Power distribution network energy management method based on improved multi-target marine predator optimization control algorithm

Abstract

The invention belongs to the technical field of electric energy management, and particularly relates to a power distribution network energy management method based on an improved multi-target marine predator optimization control algorithm. According to the method, a multi-objective optimization function including comprehensive system operation cost, clean energy consumption rate, main network interaction power and the like is constructed, demand side grading response is introduced according to load characteristics and sensitivity of the load characteristics to real-time electricity price, and a power distribution network multi-objective optimization operation model is established in combination with constraint conditions of the power distribution network. A reference point selection method and a dominating power updating strategy are introduced into the marine predator algorithm, the power distribution network optimization model is solved based on the improved multi-target marine predator algorithm, and an optimal operation scheme of the power distribution network is further obtained. The power distribution network multi-target optimization model established by the invention is more effective and closer to the real situation, and the provided improved marine predator algorithm can more efficiently and accurately seek a power distribution network multi-target optimal operation scheme.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric energy management, and in particular relates to a distribution network energy management method based on an improved multi-objective marine predator optimization control algorithm. Background Art

[0002] With the continuous development of science and technology and heavy industry technology, the global energy crisis is becoming increasingly severe. At the same time, the new power supply and distribution system that integrates renewable energy generation, energy storage, and load can promote the grid connection of renewable energy generation, reduce the use of fossil energy, and improve environmental pollution. Therefore, the relevant technologies of the new power supply and distribution system have become a research hotspot. Due to the uncertainty of renewable energy, it poses a great threat to its safe and reliable power supply. Therefore, a reasonable distribution network optimization and scheduling plan has become a hot spot and difficulty in the development of the distribution network.

[0003] For the optimal dispatching of distribution network, it is necessary to first establish the optimal dispatching model of distribution network, and then solve it through traditional mathematical methods or intelligent optimization algorithms to obtain the optimal operation plan. In reality, the load peak-valley difference is large, and the power fluctuation of interaction with the large power grid is large, which threatens the security of the distribution network system. In addition, the penetration rate of renewable energy in the distribution network is increasing, and its uncertainty increases the source-load mismatch problem in the distribution network, resulting in a low renewable energy absorption rate. Existing research generally improves the renewable energy absorption rate and reduces the load peak-valley gap by reducing and transferring the demand side load. However, the sensitivity of different types of loads to time-of-use peak-valley electricity prices is insufficient, and it is not close enough to the actual load level. It is impossible to reasonably use the demand side response to solve the problem of low renewable energy absorption rate caused by source-load mismatch and large load peak-valley difference causing a large impact on the large power grid at a certain moment.

[0004] The distribution network optimization dispatch problem is a nonlinear, multi-constrained, high-dimensional problem. At present, there are generally two ways to deal with multi-objective optimization problems. One is to process the multi-objective function into a single objective function, but the result loses diversity; the other is to use a multi-objective intelligent algorithm to directly solve it. The result is a non-inferior solution set, which is further screened. For the multi-objective economic dispatch optimization problem of distribution networks, adopting a suitable multi-objective optimization algorithm can effectively improve the accuracy and speed of the solution. Summary of the invention

[0005] The purpose of the present invention is to overcome the above-mentioned problems existing in the traditional technology and to provide a distribution network energy management method based on an improved multi-objective marine predator optimization control algorithm.

[0006] In order to achieve the above technical objectives and the above technical effects, the present invention is implemented through the following technical solutions:

[0007] The present invention provides a distribution network energy management method based on an improved multi-objective marine predator optimization control algorithm, comprising the following steps:

[0008] Step 1: Construct a multi-objective optimization function including system operation cost, clean energy consumption rate and main grid interactive power fluctuation. According to the load characteristics and its sensitivity to real-time electricity prices, introduce the demand side hierarchical response, combine the distribution network constraints, and establish a multi-objective optimization operation model for the distribution network.

[0009] Step 2: Initialize the multi-objective optimization model of the distribution network, and set the population number N and the maximum number of iterations Iter_Max according to the number of dispatchable micro-sources in the distribution network;

[0010] Step 3: According to the working constraints of each micro-source in the distribution network, the population Prey is randomly and uniformly initialized, and the current iteration number Iter is set to 0;

[0011] Step 4: According to the number of objective functions in the optimization model and the number of segments on each dimension of the target, a certain number of uniformly distributed reference points are generated on the hyperplane;

[0012] Step 5: According to the set objective function, calculate the fitness value of each individual, and perform fast non-dominated sorting on the individual fitness values ​​to form a non-dominated layer. The ocean memory stores the current population to Prey_old;

[0013] Step 6: Randomly select an individual in the highest dominance layer as the elite individual, perform normal predation process calculations according to the dominance level, and obtain a new population Prey; update the population under the influence of abnormal predation vortex and FADs to obtain the population Prey_fads; randomly select the individual position of the population for Gaussian mutation to obtain the population Prey_gau;

[0014] Step 7: Merge the four populations of Prey_old, Prey, Prey_fads, and Prey_gau, remove duplicate individuals, perform fast non-dominated sorting, and select N better individuals based on the reference point to enter the next generation population;

[0015] Step 8: Determine whether the iteration stop condition is met. If yes, stop the iteration and output the Pareto optimal solution set and the corresponding Pareto optimal frontier; otherwise, add 1 to the number of iterations and go to step 5 to continue the iterative calculation;

[0016] Step 9: Further process the Pareto solution set, calculate the satisfaction value of the solution according to the fuzzy membership method, and select the best distribution network operation plan.

[0017] Furthermore, in step 1, the multi-objective optimization model of the distribution network takes the minimum total operating cost of the distribution network, the maximum renewable energy consumption rate, and the optimal interconnection line power fluctuation as the optimization objectives.

[0018] Furthermore, the economic target is represented by the total operating cost of the distribution network, which includes the operation and maintenance costs of each micro-source, the load transfer compensation fee, and the cost of purchasing and selling electricity to the grid; the specific mathematical expression is as follows:

[0019]

[0020] Among them, C f (t) represents the fuel consumption cost of each distributed generation at time t, C OM (t) represents the maintenance cost of each power generation unit at time t; C L (t) Compensation for power outages caused by load shedding; C grid Represents the electricity transaction fee between the distribution network and the main grid.

[0021] Furthermore, the environmental protection index is expressed by the wind and solar power abandonment rate. Reducing the wind and solar power abandonment rate will help solve the problem of wind and solar power consumption and improve the utilization rate of clean energy. The specific mathematical expression is as follows:

[0022]

[0023] Among them, T is the scheduling period, which is 24 hours a day; P pv (t) and P wt (t) are the actual power generated in one day; P p ' v (t) and P w ' t (t) are the power consumed by renewable energy in a day.

[0024] Furthermore, the safety index is reflected in the interactive power fluctuation between the Zhu power grid and the distribution network system connection line; the specific mathematical expression is as follows:

[0025]

[0026] Furthermore, in the optimization process, it is necessary to constrain the power balance and the upper and lower limits of the micro-source output. The constraints include:

[0027] P pv (t)+P WT (t)+P BA (t)+P MT (t)+P grid (t) = P load (t)

[0028]

[0029] SoC max ≤SoC(t)≤SoC min

[0030]

[0031] The above formulas respectively represent the power balance constraint, the micro gas turbine ramp constraint, the micro gas turbine output constraint, the energy storage SoC constraint, the energy storage output constraint and the interconnection line power constraint.

[0032] Furthermore, in step 1, the demand-side graded response is: according to the degree of response of different types of loads to real-time electricity price changes, the loads are divided into three levels: A, B and C; among them, A is an easily transferable load that is sensitive to electricity price changes, and the mathematical model of this stage can be reflected by the load transfer rate function; B is a transferable load with elastic adjustment, and the mathematical model of this stage can be described as an electricity price elasticity coefficient matrix; C is a non-transferable load with rigid demand, which is not affected by electricity prices.

[0033] Furthermore, the mathematical model of level a is:

[0034]

[0035] Among them, Δp is the absolute value of the change in electricity price; f max is the maximum load transfer rate in the saturated zone; a is the dead zone threshold; f max / K+a is the inflection point in the saturation region, and K is the slope of the transfer rate curve in the linear region;

[0036] Assume that the time period of the day when electricity prices rise is I = {i1, i2, ..., i m}, the period of time when the electricity price drops is J = {j1,j2,…,j n}; The load transfer in each period can be described as:

[0037]

[0038] Where t1, t4∈I, t2, t3∈J; Δp(t) represents the difference between the real-time electricity price and the basic electricity price during period t; ΔP L (i,j) indicates the load transfer from period i to period j;

[0039] The load level after the a-level load demand response is:

[0040] P L ' a (t) = P La (t)-ΔP L (i,j)

[0041] Furthermore, the mathematical model of level b is:

[0042] d′=d+λ×d

[0043] Vectors d and d′ respectively describe the load power consumption before and after the user performs demand response;

[0044]

[0045] ΔL i (m) = L i (m)-L i0 (m)

[0046] ΔC i (n) = C i (n)-C i0 (n)

[0047]

[0048] Among them, λ mn is the demand elasticity coefficient of user i. When time m is equal to n, it is the self-elasticity coefficient. Since the user demand will decrease accordingly when the electricity price rises, it is a negative value. When time m is not equal to n, it is the cross-elasticity coefficient, which describes that when the electricity price decreases at time n, the user will transfer part of the load at time m to time n. ΔL i (m) is the change in charge; ΔC i (n) is the change in electricity price; L i0 It is the amount of electricity before the implementation of time-of-use electricity price; C i0 (n) is the electricity price before the implementation of time-of-use electricity price; L i (m) is the amount of electricity after the time-of-use electricity price is implemented; C i (n) The electricity price after the implementation of time-of-use electricity price;

[0049] The load in each period after implementing the hierarchical demand response model is:

[0050] P l ' oad (t) = P L ' a (t)+P L ' b (t)+P Lc (t)

[0051] Furthermore, in step 6, the calculation process of predation according to the dominance level is as follows:

[0052] First, set the maximum value LC and minimum value UC of the dominant solution; when the population is updated, it is updated according to the dominant solution of the current individual; if the dominant solution of the current individual R i If it is less than LC, the update rule is:

[0053]

[0054] in, is the update step size, and are the elite matrix and the population matrix respectively, represents the Brownian vector, P is a constant, which is 0.5 in this algorithm; if the R of the current solution i If it is greater than UC, the update rule is:

[0055]

[0056] in, represents the random vector of Levy motion, CF represents the adaptive parameter controlling the predator's moving step; similarly, the current individual's R i If it is greater than UC and less than LC, the update rule is:

[0057] The first part of the population, i = 1, 2, 3, ... n / 2

[0058]

[0059] The second part of the population, i = n / 2, ... n

[0060]

[0061]

[0062] The beneficial effects of the present invention are:

[0063] 1. The present invention proposes a hierarchical demand response model. Taking full account of load characteristics and the response to real-time electricity prices, a response model of different levels of load to changes in real-time electricity prices is established. The model is closer to the actual load level and effectively improves problems such as large load peak-valley gap, source-load mismatch, and low renewable energy consumption rate.

[0064] 2. The present invention proposes an improved multi-objective marine predator algorithm for distribution network operation optimization. The Pareto dominance idea and the reference point-based selection method are introduced to expand the marine predator algorithm to the multi-objective field. At the same time, the marine predator algorithm is improved based on the dominance update, which can effectively improve the ability of the population to jump out of the local optimum and avoid falling into the local optimum. The Gaussian mutation operator is used to focus on searching the local area near the individual, which can effectively improve the solution efficiency of the algorithm.

[0065] Of course, any product implementing the present invention does not necessarily need to achieve all of the above advantages at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0067] Figure 1 is the algorithm flow chart;

[0068] Figure 2 Generate schematic diagrams for reference points;

[0069] Figure 3 It is a schematic diagram of individual association;

[0070] Figure 4 The standard test function results

[0071] Figure 5 Run an optimal solution study for a distribution network. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] This embodiment provides a distribution network energy management method based on an improved multi-objective marine predator optimization control algorithm, comprising the following steps:

[0074] Step 1: Construct a multi-objective optimization function including system operation cost, clean energy consumption rate and main grid interactive power fluctuation, introduce demand-side hierarchical demand response, add constraints, and establish a distribution network multi-objective optimization operation mathematical model;

[0075] Step 2: Initialize the multi-objective optimization model of the distribution network, and set the population number N and the maximum number of iterations Iter_Max according to the number of dispatchable micro-sources in the distribution network;

[0076] Step 3: According to the working constraints of each micro-source in the distribution network, the population Prey is initialized using uniform random distribution, and the current number of iterations Iter is set to 0;

[0077] Step 4: Generate a certain number of evenly distributed reference points according to the number of objective functions in the optimization model and the number of segments on each dimension of the target;

[0078] Step 5: According to the set objective function, calculate the fitness value of each individual, and perform fast non-dominated sorting on the individual fitness values ​​to form a non-dominated layer. The ocean memory stores the current population to Prey_old;

[0079] Step 6: Randomly select an individual in the highest dominance layer as an elite individual, and perform normal predation according to the dominance level to obtain a new population Prey. The population is updated under the influence of abnormal predation vortices and FADs to obtain the population Prey_fads. Randomly select the individual positions of the population for Gaussian mutation to obtain the population Prey_gau;

[0080] Step 7: Merge the four populations Prey_old, Prey, Prey_fads, and Prey_gau. Remove duplicate individuals, perform fast non-dominated sorting, and then select the best N individuals to enter the next generation population based on the reference point selection operation;

[0081] Step 8: If the iteration stop condition is met, the iteration is terminated and the Pareto optimal solution set and the corresponding Pareto optimal frontier are output. Otherwise, Iter = Iter + 1, go to step 5, and iterate the calculation in a loop;

[0082] Step 9: Further process the Pareto solution set, calculate the satisfaction value of the solution according to the fuzzy membership method, and select the best distribution network operation plan.

[0083] To achieve the above steps, it is necessary to clarify the multi-objective optimization model of the distribution network and the multi-objective marine predator algorithm. The following is a detailed description:

[0084] 1. Distribution Network Optimization Model

[0085] Common distribution network systems include wind power generation (WT) systems, photovoltaic power generation (PV) systems, micro gas turbines (MT), and energy storage systems (BA). This embodiment takes into account the three aspects of economy, environmental protection, and safety, and establishes a multi-objective optimization model for the distribution network with the minimum total operating cost of the distribution network, the highest renewable energy consumption rate, and the best interconnection line power fluctuation as the optimization goals. The economic target is represented by the total operating cost of the distribution network, which includes the operation and maintenance costs of each micro-source, load transfer compensation costs, and the cost of purchasing and selling electricity to the power grid. Its mathematical expression is as follows:

[0086]

[0087] C f (t) represents the fuel consumption cost of each distributed generation at time t, C OM (t) represents the maintenance cost of each power generation unit at time t; C L (t) Compensation for power outages caused by load shedding; Cgrid Represents the electricity transaction fee between the distribution network and the main grid.

[0088]

[0089] Environmental protection indicators are expressed by the wind and solar power abandonment rate. Reducing the wind and solar power abandonment rate will help solve the problem of wind and solar power consumption and improve the utilization rate of clean energy. T is the scheduling period, which is 24 hours a day; P pv (t) and P wt (t) are the actual power generated in one day; P p ' v (t) and P w ' t (t) are the power consumed by renewable energy in a day.

[0090]

[0091] The safety index is reflected in the interactive power fluctuations between the Zhu power grid and the distribution network system connection lines.

[0092] During the optimization process, power balance and upper and lower limits of micro-source output need to be constrained. The constraints include:

[0093] P pv (t)+P WT (t)+P BA (t)+P MT (t)+P grid (t) = P load (t)

[0094]

[0095] SoC max ≤SoC(t)≤SoC min

[0096]

[0097] The above respectively represent power balance constraint, micro gas turbine ramp constraint, micro gas turbine output constraint, energy storage SoC constraint, energy storage output constraint, and interconnection line power constraint.

[0098] 2. Hierarchical Demand Response

[0099] Taking into full consideration the response of different types of loads to real-time electricity price changes, the loads are divided into three levels: a, b, and c according to easy transfer, transferable, and non-transferable loads. Level a is easy transfer load, which is more sensitive to electricity price changes. Its mathematical model can be reflected by the transfer rate function:

[0100]

[0101] Where: Δp is the absolute value of the change in electricity price; f max is the maximum load transfer rate in the saturated zone; a is the dead zone threshold; f max / K+a is the inflection point in the saturation region, and K is the slope of the transfer rate curve in the linear region. Assume that the period of rising electricity prices in a day is I = {i1, i2, ..., i m}, the period of time when the electricity price drops is J = {j1,j2,…,j n}. The load transfer in each period can be described as:

[0102]

[0103] Where: t1, t4∈I, t2, t3∈J; Δp(t) represents the difference between the real-time electricity price and the basic electricity price during period t; ΔP L (i,j) indicates the load transfer from period i to period j; therefore, the load level after the a-level load demand response is:

[0104] P L ' a (t) = P La (t)-ΔP L (i,j)

[0105] The transferable load is class B load, and the commonly used elastic price matrix is ​​used to characterize the demand-side response. The adjusted electricity consumption can be described as:

[0106] d′=d+λ×d

[0107] Vectors d and d′ respectively describe the load power consumption before and after the user performs demand response.

[0108]

[0109] ΔL i (m) = L i (m)-L i0 (m)

[0110] ΔC i (n) = C i (n)-C i0 (n)

[0111]

[0112] Where: mn is the demand elasticity coefficient of user i. When time m is equal to n, it is the self-elasticity coefficient. Since the user demand will decrease accordingly when the electricity price rises, it is a negative value. When time m is not equal to n, it is the cross-elasticity coefficient, which describes that when the electricity price decreases at time n, the user will transfer part of the load at time m to time n. ΔL i(m) is the change in charge; ΔC i (n) is the change in electricity price; L i0 It is the amount of electricity before the implementation of time-of-use electricity price; C i0 (n) is the electricity price before the implementation of time-of-use electricity price; L i (m) is the amount of electricity after the time-of-use electricity price is implemented; C i (n) is the electricity price after the implementation of time-of-use electricity price.

[0113] In summary, the load in each period after implementing the hierarchical demand response model is:

[0114] P l ' oad (t) = P L ' a (t)+P L ' b (t)+P Lc (t)

[0115] Class C load has rigid demand, and the load remains almost unchanged in different time periods.

[0116] 3. Multi-objective marine predator algorithm

[0117] The multi-objective marine predator algorithm proposed in this embodiment applies the Pareto dominance concept to the MPA algorithm and expands it to the multi-objective field. With reference to the elite retention strategy of NSGA-Ⅲ, the reference point-based selection method is applied. The basic marine predator algorithm is improved based on the update of dominance, which effectively improves the ability of individuals to jump out of the local optimum and avoid falling into the local optimum. At the same time, the introduction of the Gaussian mutation operator to focus on the search of the local area near the individual can effectively improve the solution efficiency of the algorithm. The algorithm steps are as follows: Figure 1 , as follows:

[0118] Initialize according to the multi-objective optimization model of the distribution network, and determine the number of populations N and the maximum number of iterations Iter_Max by the number of dispatchable micro-sources in the distribution network;

[0119] To meet the output constraints of each micro-source in the distribution network, the population Prey is initialized using uniform random distribution, and the current iteration number Iter is set to 0; the initialization formula is:

[0120] x0=x min +rand*(x max -x min )

[0121] x0 is the initial position, x max and x min are the upper and lower limits of the search space respectively; rand is a random number vector between 0 and 1.

[0122] According to the number of objective functions in the optimization model and the number of segments on each dimension of the objective, uniformly distributed reference points are generated; generally, the reference points are directly generated in the form of equal distribution, or the desired reference points are set according to the decision maker's preferences. The purpose is to pre-set uniformly distributed reference points in the target space to guide the evolution of individuals. This embodiment adopts the system method proposed in the article "Sampling Reference Points on the Pareto Fronts of Multi-Objective Optimization Problems" for structured generation. For M objective functions with equal scores p, H reference points can be obtained, and the size of H can be obtained by the following formula.

[0123]

[0124] For example, a three-objective function divided into four equal parts can evenly generate 15 reference points, generating a schematic diagram as follows Figure 2 shown.

[0125] According to the set distribution network objective function, the fitness value of each solution is calculated, and the individual fitness values ​​are quickly non-dominated to form a non-dominated layer. The ocean memory stores the current population to Prey_old;

[0126] Randomly select an individual from the highest dominance layer as an elite individual, and perform normal predation according to the dominance level to obtain a new population Prey. The population is updated under the influence of abnormal predation vortex and FADs to obtain the population Prey_fads. Randomly select the individual position of the population for Gaussian mutation to obtain the population Prey_gau; the dominance level update process is as follows:

[0127] First, set the maximum value LC and minimum value UC of the dominant solution. When the population is updated, it is updated according to the dominant solution of the current individual. If the dominant solution R of the individual i If it is less than LC, the update rule is:

[0128]

[0129] In the above formula, is the update step size, and are the elite matrix and the population matrix respectively, represents the Brownian vector, P is a constant, which is 0.5 in this algorithm. i If it is greater than UC, the update rule is:

[0130]

[0131] In the above formula, represents the random vector of Levy motion, and CF represents the adaptive parameter that controls the predator’s moving step. Similarly, the R i If it is greater than UC and less than LC, the update rule is:

[0132] The first part of the population, i = 1, 2, 3, ... n / 2

[0133]

[0134] The second part of the population, i = n / 2, ... n

[0135]

[0136] Merge the four populations Prey_old, Prey, Prey_fads, and Prey_gau. Remove duplicate populations, perform fast non-dominated sorting to obtain the dominance level of each individual, and then perform a reference point-based selection operation to select the better N individuals to enter the next generation population;

[0137] The reference point-based selection operation includes normalization of individuals in the population, association of reference points, and retention of microhabitats. First, since the reference point-based individual selection method is performed on a normalized hyperplane, we need to transform the objective function of the population into the normalized target space. Then, before the reference point-based selection process is performed, the individuals in the population need to be associated. In the hyperplane, the line connecting each reference point and the origin of the coordinate system corresponds to the reference line of the point. The vertical distance from each individual in the selected population to all reference lines is calculated, and the reference point corresponding to the reference line with the shortest distance is the reference point associated with the individual. After completing the population individual association operation, there may be only one individual associated with a certain reference point, or there may be multiple individuals associated with it, or there may be no individual associated with the reference point. Finally, the target individuals are selected through the microhabitat operation to enter the next generation. The individual association reference point operation is as follows: Figure 3 shown.

[0138] If the maximum number of iterations is reached, the iteration is stopped and the Pareto optimal solution set and the corresponding Pareto optimal frontier are output. Otherwise, Iter = Iter + 1, and go to step 2.5.

[0139] Further process the Pareto solution set, calculate the satisfaction value of the solution according to the fuzzy membership function, and select the best distribution network operation plan. k , the satisfaction of its ith sub-objective function It can be expressed as follows:

[0140]

[0141] are the maximum and minimum values ​​of the ith sub-objective function respectively.

[0142] For a non-inferior solution x k , and its comprehensive satisfaction is expressed as:

[0143]

[0144] M and N represent the number of non-inferior solutions and objective functions, respectively. k The larger the value of , the higher the comprehensive satisfaction of this non-inferior solution, and the better the non-inferior solution.

[0145] Figure 4 The results of running the algorithm of this embodiment in a standard multi-objective test function are shown. The proposed algorithm can completely cover the real Pareto optimal frontier of the test function and is evenly distributed on the real Pareto optimal frontier. It can be concluded that the multi-objective marine predator algorithm proposed in this paper can correctly search for the Pareto optimal frontier of the test function, and the obtained solution is evenly distributed on the Pareto optimal frontier of the test function, has good convergence and covers a wide range.

[0146] Figure 5 This is an example of the optimal operation scheme of a distribution network using an improved multi-objective marine predator algorithm. In the photovoltaic output range, most of the energy in the system is provided by photovoltaic units. If the photovoltaic power is in excess, the energy storage is charged first. In the 10:00-15:00 period in the figure, part of the load in this stage is transferred to the 0:00-5:00 period according to the graded response of demand-side users. Therefore, in the 8:00-15:00 period, photovoltaic output can supply most of the user loads, greatly improving the local consumption rate of renewable energy. In addition, part of the load during the peak period of 18:00-22:00 is also transferred to the valley period, which reduces the power fluctuation of the interconnection line, improves the stability of the microgrid system, and also reduces the overall operating cost of the distribution network.

[0147] In summary, this embodiment provides an improved multi-objective marine predator algorithm for optimizing the operation of distribution networks. First, a multi-objective optimization model of distribution networks that considers hierarchical demand response and takes into account economy, environmental protection, and safety is established. Secondly, the marine predator algorithm is extended to the multi-objective field by using Pareto dominance, and the marine predator algorithm is improved based on dominance update, which can effectively improve the ability of the population to jump out of the local optimum and avoid falling into the local optimum; and the Gaussian mutation operator is used to focus on searching the local area near the individual to maintain species diversity, which can effectively improve the algorithm's solution efficiency. Finally, the algorithm of this embodiment is used to solve the model, thereby achieving economic, environmental, safe and stable operation of the distribution network.

[0148] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A distribution network energy management method based on an improved multi-objective marine predator optimization control algorithm, characterized in that: The following steps are involved: Step 1: Construct a multi-objective optimization function including system operation cost, clean energy consumption rate and main grid interactive power fluctuation. According to the load characteristics and its sensitivity to real-time electricity prices, introduce the demand side hierarchical response, combine the distribution network constraints, and establish a multi-objective optimization operation model for the distribution network. Step 2: Initialize the multi-objective optimization model of the distribution network, and set the population number N and the maximum number of iterations Iter_Max according to the number of dispatchable micro-sources in the distribution network; Step 3: According to the working constraints of each micro-source in the distribution network, the population Prey is randomly and uniformly initialized, and the current iteration number Iter is set to 0; Step 4: According to the number of objective functions in the optimization model and the number of segments on each dimension of the target, a certain number of uniformly distributed reference points are generated on the hyperplane; Step 5: According to the set objective function, calculate the fitness value of each individual, and perform fast non-dominated sorting on the individual fitness values ​​to form a non-dominated layer. The ocean memory stores the current population to Prey_old; Step 6: Randomly select an individual in the highest dominance layer as the elite individual, perform normal predation process calculations according to the dominance level, and obtain a new population Prey; update the population under the influence of abnormal predation vortex and FADs to obtain the population Prey_fads; randomly select the individual position of the population for Gaussian mutation to obtain the population Prey_gau; Step 7: Merge the four populations of Prey_old, Prey, Prey_fads, and Prey_gau, remove duplicate individuals, perform fast non-dominated sorting, and select N better individuals based on the reference point to enter the next generation population; Step 8: Determine whether the iteration stop condition is met. If yes, stop the iteration and output the Pareto optimal solution set and the corresponding Pareto optimal frontier; otherwise, add 1 to the number of iterations and go to step 5 to continue the iterative calculation; Step 9: Further process the Pareto solution set, calculate the satisfaction value of the solution according to the fuzzy membership method, and select the best distribution network operation plan.

2. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 1 is characterized in that: In step 1, the multi-objective optimization model of the distribution network takes the minimum total operating cost of the distribution network, the maximum renewable energy consumption rate, and the optimal interconnection line power fluctuation as the optimization objectives.

3. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 2 is characterized in that: The economic target is represented by the total operating cost of the distribution network, which includes the operation and maintenance costs of each micro-source, load transfer compensation costs, and the cost of purchasing and selling electricity to the grid; the specific mathematical expression is as follows: Among them, C f (t) represents the fuel consumption cost of each distributed generation at time t, C OM (t) represents the maintenance cost of each power generation unit at time t; C L (t) Compensation for power outages caused by load shedding; C grid Represents the electricity transaction fee between the distribution network and the main grid.

4. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 3 is characterized in that: Environmental protection indicators are expressed by the wind and solar power abandonment rate. Reducing the wind and solar power abandonment rate will help solve the problem of wind and solar power consumption and increase the utilization rate of clean energy. The specific mathematical expression is as follows: Among them, T is the scheduling period, which is 24 hours a day; P pv (t) and P wt (t) are the actual power generated in one day; P p ' v (t) and P w ' t (t) are the power consumed by renewable energy in a day.

5. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 4 is characterized in that: The safety index is reflected in the interactive power fluctuation between the Zhu power grid and the distribution network system connection line; the specific mathematical expression is as follows:

6. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 5 is characterized in that: During the optimization process, power balance and upper and lower limits of micro-source output need to be constrained. The constraints include: P pv (t)+P WT (t)+P BA (t)+P MT (t)+P grid (t)=P load (t) SoC max ≤SoC(t)≤SoC min The above formulas respectively represent the power balance constraint, the micro gas turbine ramp constraint, the micro gas turbine output constraint, the energy storage SoC constraint, the energy storage output constraint and the interconnection line power constraint.

7. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 1 is characterized in that: In step 1, the demand-side graded response is as follows: according to the degree of response of different types of loads to real-time electricity price changes, the loads are divided into three levels: A, B and C; among them, A is the easily transferable load that is sensitive to electricity price changes, and the mathematical model of this stage can be reflected by the load transfer rate function; B is the elastically adjusted transferable load, and the mathematical model of this stage can be described as the electricity price elasticity coefficient matrix; C is the non-transferable load with rigid demand, which is not affected by electricity prices.

8. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 7 is characterized in that: The mathematical model of level A is: Among them, Δp is the absolute value of the change in electricity price; f max is the maximum load transfer rate in the saturated zone; a is the dead zone threshold; f max / K+a is the inflection point in the saturation region, and K is the slope of the transfer rate curve in the linear region; Assume that the time period of the day when electricity prices rise is I = {i1, i2, ..., i m }, the period of time when the electricity price drops is J = {j1,j2,…,j n }; The load transfer in each period can be described as: Where t1, t4∈I, t2, t3∈J; Δp(t) represents the difference between the real-time electricity price and the basic electricity price during period t; ΔP L (i,j) indicates the load transfer from period i to period j; The load level after class a load demand response is: P L ' a (t)=P La (t)-ΔP L (i,j) 9. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 8 is characterized in that: The mathematical model of level b is: d′=d+λ×d Vectors d and d′ respectively describe the load power consumption before and after the user performs demand response; ΔL i (m)=L i (m)-L i0 (m) ΔC i (n)=C i (n)-C i0 (n) Among them, λ mn is the demand elasticity coefficient of user i. When time m is equal to n, it is the self-elasticity coefficient. Since the user demand will decrease accordingly when the electricity price rises, it is a negative value. When time m is not equal to n, it is the cross-elasticity coefficient, which describes that when the electricity price decreases at time n, the user will transfer part of the load at time m to time n. ΔL i is the change in charge; ΔC i (n) is the change in electricity price; L i0 (m) is the amount of electricity before the implementation of time-of-use electricity prices; C i0 (n) is the electricity price before the implementation of time-of-use electricity price; L i (m) is the amount of electricity after the time-of-use electricity price is implemented; C i (n) The electricity price after the implementation of time-of-use electricity price; The load in each period after implementing the hierarchical demand response model is: P l ' oad (t)=P L ' a (t)+P L ' b (t)+P Lc (t) 10. The distribution network energy management method based on the improved multi-objective marine predator optimization control algorithm according to claim 1 is characterized in that: In step 6, the calculation process of predation based on dominance level is as follows: First, set the maximum value LC and minimum value UC of the dominant solution; when the population is updated, it is updated according to the dominant solution of the current individual; if the dominant solution of the current individual R i If it is less than LC, the update rule is: in, is the update step size, and are the elite matrix and the population matrix respectively, represents the Brownian vector, P is a constant, which is 0.5 in this algorithm; if the R of the current solution i If it is greater than UC, the update rule is: in, represents the random vector of Levy motion, CF represents the adaptive parameter controlling the predator's moving step; similarly, the current individual's R i If it is greater than UC and less than LC, the update rule is: The first part of the population, i = 1, 2, 3, ... n / 2 The second part of the population, i = n / 2, ... n

Images