A multi-objective reservoir group scheduling cooperative search method and system

Abstract

The application discloses a kind of reservoir group multi-objective scheduling cooperation search method and system, comprising the following steps: ① based on the basic information of reservoir establishes cascade reservoir multi-objective scheduling model;② constraint condition is added to model;③ construct a kind of reservoir group multi-objective scheduling cooperation search method and system solve reservoir optimal scheduling scheme, by using team communication strategy, internal competition strategy in the whole solution space is widely searched, simultaneously by self-learning strategy and self-adapting jump out local optimum strategy jump out local optimum, realize the dynamic balance of global exploration and local exploration.The application can effectively improve reservoir group scheduling calculation efficiency, quickly give reservoir group scheduling scheme set, with less calculation parameters, convergence speed is fast, global optimization ability is strong and not easy to fall into local optimum and the like advantage, while avoiding complex parameter adjustment process, programming is simple to realize, widely applicable scene, for complex reservoir group multi-objective engineering problem solution provides more effective technical method.

Description

Technical Field

[0001] This invention belongs to the field of reservoir optimization scheduling technology, specifically relating to a multi-objective scheduling cooperative search method and system for reservoir groups. Background Technology

[0002] Optimal scheduling of reservoir groups is a typical large-scale, multi-dimensional, multi-objective, multi-stage stochastic nonlinear dynamic programming problem. It is based on systems engineering theory and uses modern computer technology and optimization techniques to seek satisfactory scheduling planning methods, scheduling plans, and real-time scheduling schemes that meet the scheduling principles to guide reservoir operation.

[0003] Because reservoir scheduling is closely related to many complex factors such as society, economy, nature, environment, and engineering, and because there are constraints such as response time differences between the states and decision variables of cascade power stations, the joint scheduling of reservoir group hydropower systems has become a rather thorny optimization problem, and has been a research hotspot for many water conservancy experts and scholars for many years.

[0004] Currently, the main approaches to addressing the multi-objective problem of reservoirs fall into three categories:

[0005] The first type of method is the prior method, which is based on the decision-maker's preference structure. It uses weights and priorities to represent decision preferences, directly transforming the multi-objective optimization problem into a single-objective optimization problem for solution. The second type of method is the posterior method, also known as the non-dominated solution generation and optimization method. This mainly involves generating a set of non-dominated solutions, then evaluating and optimizing this set to find the relatively optimal solution. This includes posterior methods based on mathematical programming, such as the orthogonal boundary crossover algorithm and the continuous Pareto optimization algorithm, as well as multi-objective evolutionary algorithms, such as NSGA-II, NSGA-III, and MOPSO. The third type of method is the interactive solution method, which relies on the continuous interaction between the decision-maker and the optimization model to find the relatively optimal solution.

[0006] Prior methods and interactive solving methods require decision-makers to provide preference information before or during optimization, lacking flexibility and feasibility. Furthermore, the optimal solution determined in prior methods may not ultimately reflect the decision-maker's true preferences. In contrast, posterior methods involve independent processes for generating non-dominated solutions and making decision-making. They can provide decision-makers with a set of solutions to choose from, making them more valuable in engineering applications. Summary of the Invention

[0007] The purpose of this invention is to propose a cooperative search method and system for multi-objective scheduling of reservoir groups, in order to solve the problems of lack of flexibility and feasibility in the optimal scheduling of reservoir groups in the prior art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a multi-objective scheduling cooperative search method for a group of reservoirs, comprising the following steps:

[0009] S1. Based on basic reservoir information data, construct an optimal scheduling model for a cascade reservoir group with the objectives of maximizing power generation and maximizing minimum output. At the same time, add constraints to the optimal scheduling model for the cascade reservoir group; preset the maximum number of iterations, population size, dominant population size, and determine the outflow from the reservoir during the scheduling period as the decision variable.

[0010] S2. A multi-objective cooperative search method for reservoir group scheduling is adopted to solve for the optimal reservoir scheduling scheme: specifically as follows:

[0011] S201. Initialization: Generate an initial population based on the preset range of decision variables, store the individuals in the initial population as their optimal positions, calculate the fitness of individuals in the population, and then obtain the dominant population and the optimal individuals in the dominant population based on the fast non-dominated sorting method.

[0012] S202, Execution Team Communication Strategy: Based on the dominant population, optimal individual, and optimal individual position, generate communication individuals using random communication methods;

[0013] S203. Implement self-learning strategies: Learn based on the upper and lower limits of communication individuals and decision variables to generate learning individuals;

[0014] S204. Implement internal competition strategies: determine offspring by using dynamic selection weights based on dominance relationships among communicating and learning individuals;

[0015] S205. Update the dominant population and the optimal position of individuals: Calculate the fitness of the current population and the dominant population, obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, update the dominant population, then use the conventional update method to determine the optimal position of individuals, and then use the adaptive escape from local optima strategy to dynamically replace the optimal position of individuals.

[0016] S206. Perform iterative updates. If the preset maximum number of iterations is reached, the optimal reservoir scheduling scheme is obtained, and the fitness value of the dominant population is output. Otherwise, return to step S202.

[0017] Furthermore, in step S1 above, the objective function for maximizing power generation in the cascade reservoir group optimal scheduling model is:

[0018]

[0019] Where E is the total power generation of all reservoirs during the time period; K is the number of reservoirs; J is the total number of time periods; P k,j Q k,j H k,j These represent the power output, power generation flow, and head of the k-th reservoir in the j-th time period, respectively; A kt is the output coefficient of the k-th reservoir; j Let j be the number of hours in the j-th time period;

[0020] Minimum output maximum objective function:

[0021]

[0022] Where F represents the minimum output of all reservoirs during the time period.

[0023] Furthermore, the aforementioned multi-objective cooperative search method for reservoir group scheduling adds the following constraints to the optimal scheduling model of the cascade reservoir group:

[0024] Output constraints

[0025]

[0026] in, These are the minimum and maximum outputs of the k-th reservoir during the j-th time period, respectively.

[0027] Storage capacity constraints

[0028]

[0029] in, These are the minimum and maximum reservoir capacities for the k-th reservoir during the j-th time period, respectively.

[0030] Water balance constraints

[0031]

[0032] Among them, I k,j O k,j These represent the inflow and outflow of the k-th reservoir during the j-th time period, respectively. Remove the lost water volume from the outflow from the upstream reservoir;

[0033] Outbound flow constraints

[0034]

[0035] in, Let $\frac{1}{2}$ be the minimum and maximum outflow from the $k$ reservoir during the $j$ time period, respectively.

[0036] Power generation flow constraints

[0037]

[0038] in, These are the minimum and maximum power generation flows of the k-th reservoir during the j-th time period, respectively.

[0039] Initial and final storage capacity constraints

[0040]

[0041] in, These are the initial and final storage capacities of the k-th reservoir, respectively.

[0042] Furthermore, the aforementioned step S201 includes the following sub-steps:

[0043] S201-1. Randomly generate an initial population within the preset range of decision variables, store the individuals in the initial population as their optimal position pbest, and calculate their fitness; the decision variables are as follows:

[0044]

[0045] Where, x i,j Let j be the value of the decision variable for the i-th individual; Indicates the interval A function to generate uniformly distributed random numbers; N is the population size; D is the number of decision variables;

[0046] S201-2. Obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, store the non-dominated front individuals of the feasible solution set in the dominant population gbest, and if there is no feasible solution, store the non-dominated front individuals with the smallest constraint violation value among the infeasible solutions in the dominant population size gbest.

[0047] S2201-3. In the dominant population gbest, individuals are sorted in descending order according to their crowding distance. One individual is randomly selected from the top K individuals and stored in the optimal individual gbest. ind If gbest does not store K individuals, then randomly select an individual from the dominant population gbest and store it in the optimal individual gbest. ind middle.

[0048] Furthermore, the aforementioned step S202, the execution team communication strategy, specifically involves: using the optimal population (gbest) and the optimal individual (gbest) as the means. ind The communication individual u is generated from three aspects: the individual's optimal position (pbest), and two random communication methods, as shown in the following formula:

[0049]

[0050] A i,j =log(1 / φ(0,1))·(gbest ind,j -x i,j (33)

[0051] B i,j =α·φ(0,1)·(gbestm,j -x i,j (34)

[0052]

[0053] Among them, u i,j A is the value of the j-th decision variable for the i-th individual, generated through communication strategies; i,j Indicates from gbest ind,j Updated information at location B; i,j This represents the information updated from gbest; m is the randomly selected individual index; C i,j This represents the information updated from pbest; α and β are the adjustments made to B, respectively. i,j and C i,j Adjustment factor for the degree of influence.

[0054] Furthermore, the aforementioned step S203, implementing the self-learning strategy, specifically involves generating a learning individual v by learning the range between the current individual and the upper and lower limits of the decision variables, as shown in the following formula:

[0055]

[0056]

[0057]

[0058] s i,j =x i,j +d i,j ·φ(-1,1) (39)

[0059]

[0060]

[0061] Among them, v i,j r is the value of the j-th decision variable for the i-th individual generated through a self-learning strategy; i,j p i,j and s i,j It refers to a self-learning individual generated based on self-learning strategies; d i,j It is the dynamic range of change generated based on the upper and lower limits of the decision variables; t and T are the current iteration number and the maximum iteration number of the algorithm, respectively; c j This represents the mean of the upper and lower limits of the decision variable.

[0062] Furthermore, the aforementioned step S204, which implements an internal competition strategy, specifically involves determining offspring based on dominance relationships between the communicating individual u and the self-learning individual v. When u and v do not dominate each other, dynamic selection weights are used to determine offspring, as shown in the following formula:

[0063]

[0064] Where, x i For the final determined offspring; u i and v i γ represents the individuals generated by the communication strategy and the self-learning strategy; γ is the balance coefficient between the communication strategy and the selection strategy; t and T are the current iteration number and the maximum iteration number of the algorithm, respectively.

[0065] Furthermore, the aforementioned step S205 includes the following sub-steps:

[0066] S205-1. Calculate the fitness of the current population and the dominant population, and obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm. If the number of non-dominated fronts of the feasible solution set does not exceed the number of dominant populations, store the non-dominated fronts of the feasible solution set in the dominant population gbest. Otherwise, calculate the crowding distance value of all individuals, sort the individuals in descending order according to the crowding distance value, and remove individuals that exceed the number of dominant populations according to the sorted sequence number.

[0067] S205-2. The optimal position of an individual is determined using the conventional update method, as shown in the following formula:

[0068]

[0069] Among them, pbest i For the individual's optimal position; x i For the current individual;

[0070] S205-3. Adopt an adaptive strategy to escape local optima and dynamically replace the individual's optimal position as follows:

[0071]

[0072] Where, m q It randomly selects q individuals from the population N to replace the current individual's optimal position; floor is the floor function.

[0073] Another aspect of this invention proposes a multi-objective scheduling cooperative search system for a group of reservoirs, comprising:

[0074] The model building module is configured to perform the following actions: Based on the basic information data of the reservoirs, construct an optimal scheduling model for a cascade reservoir group with the objectives of maximizing power generation and maximizing minimum output, and add constraints to the optimal scheduling model for the cascade reservoir group; preset the maximum number of iterations, population size, dominant population size, and determine the outflow from the reservoir during the scheduling period as the decision variable;

[0075] The module for solving the multi-objective scheduling cooperative search method for reservoir groups includes an initialization unit, a team communication strategy unit, a self-learning strategy unit, an internal competition strategy unit, a dominant population and individual optimal position update unit, and a reservoir optimal scheduling scheme acquisition unit.

[0076] The initialization unit is configured to perform the following actions: based on the preset range of decision variables, initialize the population, the optimal position of individuals in the population, and the fitness of individuals, and then obtain the dominant population and the optimal individuals in the dominant population based on the fast non-dominated sorting method;

[0077] The team communication strategy unit is configured to perform the following actions: generate communication individuals based on the dominant population, the best individual, and the individual's best position, using a random communication method;

[0078] The self-learning strategy unit is configured to perform the following actions: learn based on the upper and lower limits of the communication individual and decision variables to generate a learning individual;

[0079] The internal competition strategy unit is configured to perform the following actions: determine offspring by dynamically selecting weights based on dominance relationships among communicating and learning individuals;

[0080] The dominant population and individual optimal position update unit is configured to perform the following actions: calculate the fitness of the current population and the dominant population, obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, update the dominant population, then use the conventional update method to determine the optimal position of the individual, and then use the adaptive escape from local optima strategy to dynamically replace the optimal position of the individual.

[0081] The reservoir optimal scheduling scheme acquisition unit is configured to perform the following actions: perform iterative updates; if the preset maximum number of iterations is reached, the optimal scheduling scheme for the reservoir is obtained and the fitness value of the dominant population is output; otherwise, it is returned to the team communication strategy unit.

[0082] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0083] 1. This method has advantages such as simple implementation, fast convergence speed, strong global optimization ability and not easily getting trapped in local optima, and wide applicability. It can comprehensively consider the combined benefits of power generation and output in reservoir optimization scheduling, providing a new approach to solving the multi-objective optimization scheduling problem of reservoirs.

[0084] 2. This method can conduct a broad search across the entire solution space through team communication and internal competition strategies, resulting in fast convergence. Furthermore, through self-learning and escaping local optima strategies, the algorithm is less prone to getting trapped in local optima and can escape them, achieving a good balance between global and local exploration.

[0085] 3. This method only requires two learning coefficients and one self-learning adjustment coefficient as input, which requires fewer parameters, avoids a complex parameter tuning process, and is simple to program.

[0086] 4. This method can achieve global optimization for reservoir group scheduling problems, further improve computational efficiency and accuracy, and quickly provide a set of reservoir group scheduling schemes. This provides a more effective method for solving complex multi-objective engineering problems of reservoir groups and provides decision-makers with the option to select the optimal scheme. Attached Figure Description

[0087] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0088] Figure 2 This is a flowchart illustrating the cooperative search algorithm of the present invention.

[0089] Figure 3 This is a schematic diagram showing the comparison results of the model with the maximum power generation in a reservoir according to the present invention.

[0090] Figure 4 This is a schematic diagram showing the comparison results of the present invention in the minimum output maximum model.

[0091] Figure 5 This is the Pareto preface diagram of the multi-objective scheduling model of this invention.

[0092] Figure 6 The water level change curve is plotted from the solution with the highest power generation in the Pareto preface solution set. Detailed Implementation

[0093] To better understand the technical content of the present invention, specific embodiments are described below in conjunction with the accompanying drawings.

[0094] In this invention, various aspects of the invention are described with reference to the accompanying drawings, in which numerous illustrative embodiments are shown. Embodiments of the invention are not limited to those depicted in the drawings. It should be understood that the invention is implemented through any of the various concepts and embodiments described above, as well as the concepts and embodiments described in detail below, because the concepts and embodiments disclosed herein are not limited to any particular implementation. Furthermore, some aspects of the invention disclosed may be used alone or in any suitable combination with other aspects of the invention disclosed.

[0095] like Figure 1 As shown, a cooperative search method for multi-objective scheduling of reservoir groups includes the following steps:

[0096] S1. Based on the basic information data of the reservoirs, construct an optimal scheduling model for a cascade reservoir group with the objectives of maximizing power generation and maximizing minimum output. At the same time, add constraints to the optimal scheduling model for the cascade reservoir group; preset the maximum number of iterations, population size, dominant population size, and determine the outflow from the reservoir during the scheduling period as the decision variable.

[0097] The objective function for maximizing power generation in the optimal scheduling model of a cascade reservoir group is:

[0098]

[0099] Where E is the total power generation of all reservoirs during the time period; K is the number of reservoirs; J is the total number of time periods; P k,j Q k,j H k,j These represent the power output, power generation flow, and head of the k-th reservoir in the j-th time period, respectively; A k t is the output coefficient of the k-th reservoir; j Let j be the number of hours in the j-th time period.

[0100] Minimum output maximum objective function:

[0101]

[0102] Where F represents the minimum output of all reservoirs during the time period.

[0103] The specific constraints added to the optimal scheduling model for this cascade reservoir group include:

[0104] Output constraints

[0105]

[0106] in, These represent the minimum and maximum output of the k-th reservoir during the j-th time period, respectively.

[0107] Storage capacity constraints

[0108]

[0109] in, These are the minimum and maximum reservoir capacities for the k-th reservoir during the j-th time period, respectively.

[0110] Water balance constraints

[0111]

[0112] Among them, I k,j O k,j These represent the inflow and outflow of the k-th reservoir during the j-th time period, respectively. The outflow from the upstream reservoir is reduced by the amount of water lost.

[0113] Outbound flow constraints

[0114]

[0115] in, These are the minimum and maximum outflow rates of the k-th reservoir during the j-th time period, respectively.

[0116] Power generation flow constraints

[0117]

[0118] in, These represent the minimum and maximum power generation flows of the k-th reservoir during the j-th time period, respectively.

[0119] Initial and final storage capacity constraints

[0120]

[0121] in, These are the initial and final storage capacities of the k-th reservoir, respectively.

[0122] S2. The optimal reservoir scheduling scheme is solved by adopting a multi-objective scheduling cooperative search method for reservoir groups.

[0123] The specific solution process for step S2 is as follows: Figure 2 As shown, it includes the following steps: S201 to S206;

[0124] S201. Initialization: Generate an initial population based on a preset range of decision variables, store individuals in the initial population as their optimal positions, calculate the fitness of individuals in the population, and then obtain the dominant population and its optimal individuals based on the fast non-dominated ranking method. Step S201 includes the following sub-steps:

[0125] S201-1. Randomly generate an initial population within the preset range of decision variables, store the individuals in the initial population as their optimal position pbest, and calculate their fitness; the decision variables are as follows:

[0126]

[0127] Where, x i,j Let j be the value of the decision variable for the i-th individual; Indicates the interval A function that generates uniformly distributed random numbers; N is the population size; D is the number of decision variables.

[0128] S201-2. Obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm. Store the non-dominated front individuals of the feasible solution set in the dominant population gbest. If there is no feasible solution, store the non-dominated front individuals with the smallest constraint violation value among the infeasible solutions in the dominant population size gbest.

[0129] S2201-3. In the dominant population gbest, individuals are sorted in descending order according to their crowding distance. One individual is randomly selected from the top K individuals and stored in the optimal individual gbest. ind If gbest does not store K individuals, then randomly select an individual from the dominant population gbest and store it in the optimal individual gbest. ind In this example, K is N / 10.

[0130] S202, Execution Team Communication Strategy: Based on the dominant population, optimal individual, and optimal individual position, communication individuals are generated using a random communication method, specifically:

[0131] Through optimal population gbest, optimal individual gbest ind The communication individual u is generated from three aspects: the individual's optimal position (pbest), and two random communication methods, as shown in the following formula:

[0132]

[0133] A i,j =log(1 / φ(0,1))·(gbest ind,j -x i,j (55)

[0134] B i,j =α·φ(0,1)·(gbest m,j -x i,j (56)

[0135]

[0136] Among them, u i,j A is the value of the j-th decision variable for the i-th individual, generated through communication strategies; i,j Indicates from gbest ind,j Updated information at location B; i,j This represents the information updated from gbest; m is the randomly selected individual index; C i,j This represents the information updated from pbest; α and β are the adjustments made to B, respectively. i,j and C i,jAdjustment coefficient for the degree of influence. S203, Implementing a self-learning strategy: Learning is conducted based on the upper and lower limits of the communication individual and decision variables to generate a learning individual. The specific process is as follows:

[0137] The learning individual v is generated by learning the extent to which the current individual's range relative to the upper and lower limits of the decision variable is determined, as shown in the following formula:

[0138]

[0139]

[0140]

[0141] s i,j =x i,j +d i,j ·φ(-1,1) (61)

[0142]

[0143]

[0144] Among them, v i,j r is the value of the j-th decision variable for the i-th individual generated through a self-learning strategy; i,j p i,j and s i,j It refers to a self-learning individual generated based on self-learning strategies; d i,j It is the dynamic range of change generated based on the upper and lower limits of the decision variables; t and T are the current iteration number and the maximum iteration number of the algorithm, respectively; c j This represents the mean of the upper and lower limits of the decision variable.

[0145] S204. Implementing an internal competition strategy: Determining offspring through dynamic selection weights based on dominance relationships among communicating and learning individuals. The specific process is as follows:

[0146] In the communication individual u and the self-learning individual v, offspring are determined based on the dominance relationship. When u and v do not dominate each other, dynamic selection weights are used to determine offspring, as shown in the following formula:

[0147]

[0148] Where, x i For the final determined offspring; u i and v i γ represents the individuals generated by the communication strategy and the self-learning strategy; γ is the balance coefficient between the communication strategy and the selection strategy; t and T are the current iteration number and the maximum iteration number of the algorithm, respectively.

[0149] S205. Update the dominant population and the optimal position of individuals: Calculate the fitness of the current population and the dominant population, obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, update the dominant population, and then use the conventional update method to determine the optimal position of individuals. After that, adopt an adaptive strategy of escaping local optima to dynamically replace the optimal position of individuals. Specifically, this includes sub-steps S205-1 to S205-3:

[0150] S205-1. Calculate the fitness of the current population and the dominant population, and obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm. If the number of non-dominated fronts of the feasible solution set does not exceed the number of dominant populations, store the non-dominated fronts of the feasible solution set in the dominant population gbest. Otherwise, calculate the crowding distance value of all individuals, sort the individuals in descending order according to the crowding distance value, and remove individuals that exceed the number of dominant populations according to the sorted sequence number.

[0151] S205-2. The optimal position of an individual is determined using the conventional update method, as shown in the following formula:

[0152]

[0153] Among them, pbest i For the individual's optimal position; x i For the current individual.

[0154] S205-3. Adopt an adaptive strategy to escape local optima and dynamically replace the individual's optimal position as follows:

[0155]

[0156] Where, m q It randomly selects q individuals from the population N to replace the current individual's optimal position; floor is the floor function.

[0157] S206. Perform iterative updates. If the preset maximum number of iterations is reached, the optimal reservoir scheduling scheme is obtained, and the fitness value of the dominant population is output. Otherwise, return to step S202.

[0158] In another aspect, this invention proposes a multi-objective scheduling cooperative search system for reservoir groups, comprising:

[0159] The model building module is configured to perform the following actions: Based on the basic information data of the reservoirs, construct an optimal scheduling model for a cascade reservoir group with the objectives of maximizing power generation and maximizing minimum output, and add constraints to the optimal scheduling model for the cascade reservoir group; preset the maximum number of iterations, population size, dominant population size, and determine the outflow from the reservoir during the scheduling period as the decision variable;

[0160] The module for solving the multi-objective scheduling cooperative search method for reservoir groups includes an initialization unit, a team communication strategy unit, a self-learning strategy unit, an internal competition strategy unit, a dominant population and individual optimal position update unit, and a reservoir optimal scheduling scheme acquisition unit.

[0161] The initialization unit is configured to perform the following actions: based on the preset range of decision variables, initialize the population, the optimal position of individuals in the population, and the fitness of individuals, and then obtain the dominant population and the optimal individuals in the dominant population based on the fast non-dominated sorting method;

[0162] The team communication strategy unit is configured to perform the following actions: generate communication individuals based on the dominant population, the best individual, the individual's best position, and using random communication methods.

[0163] The self-learning strategy unit is configured to perform the following actions: learn based on the upper and lower limits of the communication individual and decision variables to generate a learning individual;

[0164] The internal competition strategy unit is configured to perform the following actions: determine offspring by dynamically selecting weights based on dominance relationships among communicating and learning individuals;

[0165] The dominant population and individual optimal position update unit is configured to perform the following actions: calculate the fitness of the current population and the dominant population, obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, update the dominant population, then use the conventional update method to determine the optimal position of the individual, and then use the adaptive escape from local optima strategy to dynamically replace the optimal position of the individual.

[0166] The reservoir optimal scheduling scheme acquisition unit is configured to perform the following actions: perform iterative updates; if the preset maximum number of iterations is reached, the optimal scheduling scheme for the reservoir is obtained and the fitness value of the dominant population is output; otherwise, it is returned to the team communication strategy unit.

[0167] The established multi-objective optimization scheduling model for power generation and output was solved using the present invention and the NSGA-III method, respectively. The parameter settings are pre-defined as shown in Table 1.

[0168] Table 1 Parameter Setting Details

[0169]

[0170] Results Analysis: To compare the stability and optimization breadth of this invention and NSGA-III in single-objective and multi-objective reservoir scheduling problems, 15 independent calculations were performed on both single-objective and multi-objective scheduling problems. The calculation results were analyzed, and the Pareto preface diagram was plotted using the relatively optimal multi-objective scheduling results. See Table 2 for details. Figures 3 to 6 .

[0171] Table 2 Comparison of Results

[0172]

[0173] As shown in Table 2, the present invention outperforms the NSGA-III scheme in terms of maximum, minimum, mean, and standard deviation of the maximum power generation model. Furthermore, it outperforms the NSGA-III scheme in terms of maximum, minimum, and mean of the minimum output model. See Table 2 for details of the independently calculated results. Figure 3 and Figure 4 Therefore, in single-objective scheduling problems, the stability and optimization breadth of this invention have significant advantages over NSGA-III.

[0174] Depend on Figure 5 It can be seen that the Pareto front results obtained by the NSGA-III algorithm are dominated by the present invention. At the same time, taking the Pareto front balancing scheme, it can be seen that the balancing scheme of the present invention is better than the NSGA-III balancing scheme in terms of power generation and minimum output, as shown in Table 2. Therefore, from the perspective of algorithm optimization and dominance relationship, the present invention performs the best.

[0175] In the Pareto preamble obtained from the two algorithms, the solution with the highest power generation was selected to plot the water level change curve. See details below. Figure 6 It can be seen that in order to maximize power generation, each reservoir stores as much water as possible to increase the water head and thus increase power generation. Therefore, the scheduling process of this invention meets the actual scheduling needs, satisfies the feasibility of actual scheduling problems, and can make the power generation of cascade reservoirs more coordinated with the grid demand, thereby promoting the full utilization of water resources by cascade reservoirs.

[0176] While the present invention has been described above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A multi-objective scheduling cooperative search method for a reservoir group, characterized in that, The method comprises the following steps: S1, based on the basic information data of the reservoir, a cascade reservoir group optimal scheduling model with the maximum power generation and the maximum minimum output as the target is constructed, and constraint conditions are added to the cascade reservoir group optimal scheduling model; The preset maximum iteration number, population size, and dominant population size are determined, and the reservoir discharge flow in the reservoir scheduling period is determined as the decision variable; S2, the reservoir optimal scheduling scheme is solved by adopting a reservoir group multi-objective scheduling cooperative search method: specifically, S201, initialization: based on the preset range of the decision variable, an initial population is generated, the individual in the initial population is stored as the individual optimal position, the fitness of the individual in the population is calculated, and then the dominant population and the optimal individual in the dominant population are obtained based on the fast non-dominated sorting method; S202, execute team communication strategy: based on the dominant population, the optimal individual, the individual optimal position, and the random communication method, a communication individual is generated; S203, execute self-learning strategy: based on the communication individual and the upper and lower limit range of the decision variable, learning is performed, and a learning individual is generated; S204, execute internal competition strategy: in the communication individual and the learning individual, the offspring is determined by the dominance relationship through dynamic selection weight; specifically, in the communication individual u and the self-learning individual v, the offspring is determined by the dominance relationship, and when u and v do not dominate each other, the offspring is determined by dynamic selection weight, as follows: where x i is the final determined offspring; u i and v i are individuals generated by the communication strategy and the self-learning strategy; γ is the balance coefficient of the communication strategy and the selection strategy; t, T are the current iteration number and the maximum iteration number of the algorithm, respectively. S205, update the dominant population and the individual optimal position: calculate the fitness of the current population and the dominant population, obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, update the dominant population, then determine the optimal position of the individual by adopting the conventional update method, and then dynamically replace the individual optimal position by adopting the adaptive jump out of local optimal strategy; comprising the following sub-steps: S205-1, calculate the fitness of the current population and the dominant population, and obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, if the number of non-dominated frontiers of the feasible solution set does not exceed the number of the dominant population, store the non-dominated frontiers of the feasible solution set in the dominant population gbest, otherwise calculate the crowding distance value of all individuals, sort the individuals in descending order according to the crowding distance value, and remove the individuals exceeding the number of the dominant population according to the sequence number; S205-2, the optimal position of the individual is determined by adopting the conventional update method as follows: where pbest i is the individual best position; x i is the current individual; S205-3, the individual optimal position is dynamically replaced by adopting the adaptive jump out of local optimal strategy as follows: where m q is a random selection of q individuals in population N to replace the current individual best position; floor is the floor function; S206, perform iteration update, if the preset maximum iteration number is reached, the reservoir optimal scheduling scheme is obtained, and the fitness value of the dominant population is output, otherwise return to step S202.

2. The multi-objective scheduling cooperative search method for a reservoir group according to claim 1, characterized in that, In step S1, the objective function of the maximum power generation of the cascade reservoir group optimal scheduling model is: Wherein, E is the total power generation of all reservoirs in the period; K is the number of reservoirs; J is the total number of periods; P k,j , Q k,j , H k,j are the output, power generation flow, and water head of the kth reservoir in the jth period, respectively; A k is the output coefficient of the kth reservoir; t j is the number of hours in the jth period. The minimum output maximum objective function is: Wherein, F is the minimum output of all reservoirs in the period.

3. The multi-objective scheduling cooperative search method for a reservoir group according to claim 1, characterized in that, The constraint conditions added to the cascade reservoir group optimal scheduling model specifically include: Output constraint wherein, respectively the minimum and maximum output of the kth reservoir in the jth period. Reservoir capacity constraint wherein, respectively the minimum and maximum reservoir capacity of the kth reservoir in the jth period. Water balance constraint where I k,j , O k,j are the interval inflow and outflow of the kth reservoir in the jth period, respectively. is the outflow of the upstream reservoir minus the loss of water. Discharge flow constraint wherein, respectively the minimum and maximum outflow of the kth reservoir in the jth period; power generation flow constraints wherein, respectively the minimum and maximum power generation flow of the kth reservoir in the jth period; initial and final reservoir storage constraints wherein, V0and Vf are the initial and final reservoir capacities, respectively, of the kth reservoir.

4. The multi-objective scheduling cooperative search method for a reservoir group according to claim 1, characterized in that, Step S201 comprises the following sub-steps: S201-1, randomly generate an initial population in a preset range of decision variables, store the individuals in the initial population as individual optimal positions pbest, and calculate individual fitness; the decision variable is as follows: where x i,j is the value of the jth decision variable of the ith individual; denotes the interval a function that generates a uniformly distributed random number; N is the population size; D is the number of decision variables; S201-2, obtain the dominance relation of the solution set based on the fast non-dominated sorting algorithm, store the non-dominated front individuals of the feasible solution set in the dominant population gbest, if there is no feasible solution, store the non-dominated front individual with the minimum constraint violation value in the infeasible solution in the dominant population size gbest; S2201-3, sort the individuals in the dominant population gbest in descending order according to the distance value of the crowding distance sorting, randomly select one individual in the first K individuals and store it in the optimal individual gbest ind , if there is no K individuals stored in the dominant population gbest, randomly select one individual in the dominant population gbest and store it in the optimal individual gbest ind .

5. The multi-objective scheduling cooperative search method for a reservoir group according to claim 1, characterized in that, Step S202, the specific team communication strategy, involves: using the optimal population (gbest) and the optimal individual (gbest) as the key factors. ind The communication individual u is generated from three aspects: the individual's optimal position (pbest), and two random communication methods, as shown in the following formula: A i,j = log(l / φ(0,1)) · (gbest ind,j -x i,j ) B i,j = a · φ(0, 1) · (gbest m,j -x i,j ) where, u i,j is the value of the jth decision variable of the ith individual generated by the communication strategy; A i,j represents the information updated from gbest ind,j ; B i,j represents the information updated from gbest i,j ; m is the randomly selected individual number; C i,j represents the information updated from pbest i,j ; and α and β are the adjustment coefficients of the influence degree of B 6. The multi-objective scheduling cooperative search method for a reservoir group according to claim 1, characterized in that, Step S203 performs a self-learning strategy, specifically: learning from the current individual and the upper and lower limits of the decision variable to generate a learning individual v, as follows: s i,j = x i,j + d i,j · φ(-1, 1) wherein v i,j is the value of the jth decision variable of the ith individual generated by the self-learning strategy; r i,j , p i,j and s i,j are self-learning individuals generated according to the self-learning strategy; d i,j is a dynamically changing range generated according to the upper and lower limits of the decision variable; t, T are the current iteration number and the maximum iteration number of the algorithm, respectively; c j is the mean of the upper and lower limits of the decision variable.

7. A multi-objective reservoir group scheduling cooperative search system, characterized in that, including: The model construction module is configured to perform the following actions: based on the basic information data of the reservoir, construct a cascade reservoir group optimal scheduling model with the maximum power generation and the maximum minimum output as the target, and add constraint conditions to the cascade reservoir group optimal scheduling model; The preset maximum iteration number, population size, and dominant population size are determined, and the reservoir discharge during the reservoir scheduling period is determined as the decision variable; the reservoir group multi-objective scheduling cooperative search method solving module includes an initialization unit, a team communication strategy unit, a self-learning strategy unit, an internal competition strategy unit, a dominant population and individual optimal position updating unit, and a reservoir optimal scheduling scheme acquisition unit: the initialization unit is configured to perform the following actions: based on the preset range of decision variables, initialize the population, the individual optimal position in the population, and the individual fitness, and then obtain the dominant population and the optimal individual in the dominant population based on the fast non-dominated sorting method; The team communication strategy unit is configured to perform the following actions: based on the dominant population, the optimal individual, and the individual optimal position, generate a communication individual using a random communication method; The self-learning strategy unit is configured to perform the following actions: learning from the communication individual and the upper and lower limit range of the decision variable to generate a learning individual; The internal competition strategy unit is configured to perform the following actions: in the communication individual and the learning individual, determine the offspring by adopting a dynamic selection weight based on the dominance relationship; specifically: in the communication individual u and the self-learning individual v, determine the offspring by adopting a dynamic selection weight based on the dominance relationship, when u and v do not dominate each other, determine the offspring by adopting a dynamic selection weight, as follows: where x i is the final determined offspring; u i and v i are individuals generated by the communication strategy and the self-learning strategy; γ is the balance coefficient of the communication strategy and the selection strategy; t, T are the current iteration number and the maximum iteration number of the algorithm, respectively. The dominant population and individual optimal position updating unit is configured to perform the following actions: calculate the fitness of the current population and the dominant population, obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, update the dominant population, and then determine the optimal position of the individual by adopting the conventional updating method, and then dynamically replace the individual optimal position by adopting the self-adaptive jump out of local optimal strategy; specifically perform the following sub-steps: S205-1, calculate the fitness of the current population and the dominant population, and obtain the dominance relationship of the solution set based on the fast non-dominated sorting algorithm, if the number of non-dominated frontiers of the feasible solution set does not exceed the number of dominant populations, store the non-dominated frontiers of the feasible solution set in the dominant population gbest, otherwise calculate the crowding distance value of all individuals, and sort the individuals in descending order according to the crowding distance value, and remove the individuals exceeding the number of dominant populations according to the sorted serial number; S205-2, determine the optimal position of the individual by adopting the conventional updating method as follows: where pbest i is the individual best position; x i is the current individual; S205-3, dynamically replace the individual optimal position by adopting the self-adaptive jump out of local optimal strategy as follows: where m q is a random selection of q individuals in population N to replace the current individual best position; floor is the floor function; S206, if the preset maximum iteration number is reached, the optimal reservoir scheduling scheme is obtained, and the fitness value of the dominant population is output, otherwise the team communication strategy unit is returned to execute; The reservoir optimal scheduling scheme obtaining unit is configured to perform the following actions: iteratively updating, if the preset maximum iteration number is reached, the optimal reservoir scheduling scheme is obtained, and the fitness value of the dominant population is output, otherwise the team communication strategy unit is returned to execute.

Images