crisis, the high penetration of distributed power sources such as photovoltaic and wind power has led to the development of distribution networks into active distribution networks. The power injected by intermittent distributed power sources is highly random, making the traditional distribution network reconstruction scheme no longer applicable; in addition, the voltage fluctuations or overvoltages caused by the high penetration of distributed power sources connected to the grid cause them to disconnect from the grid, which seriously restricts the ability of active distribution networks to absorb renewable energy. Therefore, it is of great significance to study the coordinated optimization of dynamic reconstruction of distribution networks and reactive voltage adjustment considering the time-varying and uncertain nature of distributed power sources to ensure the safe and economical operation of distribution networks.

[0003] As key active management strategies, network reconstruction, DG output dispatch and reactive power optimization have been explored by many scholars and have achieved a series of research results. The technical solutions disclosed in the Chinese patent "A distribution network reconstruction method for improving the capacity of accepting new energy" (CN109741207A) and the Chinese patent "A distribution network reactive power optimization method based on adaptive particle swarm algorithm" (CN113489019A) only consider a single active management measure, and do not consider the coordinated optimization between multiple active management strategies, which makes it difficult to give full play to the control capabilities of various controllable devices in the active distribution network.

[0004] Although the Chinese patent "Active distribution network reconstruction and reactive power joint robust optimization method considering extreme scenarios" (CN112671047A) takes network reconstruction and reactive power optimization into consideration, the method is limited to static reconstruction and does not fully consider the impact of the timing fluctuations of clean energy on the active distribution network.

[0005] The Chinese patent "A method for optimizing and dispatching active distribution networks considering dynamic network reconstruction" (CN110021966A) realizes dynamic reconstruction and joint optimization of reactive power, but does not take into account the randomness of clean energy, which limits its application in practice. For active distribution networks with high penetration of renewable energy, it is difficult to ensure the voltage safety of all nodes at the end of the line by simply utilizing the flexibility of the distribution network or reactive power compensation devices. The active distribution network has abundant voltage regulation resources, but the role of energy storage battery charging and discharging in voltage control has not been fully explored.

[0006] The whale optimization algorithm is a swarm intelligence optimization algorithm that simulates the foraging behavior of humpback whales. It seeks the global optimal solution to the optimization problem through the mechanism of encirclement predation, spiral update, and prey search. Compared with other optimization algorithms, the whale optimization algorithm has a simple principle, few adjustment parameters, and strong optimization ability. However, in the process of solving the reconstruction problem, there are problems such as a large number of infeasible solutions in the randomly generated initial population, slow convergence, and easy to fall into the local optimum. Summary of the invention

[0007] The purpose of the present invention is to provide a two-layer optimization scheduling method for dynamic reconstruction and reactive voltage adjustment based on an improved whale optimization algorithm. The method is a two-layer optimization scheduling method that can take into account the flexibility of the distribution network structure and the voltage regulation resource capacity. The method constructs a two-layer optimization scheduling model for active distribution networks with dynamic network reconstruction and reactive voltage adjustment; at the same time, the method uses a hybrid solution strategy combining a decimal-coded improved whale optimization algorithm and an interior point method to solve the problem. In view of the problems that the randomly generated initial population is full of a large number of infeasible solutions, the convergence speed is slow, and it is easy to fall into the local optimum in the process of solving the reconstruction problem with the whale optimization algorithm, the genetic algorithm crossover mutation idea, the nonlinear change convergence factor and the adaptive weight strategy are introduced for improvement, which can maintain population diversity and quickly find the global optimal solution.

[0008] The technical solution adopted by the present invention is:

[0009] A two-layer optimization dispatching method based on dynamic reconstruction and reactive power voltage adjustment of improved whale optimization algorithm.

[0010] Step 1: Collect active distribution network technical parameters and establish distribution network node system;

[0011] Step 2: Taking full account of the uncertainty of the output of distributed wind turbines and photovoltaic power generation systems, a probabilistic model is used to describe them; a two-layer optimization dispatch model of active distribution network with dynamic network reconstruction and reactive voltage adjustment is constructed:

[0012] The upper optimization model aims to minimize the comprehensive operation cost of the distribution network, adopts chance-constrained programming to deal with the randomness of wind and solar power output, and uses dynamic reconstruction of chance constraints as a means to achieve economic operation of the distribution network over a long period of time.

[0013] The lower optimization model aims to minimize the total deviation of system nodes, and takes static VAR compensation, micro gas turbine and energy storage output as control objects to optimize the system voltage quality.

[0014] Step 3: A hybrid solution strategy combining the improved whale optimization algorithm and the interior point method is used to solve the active distribution network two-layer optimization dispatching model in step 2. The upper optimization model transmits the grid structure to the lower layer, and the lower optimization model feeds back the optimal output of static VAR compensation, micro gas turbines and energy storage to the upper model, and iterates repeatedly until the optimal grid structure and the optimal output of controllable equipment are found to minimize the objective function value.

[0015] In the step 1, the parameters are initialized, and an active distribution network topology including two types of distributed power sources, wind power and photovoltaic power, is established. First, the nodes and branches are numbered, and then various controllable devices such as static VAR compensation, energy storage and micro gas turbines are connected to the designated nodes to achieve coordinated optimization between the integration of distributed power sources into the active distribution network and various active management strategies.

[0016] In step 2, considering the random characteristics of wind and solar power output, a probabilistic model is used to describe it; the established mathematical model of active distribution network double-layer optimization dispatch is as follows:

[0017] ① Randomness model of distributed wind turbine output:

[0018] The variable speed constant frequency wind power generation model is adopted, and the wind turbine output is simulated using Monte Carlo simulation and two-parameter Weibull distribution. The relationship between wind turbine output power and wind speed is as follows:

[0019]

[0020] Where: P t w is the active power output of the fan at time t, P r is the rated power of the generator; v t is the wind speed at time t, v ci is the cut-in wind speed, v r is the rated wind speed, v co is the cut-out wind speed, where A=P r / (v r -v ci ), B = -Av ci .

[0021] The expected value of the fan output power can be obtained from the above formula:

[0022]

[0023] Where: E(P DG,t ) is the expected value of the fan output power, f(v t ) is the wind speed probability density function. In the present invention, the wind turbine is a constant power factor model, and the wind turbine generator can be simplified to a PQ node;

[0024] ② Photovoltaic output model:

[0025] Since the photovoltaic illumination intensity satisfies the Beta distribution, the present invention assumes that the photovoltaic generator power output probability density function is as follows:

[0026]

[0027]

[0028] Where: Γ is the Gamma function, P s is the output power of the solar panel, P smax is the maximum output power of the solar panel, α and β are the shape parameters of the Beta distribution, u and δ are the standard deviation and variance of the light intensity in a certain period of time, Γ(α+β), Γ(α), and Γ(β) are the gamma function values ​​corresponding to the shape parameters in the Beta distribution.

[0029] The expected value of photovoltaic output power can be obtained:

[0030]

[0031] Where: f(P s ) is the power output probability density function of the photovoltaic generator, E(P s ) is the expected value of photovoltaic output power.

[0032] The photovoltaic power generation system can use capacitors to ensure that the power factor is constant and can be regarded as a PQ node in power flow calculation.

[0033] The established two-layer optimization dispatching model of active distribution network is as follows:

[0034] The objective function minf1 of the upper optimization model based on chance constraints has a probability constraint of:

[0035] Pr(f(X,ξ)≤f1)≥β;

[0036] Where: X is the control variable, which is the switch state of each branch; ξ is the state variable; Pr(·) represents the probability of event establishment; β is the target confidence level; f(X,ξ) is the comprehensive operating cost value of the system under state ξ, and f1 is the minimum value of f(X,ξ) when the probability level is at least β.

[0037] f(X,ξ)=C Ploss +C swi ;

[0038] Where: C Ploss is the network loss cost, C swi Cost of switch operation;

[0039]

[0040] Where: c L is the electricity price in period t; P loss,t is the equivalent network loss that meets a certain confidence level within time period t; △t = 1h;

[0041]

[0042] Where: Ω b is the set of power grid branches, l is a branch in the set of branches; M is the total number of sections after time segmentation, m is a single time period in the total segmentation; c act The cost of a single switch operation; l,m-1 、s l,m These are the switch states of branch l in period m-1 and period m, respectively, with open being 0 and closed being 1;

[0043] Node voltage constraints:

[0044] Pr(V i,min ≤V i ≤V i,max )≥β U

[0045] Where: V i,min and V i,max are the upper and lower limits of node voltage respectively; β U is the voltage constraint confidence level; Pr(V i,min ≤V i ≤V i,max ) is the probability that the node voltage is qualified.

[0046] Active and reactive power flow constraints:

[0047]

[0048] Where:

[0049] P i in and are the active power and reactive power injected into node i respectively; v i 、v j are the voltage amplitudes at node i and node j respectively; N is the total number of nodes; G ij , B ij ,θ ij are the conductance, susceptance and voltage phase angle values ​​between nodes i and j respectively, i, j = 1, 2, …, n, where n is the total number of nodes.

[0050] Branch flow constraints:

[0051] Pr(|τ ij P l|≤P l,max )≥β I ;

[0052] Where: P l is the transmission power of line l; P l,max is the upper limit of the transmission power of line l; τ ij is the switch state of branch ij, τ ij =0 means the branch is disconnected, τ ij =1 means the branch is closed;

[0053] Pr(|τ ij P l |≤P l,max ) represents the probability that the branch transmission power is qualified, β I is the line transmission power confidence level.

[0054] Distributed generation (DG) output constraints:

[0055] Pr(0≤P DG,i ≤P DG,i,max )≥β DG ;

[0056] Where: P DG,i is the DG output value at node i; P DG,max is the upper limit of DG output at node i; Pr(0≤P DG,i ≤P DG,i,max ) is the probability that the DG output at node i is qualified; β DG Output confidence level for DG.

[0057] Radial network topology constraints:

[0058] During the reconstruction and optimization process, the distribution network must meet the radial structure operation mode, and no island or ring network should occur.

[0059] Switching action times constraint:

[0060] N i ≤N i,max ;

[0061] N all ≤N all,max ;

[0062] Where: N i Represents the number of operations of a single switch in a day, and the maximum number of operations is N i,max ;

[0063] N all is the number of all switch operations in a day, and the maximum number of operations allowed is N all,max

[0064] The objective function of the lower optimization model is:

[0065] Where: T is the computing period (24 hours); n is the total number of nodes; u i,t and u i,N are the actual voltage and rated voltage of node i at time t respectively.

[0066] Micro Turbine Generator (MTG) output limit:

[0067]

[0068]

[0069] Where: P MTG,i , Q MTG,i is the active and reactive output of the micro gas turbine at node i;

[0070] are the minimum active and reactive outputs of the micro gas turbine at node i respectively;

[0071] are the maximum active and reactive outputs of the micro gas turbine at node i respectively;

[0072] Static VAR compensation device SVC constraints:

[0073]

[0074] Where: Q svc,i,t is the reactive power compensation output of node i at time t;

[0075] and are the upper and lower limits of SVC reactive power compensation at node i respectively;

[0076] Energy storage battery operation constraints:

[0077] State of charge upper and lower limits:

[0078] E SOC,min ≤E SOC,t ≤E SOC,max ;

[0079] Where: E SOC,t is the remaining energy storage capacity at time t; E SOC,min and E SOC,max are the minimum and maximum states of charge, respectively;

[0080] Charge and discharge power:

[0081]

[0082] Where: c,t The charging flag at time t is 1 when charging and 0 when not charging;

[0083] λ d,t It is the discharge flag at time t, which is 1 when discharging and 0 when not discharging;

[0084]

[0085] Where: p c,t is the actual charging power at time t; p d,t is the actual discharge power at time t;

[0086] p c,max is the maximum charging power; p d,max is the maximum discharge power.

[0087] In step 3, the upper model solution process includes the following steps:

[0088] S1: Reconstruct the whole day period into four sections: 0:00-7:00, 7:00-15:00, 15:00-21:00 and 21:00-24:00. Set the reconstruction period s to 1, generate the initial grid population, determine whether all individuals in the population meet the radial operation conditions, if not, regenerate the individual, and set the maximum number of iterations to max_iter;

[0089] S2: passing the upper grid as input to the lower optimization model, starting the lower optimization model, and calculating the upper population fitness value according to the calculation results of the lower layer and the upper objective function;

[0090] S3: Update the position of the whale population and generate the offspring population; by simulating the foraging behavior of whale bubble nets, the whale optimization algorithm searches for prey through wandering, contraction and spiral encirclement mechanisms. The position of each humpback whale represents a feasible solution to the reconstruction problem, that is, the dimension of the whale is the number of basic loops, and the position of the whale is the position number of the disconnect switch of each loop.

[0091] S4:

[0092] ⑴ Based on the traditional whale optimization algorithm, the crossover and mutation operations of the genetic algorithm are introduced to ensure the diversity of feasible solutions and improve the global search ability of the algorithm. Among them, the crossover operation is to apply the crossover operator to the group, and for the selected pairs of individuals, part of the chromosomes between them are exchanged with a probability of 0.8 to generate new individuals; the mutation operation is to apply the mutation operator to the group, and for the selected individuals, one or some genes are changed to other alleles with a probability of 0.1.

[0093] (2) Introducing the adaptive weight adjustment strategy into the whale optimization algorithm, adaptively changing the weight value according to the current distribution of the whale population, which is conducive to maintaining population diversity and jumping out of the local optimum in time;

[0094]

[0095] Where: w is the adaptive weight factor, w min 、w max are the upper and lower limits of the weight coefficient, respectively, and t is the current number of iterations, which are 1.5 and 0.4 respectively in the present invention;

[0096] ⑶ In the process of encirclement search of the whale optimization algorithm, other humpback whale individuals all approach the current optimal individual to complete the encirclement of the target prey. In the iterative search process, the control parameter a decreases linearly from 2 to 0. The control factor a cannot balance the global search ability and local development ability well due to linear changes, and it is easy to fall into the local optimum, resulting in poor optimization results. Therefore, the present invention proposes a convergence factor based on a logarithmic function that changes nonlinearly with the number of evolutionary iterations, and its expression is as follows:

[0097]

[0098] In the formula: a is the control factor; e is the natural constant; t is the current number of iterations; max_iter is the maximum number of iterations; is a component that changes logarithmically with the number of iterations, resulting in nonlinear changes in the control factor.

[0099] S5: Calculate the fitness value of the updated whale population to form a new population and an old population, rearrange the fitness values ​​of all individuals in the combined population by size, retain some individuals with smaller fitness values, eliminate the remaining individuals, and update the global optimal fitness value and the optimal individual;

[0100] Repeat steps S2-S4 until the maximum number of iterations is reached, and then output the optimal grid and optimal fitness value of the first reconstruction period.

[0101] S6: Repeat steps S1-S6 until all grids of all reconstruction periods are output.

[0102] The solution process of the lower-level optimization model is as follows:

[0103] Step 1: Input the optimization results obtained by the upper-level optimization model, and initialize the lower-level optimization model population according to the upper-level optimization model input.

[0104] Step 2: Input the distributed power and load data at the first moment of the sth reconstruction period, and calculate the fitness value of the objective function of the lower optimization model.

[0105] Step 3: Use the interior point method to solve the lower optimization model, update the lower optimization model population until the iteration termination condition is met, and obtain the optimal output reference values ​​of static VAR compensation, energy storage and micro gas turbine at the first moment of the sth reconstruction period;

[0106] Step 4: At the next moment of the sth reconstruction period, repeat steps Step 1-Step 3, and output the fitness value and the optimal output of the controllable device at each moment of the sth reconstruction period.

[0107] Step 5: Calculate the fitness based on the lower layer objective function and feed it back to the upper layer. This step is connected to step 2 of the upper layer model solution process, and it is repeated in this way.

[0108] Step 6: Output the switch states of all reconstruction periods that meet the constraints, the optimal output set of each hourly controllable device, and the optimal values ​​of the objective functions of the upper and lower optimization models.

[0109] The present invention provides a double-layer optimization dispatching method for dynamic reconstruction and reactive voltage adjustment based on an improved whale optimization algorithm, and the technical effects are as follows:

[0110] 1) The present invention coordinates the optimization of the dynamic reconstruction of the distribution network and the reactive voltage adjustment, and adopts a two-layer optimization scheduling method. Compared with the single dynamic reconstruction and reactive optimization measures, it can greatly reduce the network loss and improve the voltage quality, taking into account the economic operation of the distribution network and the system voltage stability. The present invention is capable of dealing with the randomness and volatility of renewable energy, and has a certain guiding significance for obtaining a safe and economical operation plan according to the actual situation.

[0111] 2) The present invention introduces the crossover / mutation operator of the genetic algorithm into the whale optimization algorithm to enhance the global search capability of the algorithm; adopts an adaptive weight strategy to improve the original whale optimization algorithm so that it has the ability to jump out of the local optimum; uses the convergence factor of nonlinear changes to effectively balance the global search capability and the local search capability; uses a hybrid solution strategy that combines the improved whale optimization algorithm and the interior point method to solve the two-layer optimization scheduling of the active distribution network, so that the global optimal value can be found quickly, effectively solving the economic operation problem of the active distribution network.

[0112] 3) The present invention establishes a two-layer optimization scheduling model that takes into account the flexibility of the distribution network structure and the dynamic regulation of various controllable devices. The model is a multi-period decision variable nonlinear mixed integer dynamic optimization model that includes discrete variables (interconnection switch status) and continuous variables (static VAR compensation, micro gas turbine and energy storage output). The upper model uses dynamic reconstruction as the adjustment means, and the objective function is to minimize the comprehensive operating cost. The lower model uses energy storage and micro gas turbine optimal output as the control strategy, and the objective function is to minimize the total deviation and voltage of the system nodes. A hybrid strategy combining the improved whale optimization algorithm with decimal coding and the interior point method is used for optimization. The present invention improves the whale optimization algorithm by introducing the crossover mutation of the genetic algorithm, the convergence factor of nonlinear changes and the adaptive weight strategy, and can quickly find the global optimal solution. The operation plan determined by the method of the present invention can achieve comprehensive optimization of the distribution network operation economy and the system voltage level. BRIEF DESCRIPTION OF THE DRAWINGS

[0113] Figure 1 Schematic diagram of the improved distribution network system structure.

[0114] Figure 2 This is the daily load and wind and solar power generation prediction curve.

[0115] Figure 3 This is a structural block diagram of the active distribution network double-layer optimization scheduling model of the present invention.

[0116] Figure 4 It is a flow chart of the present invention.

[0117] Figure 5 This is the optimal output diagram of the energy storage device per hour according to an embodiment of the present invention.

[0118] Figure 6 This is the optimal output diagram of the micro gas turbine per hour according to an embodiment of the present invention.

[0119] Figure 7 This is a node voltage curve diagram obtained by optimizing different scenarios in the embodiments of the present invention.

[0120] Figure 8 This is a comparison chart of network losses obtained by optimizing different situations in the embodiments of the present invention.

[0121] Fig. 9 This is a network loss convergence curve diagram of the improved whale optimization algorithm according to an embodiment of the present invention.

[0122] Fig.10 Searching for a flowchart for the improved whale optimization algorithm. DETAILED DESCRIPTION

[0123] A two-layer optimization scheduling method with dynamic network reconstruction containing high-penetration renewable energy and output regulation of multiple controllable devices. Photovoltaic and wind turbines are connected to a 33-node distribution network system. The photovoltaic power generation and wind power generation are distributed power sources with high penetration rates. The distribution network will have power backflow and node voltage over-limit problems, which will bring serious challenges to the operation of the distribution network. A variety of controllable devices such as static VAR compensation, micro gas turbines and energy storage are connected to the distribution network. Through active management measures such as output regulation of controllable devices and switch reconstruction, the distribution network is jointly optimized from the two aspects of active power and reactive power. However, continuous variables and discrete variables need to be solved at the same time, and multi-variable and multi-constrained nonlinear mixed integer dynamic optimization problems make it difficult to solve. The whale optimization algorithm is improved by introducing the crossover mutation operator of the genetic algorithm, the convergence factor of nonlinear changes and the adaptive weight adjustment strategy. The hybrid solution strategy combining the improved whale optimization algorithm and the interior point method is used to solve the above model. It has the characteristics of fast convergence speed and convergence to the global optimum, making the operation of the active distribution network safer and more economical. It includes the following steps:

[0124] Step 1: Collect active distribution network technical parameters and establish an improved distribution network node system:

[0125] The technical parameters of the active distribution network include: the system reference voltage is 12.66kV, the reference power is 10MW, and the active distribution network topology with multiple distributed power sources is established. Photovoltaic units with a rated power of 2MW and a power factor of 0.95 are installed at nodes 13, 21, and 33 respectively; wind turbines with a rated power of 2MW and a power factor of 0.9 are installed at nodes 16 and 25 respectively. Five micro gas turbines with a rated power of 100kW and a power factor of 0.85 are connected at nodes 7 and 30. Two reactive compensation devices are connected at node 18, each of which provides reactive power in the range of -100-300kvar. A battery energy storage system with a capacity of 800kW·h is connected at node 11, with a maximum charge and discharge power of 200kW. The purchase price of electricity is 0.7 yuan / (kW·h), and the cost of a single switch operation is 7 yuan. The permissible range of node voltage standard value is set to [0.95, 1.05], and the confidence level of objective function, node voltage, DG output and branch power in chance constrained programming is 0.9. Then the improved distribution network test system is as follows: Figure 1 As shown, the predicted value curves of wind, solar and load are as follows Figure 2 shown.

[0126] Step 2: Taking full account of the uncertainty of the output of distributed wind turbines and photovoltaic power generation systems, a probabilistic model is used to describe them; a two-layer optimization dispatching model of active distribution network with dynamic network reconstruction and reactive power voltage adjustment is constructed, such as Figure 3 As shown, Figure 3This is the structure diagram of the two-layer optimization dispatching model of the active distribution network of the present invention. The upper layer is for dynamic reconstruction optimization, and the lower layer is for dynamic regulation of the output of various controllable devices, and gives the objective function of minimizing the overall operating cost of the upper layer and the objective function of minimizing the total deviation of the system node voltage of the lower layer, as well as the corresponding constraints of the upper and lower layers.

[0127] ① Randomness model of distributed wind turbine output:

[0128] The present invention adopts a variable speed constant frequency wind power generation model, and uses Monte Carlo simulation and two-parameter Weibull distribution to simulate the wind turbine output. The relationship between the output power and the wind speed is as follows:

[0129]

[0130] Where: P r is the rated power of the generator, v ci is the cut-in wind speed, v r is the rated wind speed, v co is the cut-out wind speed, where A=P r / (v r -v ci ), B = -Av ci .

[0131] The expected value of the fan output power can be obtained from the above formula:

[0132]

[0133] Where: E(P DG,t ) is the expected value of the fan output power, f(v t ) is the wind speed probability density function. In the present invention, the wind turbine is a constant power factor model, and the wind turbine generator can be simplified to a PQ node.

[0134] ② Photovoltaic output model

[0135] Since the photovoltaic illumination intensity satisfies the Beta distribution, the present invention assumes that the photovoltaic generator power output probability density function is as follows:

[0136]

[0137]

[0138] Where: Γ is the Gamma function, P s is the output power of the solar panel, P smax is the maximum output power of the solar panel, α and β are the shape parameters of the Beta distribution, and u and δ are the standard deviation and variance of the light intensity in a certain period of time.

[0139] The expected value of photovoltaic output power can be obtained:

[0140]

[0141] Where: E(P s ) is the expected value of photovoltaic output power. The photovoltaic power generation system can be guaranteed to have a constant power factor by capacitors and can be regarded as a PQ node in power flow calculation.

[0142] The established two-layer optimization dispatching model of active distribution network is as follows:

[0143] The upper objective function minf1 based on the chance constraint, the probability constraint of the objective function is:

[0144] Pr(f(X,ξ)≤f1)≥β

[0145] f(X,ξ)=C Ploss +C swi

[0146]

[0147]

[0148] Where: X is the control variable, which is the switch state of each branch; ξ is the state variable; f1 is the optimal target value achieved when a certain confidence level is met; Pr(·) represents the probability of event establishment; β is the target confidence level; Ω b is the collection of power grid branches; c L is the electricity price in period t; P loss,t is the equivalent network loss that meets a certain confidence level within time period t; △t = 1h; M is the total number of time periods after division; c act The cost of a single switch operation; l,m is the switch state of branch l in period m, open is 0, closed is 1, C Ploss is the network loss cost, C swi The cost of switch operation.

[0149] Node voltage constraints:

[0150] Pr(V i,min ≤V i ≤V i,max )≥β U

[0151] Where: V i,min and V i,max are the upper and lower limits of node voltage respectively; β U is the voltage constraint confidence level.

[0152] Active and reactive power flow constraints:

[0153]

[0154] Where: P i in and are the active power and reactive power injected into node i respectively; v i is the voltage amplitude at node i; G ij , B ij ,θ ij are the conductance, susceptance and voltage phase angle values ​​between nodes i and j respectively, i, j = 1, 2, …, n, where n is the total number of nodes.

[0155] Branch flow constraints:

[0156] Pr(|τ ij P l |≤P l,max )≥β I

[0157] Where: P l is the transmission power of line l; P l,max is the upper limit of the transmission power of line l; β I is the line transmission power confidence level.

[0158] Distributed generation (DG) output constraints:

[0159] Pr(0≤P DG,i ≤P DG,i,max )≥β DG

[0160] Where: P DG,max is the upper limit of DG output at node i; β DG Output confidence level for DG.

[0161] Radial network topology constraints:

[0162] During the reconstruction and optimization process, the distribution network must meet the radial structure operation mode, and no island or ring network should occur.

[0163] Switching action times constraint:

[0164] N i ≤N i,max

[0165] N all ≤N all,max

[0166] Where: N i Represents the number of operations of a single switch in a day, and the maximum number of operations is N i,max ; N all is the number of all switch operations in a day, and the maximum number of operations allowed is Nall,max .

[0167] The lower layer objective function:

[0168]

[0169] Where: T is the computing period (24 hours); n is the total number of nodes; u i,t and u i,N are the actual voltage and rated voltage of node i at time t respectively.

[0170] Micro Turbine Generator (MTG) output limit:

[0171]

[0172]

[0173] Where: P MTG,i , Q MTG,i is the active and reactive output of the micro gas turbine at node i; are the minimum active and reactive outputs of the micro gas turbine at node i respectively; are the maximum active and reactive outputs of the micro gas turbine at node i respectively.

[0174] Static VAR compensation device SVC constraints:

[0175]

[0176] Where: is the reactive power compensation output of node i at time t; and are the upper and lower limits of SVC reactive power compensation at node i respectively.

[0177] Energy storage battery operation constraints:

[0178] State of charge upper and lower limits:

[0179] E SOC,min ≤E SOC,t ≤E SOC,max

[0180] Where: E SOC,t is the remaining energy storage capacity at time t; E SOC,min and E SOC,max are the minimum and maximum states of charge, respectively.

[0181] Charge and discharge power:

[0182]

[0183] Where: c,t is the charging flag at time t, which is 1 when charging and 0 when not charging; d,t It is the discharge flag at time t, which is 1 when discharging and 0 when not discharging.

[0184]

[0185] Where: p c,t is the actual charging power at time t; p d,t is the actual discharge power at time t; p c,max is the maximum charging power; p d,max is the maximum discharge power.

[0186] Step 3: A hybrid solution strategy combining the improved whale optimization algorithm and the interior point method is used to solve the two-layer optimization scheduling model of the active distribution network in step 2. The upper layer transmits the grid structure to the lower layer, and the lower layer feeds back the optimal output of static VAR compensation, micro gas turbines and energy storage to the upper layer. It is repeated until the optimal grid structure and the optimal output of controllable equipment are found to minimize the objective function value.

[0187] Solve the above active distribution network two-layer optimization dispatch model flow chart combined with Figure 4 illustrate.

[0188] The upper model solution process is as follows:

[0189] S1: Reconstruct the whole day period into four sections: 0:00-7:00, 7:00-15:00, 15:00-21:00 and 21:00-24:00. Set the reconstruction period s to 1, generate the initial grid population, determine whether all individuals in the population meet the radial operation conditions, if not, regenerate the individual, and set the maximum number of iterations to max_iter;

[0190] S2: Pass the upper grid as input to the lower grid, start the lower optimization model, and calculate the upper population fitness value according to the calculation results of the lower grid and the objective function of the upper grid.

[0191] S3: Update the position of the whale population and generate the offspring population; by simulating the foraging behavior of whale bubble nets, the whale optimization algorithm searches for prey through wandering, contraction and spiral encirclement mechanisms. The position of each humpback whale represents a feasible solution to the reconstruction problem, that is, the dimension of the whale is the number of basic loops, and the position of the whale is the position number of the disconnect switch of each loop.

[0192] S4:

[0193] ⑴Introduce the crossover / mutation operator of the genetic algorithm to ensure the diversity of feasible solutions and improve the global search capability of the algorithm.

[0194] (2) Introducing the adaptive weight adjustment strategy into the whale optimization algorithm, the size of the weight is adaptively changed according to the current distribution of the whale population, which is conducive to maintaining population diversity and timely jumping out of the local optimum.

[0195]

[0196] Where: w min 、w max They are the upper and lower limits of the weight coefficient respectively, and the present invention takes them as 1.5 and 0.4.

[0197] ⑶Introduce the convergence factor of nonlinear changes into the whale optimization algorithm to effectively adjust the global search capability and local development capability.

[0198]

[0199] Where: t is the current iteration number; max_iter is the maximum iteration number.

[0200] S5: Calculate the fitness value of the updated whale population to form a new population and an old population, rearrange the fitness values ​​of all individuals in the combined population by size, retain some individuals with smaller fitness values, eliminate the remaining individuals, update the global optimal fitness value and the optimal individual, repeat steps 2-4 until the maximum number of iterations is reached, and output the optimal grid and optimal fitness value of the first reconstruction period.

[0201] S6: Repeat steps S1-S6 until all grids of all reconstruction periods are output.

[0202] The solution process of the lower model is as follows:

[0203] Step 1: Input the optimization results obtained from the upper layer and initialize the lower layer population according to the upper layer input.

[0204] Step 2: Input the distributed generation and load data at the first moment of the sth reconstruction period, and calculate the fitness value of the objective function of the lower model.

[0205] Step 3: Use the interior point method to solve the lower model and update the lower population until the iteration termination condition is met to obtain the optimal output reference values ​​of static VAR compensation, energy storage and micro gas turbine at the first moment of the sth reconstruction period.

[0206] Step 4: At the next moment of the s-th reconstruction period, repeat steps 1 to 3, and output the fitness value and the optimal output of the controllable device at each moment of the s-th reconstruction period.

[0207] Step 5: Calculate the fitness based on the lower-level objective function and feed it back to the upper-level layer. This step is connected to Step 2 of the upper-level model solution process, and the process is repeated in this way.

[0208] Step 6: Output the switch states of all reconstruction periods that meet the constraints, the optimal output set of each hourly controllable device, and the optimal values ​​of the upper and lower layer objective functions.

[0209] Example:

[0210] In order to fully explore the impact of dynamic reconstruction and dynamic control strategies of controllable devices such as energy storage and micro gas turbines on the optimal operation of the distribution network, and verify the effectiveness and feasibility of the coordinated optimization model proposed in this invention, four scenarios are set for comparison:

[0211] Scenario 1: Passive distribution network without any optimization measures;

[0212] Scenario 2: No energy storage, micro gas turbines, or static VAR compensation, only dynamic network reconstruction;

[0213] Scenario 3: Connect energy storage, micro gas turbines and static VAR compensation, and only implement dynamic control strategies for controllable devices;

[0214] Scenario 4: Comprehensive consideration of dynamic network reconstruction and dynamic regulation strategies for multiple controllable devices.

[0215] The optimization results of scenarios 1-4 can be seen from Table 1. Whether it is dynamic reconstruction of the grid or dynamic regulation of controllable devices, both can reduce network losses and improve voltage quality. Compared with scenario 2, scenario 3 reduces network loss costs by 58.21%, and the total voltage deviation is reduced by 16.34%. The optimization effect on the active distribution network is more significant. This is because the controllable equipment is optimized and dispatched once an hour. While achieving continuous and rapid regulation, it can not only provide a certain amount of reactive power support, but also release active power to regulate the system flow. Reconstruction does not have the above characteristics, and has defects such as limited reconstruction times and difficulty in finding the optimal solution to the multi-dimensional reconstruction problem, which affects the optimization effect of reconstruction. Scenario 4 shows that compared with a single optimization measure, flexible regulation of the output of controllable equipment and joint optimization of the grid can enable the active distribution network to achieve a better operating state. The two optimization measures complement each other and influence each other. Figure 5 and Figure 6 It can be seen from the output of energy storage and micro gas turbines that the power emitted by controllable devices to the system increases after grid reconstruction, further improving voltage quality and reducing network losses. Giving full play to the advantages of continuous regulation of controllable devices can reduce the number of switching actions during reconstruction, making up for the increase in network losses and average voltage deviation caused by the small number of distribution network reconstructions.

[0216] Table 1 Comparison of optimization effects in various scenarios

[0217]

[0218]

[0219] The high penetration of clean energy grid connection has caused ADN voltage limit and power quality problems to become more serious. Network reconstruction can improve the voltage level of the distribution network to a certain extent, but it is difficult to ensure the voltage safety of all nodes. In order to fully explore the impact of reactive power compensation devices and energy storage on the voltage of the distribution network, two scenarios, scenario 5 and scenario 6, were set up. Scenario 5: Add reactive power compensation devices on the basis of scenario 2 to explore the impact of static reactive power compensation on reconstruction. Scenario 6: On the basis of scenario 5, further explore the role of energy storage in optimizing the operation of the distribution network. Figure 7 The optimized distribution of node voltages before and after the reactive compensation connection in scenario 2 and scenario 5 is explained. It can be seen that after the network reconstruction in scenario 2, some node voltage amplitudes still exceed the limit at 9:00 and 14:00. Scenario 5 optimizes the voltage distribution of network nodes by connecting the reactive compensation device. The maximum voltage amplitude at 9:00 did not reach the upper limit, and the maximum voltage amplitude at 14:00 still exceeded the limit, indicating that the addition of reactive compensation devices after network reconstruction plays a certain role in regulating the voltage level, which can improve the situation of voltage exceeding the limit in a few nodes and effectively reduce the network loss of the distribution network, but it still cannot effectively eliminate the risk of voltage exceeding the limit. Scenario 5 and Scenario 6 are used to compare the node voltage distribution before and after the addition of energy storage. It is not difficult to find that in scenario 4, the DG output is relatively large during the period of 12:00-15:00, which meets the load demand, and the node voltage also exceeds the limit due to the increase in power reverse transmission; after the energy storage device is connected, it absorbs active power to make the voltage amplitude lower than 1.05U N , to prevent overvoltage. In scenario 4, during the 18:00-20:00 period, the peak load is high and the DG output is low. The voltage at the end of the line is low due to increased line losses. The active power released by the energy storage device will increase the voltage amplitude of the system nodes, indicating that energy storage can improve the overall voltage quality of the system and suppress the voltage fluctuation of the system.

[0220] In order to further analyze the impact of different controllable devices on the distribution network after access, Figure 8 It shows that before adding energy storage and during periods of high photovoltaic output, reverse flow will occur at multiple time points, and active power loss will increase significantly. After adding energy storage, the voltage of all nodes can be within the specified deviation allowable range, ensuring the safe operation of the distribution network and reducing network losses to a certain extent. Compared with adding only energy storage, the addition of energy storage and micro gas turbines has also greatly reduced the network losses of the system, improving the economy of system operation, and further highlighting the advantages of distribution network reconstruction and coordinated dispatch of micro gas turbines and energy storage output in improving the operating status of the distribution network.

Claims

1. A two-layer optimization dispatching method based on dynamic reconstruction and reactive power voltage adjustment of the improved whale optimization algorithm, characterized by The following steps are involved: Step 1: Establish a distribution network node system; Step 2: Construct a two-layer optimization dispatch model for active distribution network with dynamic network reconstruction and reactive power voltage adjustment: The upper optimization model aims to minimize the comprehensive operation cost of the distribution network, adopts chance-constrained programming to deal with the randomness of wind and solar power output, and uses dynamic reconstruction of chance constraints as a means to achieve economic operation of the distribution network over a long period of time. The lower optimization model aims to minimize the total deviation of system nodes, and takes static VAR compensation, micro gas turbine and energy storage output as control objects to optimize the system voltage quality. Step 3: A hybrid solution strategy combining the improved whale optimization algorithm and the interior point method is used to solve the active distribution network double-layer optimization dispatching model in step 2; In step 2, the established mathematical model for the two-layer optimization dispatch of the active distribution network is as follows: ① Randomness model of distributed wind turbine output: The relationship between fan output power and wind speed is as follows: Where: P t w is the active power output of the fan at time t, P r is the rated power of the generator; v t is the wind speed at time t, v ci is the cut-in wind speed, v r is the rated wind speed, v co is the cut-out wind speed, where A=P r / (v r -v ci ), B = -Av ci ; The expected value of the fan output power can be obtained from the above formula: Where: E(P DG,t ) is the expected value of the fan output power, f(v t ) is the wind speed probability density function; ② Photovoltaic output model: Assume that the probability density function of photovoltaic generator power output is as follows: Where: Γ is the Gamma function, P s is the output power of the solar panel, P smax is the maximum output power of the solar panel, α and β are the shape parameters of the Beta distribution, u and δ are the standard deviation and variance of the light intensity in a certain period, Γ(α+β), Γ(α), and Γ(β) are the gamma function values ​​corresponding to the shape parameters in the Beta distribution; The expected value of photovoltaic output power can be obtained: Where: f(P s ) is the power output probability density function of the photovoltaic generator, E(P s ) is the expected value of photovoltaic output power; The established two-layer optimization dispatching model of active distribution network is as follows: The objective function minf1 of the upper optimization model based on chance constraints has a probability constraint of: Pr(f(X,ξ)≤f1)≥β; Where: X is the control variable, which is the switch state of each branch; ξ is the state variable; Pr(·) represents the probability of event establishment; β is the target confidence level; f(X,ξ) is the comprehensive operating cost value of the system under state ξ, and f1 is the minimum value of f(X,ξ) when the probability level is at least β; f(X,ξ)=C Ploss +C swi ; Where: C Ploss is the network loss cost, C swi Cost of switch operation; Where: c L is the electricity price in period t; P loss,t is the equivalent network loss that meets a certain confidence level within time period t; △t = 1h; Where: Ω b is the set of power grid branches, l is a branch in the set of branches; M is the total number of sections after time segmentation, m is a single time period in the total segmentation; c act The cost of a single switch operation; l,m-1 、s l,m These are the switch states of branch l in period m-1 and period m, respectively, with open being 0 and closed being 1; Node voltage constraints: Pr(V i,min ≤V i ≤V i,max )≥β U Where: V i,min and V i,max are the upper and lower limits of node voltage respectively; β U is the voltage constraint confidence level; Pr(V i,min ≤V i ≤V i,max ) is the probability that the node voltage is qualified; Active and reactive power flow constraints: Where: P i in and are the active power and reactive power injected into node i respectively; v i 、v j are the voltage amplitudes at node i and node j respectively; N is the total number of nodes; G ij , B ij ,θ ij are the conductance, susceptance and voltage phase angle values ​​between nodes i and j, respectively, i, j = 1, 2, ..., n, where n is the total number of nodes; Branch flow constraints: Pr(|τ ij P l |≤P l,max )≥β I , Where: P l is the transmission power of line l; P l,max is the upper limit of the transmission power of line l; τ ij is the switch state of branch ij, τ ij =0 means the branch is disconnected, τ ij =1 means the branch is closed; Pr(|τ ij P l |≤P l,max ) represents the probability that the branch transmission power is qualified, β I is the line transmission power confidence level; Distributed generation (DG) output constraints: Pr(0≤P DG,i ≤P DG,i,max )≥β DG ; Where: P DG,i is the DG output value at node i; P DG,max is the upper limit of DG output at node i; Pr(0≤P DG,i ≤P DG,i,max ) is the probability that the DG output at node i is qualified; β DG Output confidence level for DG; Switching action times constraint: N i ≤N i,max ; N all ≤N all,max ; Where: N i Represents the number of operations of a single switch in a day, and the maximum number of operations is N i,max ; N all is the number of all switch operations in a day, and the maximum number of operations allowed is N all,max The objective function of the lower optimization model is: Where: T is the computing period (24 hours); n is the total number of nodes; u i,t and u i,N are the actual voltage and rated voltage of node i at time t respectively; Micro Turbine Generator (MTG) output limit: Where: P MTG,i , Q MTG,i is the active and reactive output of the micro gas turbine at node i; are the minimum active and reactive outputs of the micro gas turbine at node i respectively; are the maximum active and reactive outputs of the micro gas turbine at node i respectively; Static VAR compensation device SVC constraints: Where: Q svc,i,t is the reactive power compensation output of node i at time t; and are the upper and lower limits of SVC reactive power compensation at node i respectively; Energy storage battery operation constraints: State of charge upper and lower limits: AND SOC,min ≤E SOC,t ≤E SOC,max ; Where: E SOC,t is the remaining energy storage capacity at time t; E SOC,min and E SOC,max are the minimum and maximum states of charge, respectively; Charge and discharge power: Where: c,t The charging flag at time t is 1 when charging and 0 when not charging; λ d,t It is the discharge flag at time t, which is 1 when discharging and 0 when not discharging; Where: p c,t is the actual charging power at time t; p d,t is the actual discharge power at time t; p c,max is the maximum charging power; p d,max is the maximum discharge power.

2. The double-layer optimization dispatching method for dynamic reconstruction and reactive voltage adjustment based on the improved whale optimization algorithm according to claim 1 is characterized in that: In step 1, the parameters are initialized to establish an active distribution network topology including two distributed power sources, wind power and photovoltaic power. First, the nodes and branches are numbered, and then static VAR compensation, energy storage and micro gas turbines are connected to the designated nodes.

3. The double-layer optimization scheduling method for dynamic reconstruction and reactive voltage adjustment based on the improved whale optimization algorithm according to claim 1 is characterized in that: In step 3, the upper model solution process includes the following steps: S1: Reconstruct the whole day period into four sections: 0:00-7:00, 7:00-15:00, 15:00-21:00 and 21:00-24:

00. Set the reconstruction period s to 1, generate the initial grid population, determine whether all individuals in the population meet the radial operation conditions, if not, regenerate the individual, and set the maximum number of iterations to max_iter; S2: passing the upper grid as input to the lower optimization model, starting the lower optimization model, and calculating the upper population fitness value according to the calculation results of the lower layer and the upper objective function; S3: Update the position of the whale population and generate the offspring population; by simulating the foraging behavior of whale bubble nets, the whale optimization algorithm searches for prey through wandering, shrinking and surrounding mechanisms, and spiral surrounding mechanisms. The position of each humpback whale represents a feasible solution to the reconstruction problem. S4： ⑴ Based on the traditional whale optimization algorithm, the crossover and mutation operations of the genetic algorithm are introduced. The crossover operation is to apply the crossover operator to the group, and for the selected pairs of individuals, part of the chromosomes between them are exchanged with a probability of 0.8 to generate new individuals; the mutation operation is to apply the mutation operator to the group, and for the selected individuals, one or some genes are changed to other alleles with a probability of 0.1; (2) Introduce the adaptive weight adjustment strategy into the whale optimization algorithm to adaptively change the weight value according to the current distribution of whale populations; Where: w is the adaptive weight factor, w min 、w max are the upper and lower limits of the weight coefficient, respectively, and t is the current number of iterations, which are 1.5 and 0.4 respectively in the present invention; ⑶ The convergence factor based on the logarithmic function changes nonlinearly with the number of evolutionary iterations, and its expression is as follows: In the formula: a is the control factor; e is the natural constant; t is the current number of iterations; max_iter is the maximum number of iterations; is the component that changes logarithmically with the number of iterations, resulting in nonlinear changes in the control factor; S5: Calculate the fitness value of the updated whale population to form a new population and an old population, rearrange the fitness values ​​of all individuals in the combined population by size, retain some individuals with smaller fitness values, eliminate the remaining individuals, and update the global optimal fitness value and the optimal individual; Repeat steps S2-S4 until the maximum number of iterations is reached, and then output the optimal grid and optimal fitness value of the first reconstruction period; S6: Repeat steps S1-S6 until all grids of all reconstruction periods are output.

4. The double-layer optimization dispatching method for dynamic reconstruction and reactive voltage adjustment based on the improved whale optimization algorithm according to claim 3 is characterized by: The solution process of the lower-level optimization model is as follows: Step 1: Input the optimization results obtained by the upper optimization model, and initialize the lower optimization model population according to the upper optimization model input; Step 2: Input the distributed power and load data at the first moment of the sth reconstruction period, and calculate the fitness value of the objective function of the lower optimization model; Step 3: Use the interior point method to solve the lower optimization model, update the lower optimization model population until the iteration termination condition is met, and obtain the optimal output reference values ​​of static VAR compensation, energy storage and micro gas turbine at the first moment of the sth reconstruction period; Step 4: Repeat Step 1-Step 3 at the next moment of the s-th reconstruction period, and output the fitness value and the optimal output of the controllable device at each moment of the s-th reconstruction period; Step 5: Calculate the fitness according to the lower layer objective function and feed it back to the upper layer. This step is connected to the step 2 of the upper layer model solution process, and it is repeated in this way; Step 6: Output the switch states of all reconstruction periods that meet the constraints, the optimal output set of each hourly controllable device, and the optimal values ​​of the objective functions of the upper and lower optimization models.

Images