Reservoir drought limit water level staging ecological flow optimization method

Abstract

The invention discloses a reservoir drought limit water level staging ecological flow optimization method. The method comprises the following steps: S1, generating an ecological window set; s2, constructing a hydrological state set, and calculating a corresponding hydrological state transition probability matrix; s3, setting physical relationship constraints; s4, constructing an improved particle swarm optimization model; s5, finally constructing a complete improved particle swarm optimization model; s6, performing iterative search on the improved particle swarm optimization model, and outputting an optimal particle solution composed of a reservoir water level control curve corresponding to the ecological window set and an ecological flow discharge sequence; and S7, converting the optimal particle solution into an executable control signal of a gate opening sequence, a flow guide unit adjusting instruction and a real-time reservoir water level set value which can be identified by a reservoir monitoring data acquisition and monitoring system, and executing the control signal through the monitoring data acquisition and monitoring system to implement stage-by-stage reservoir scheduling. According to the invention, organic unification of intelligentization of reservoir multi-target stage scheduling and engineering applicability is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir monitoring, and in particular to a reservoir drought-limited water level phased ecological flow optimization method. Background Art

[0002] With the continuous advancement of water resources regulation and river basin ecological protection, multi-objective phased reservoir operation is playing an increasingly important role in the integrated management of water supply, power generation, flood control, and ecological protection. In recent years, the dual constraints of drought-limited water levels and ecological flow have become a hot research topic in ecohydrology and water resources engineering. Achieving a dynamic balance between reservoir water level security and ecological baseflow protection in the context of uncertain water inflow and changing ecological demands is a major challenge in current reservoir operation research and engineering practice.

[0003] Existing methods for controlling reservoir drought-limited water levels and ecological flows in stages generally adopt "static linear" or "single threshold within the year" water level control lines. Water level scheduling plans are often manually set based on historical experience or deterministic rules. They lack the ability to dynamically respond to nonlinear hydrological situations such as extreme droughts and sudden floods. Traditional models mainly implement water level and flow scheduling through single-period linear programming or static threshold methods. They are unable to carry out differentiated control of the phased water demand and habitat protection goals of different ecological windows within the year, resulting in insufficient discharge of reservoir ecological flow or increased risks to water supply security in the later period in extreme drought years or climate change events.

[0004] On the other hand, although the existing multi-objective swarm intelligence algorithm can optimize reservoir scheduling plans to a certain extent, it is prone to premature convergence, local extreme values ​​and parameter adjustment rigidity in complex scenarios with high-dimensional phases, multiple constraints and extreme climates, making it difficult to achieve a dynamic balance between the dual goals of ecology and safety. In addition, the mainstream scheduling model is disconnected from the actual reservoir control system, and the calculation results are difficult to convert into executable gate adjustment instructions in real time, making it difficult to implement the model results. Information lag and execution deviation further aggravate the contradiction between ecological and water supply goals.

[0005] Therefore, there is an urgent need for a new reservoir operation optimization method that can achieve phased operation adaptation, respond to extreme hydrological situations, and enhance the ability to coordinate ecological and water supply guarantees. Summary of the Invention

[0006] One purpose of the present invention is to propose a reservoir drought-limited water level phased ecological flow optimization method, which realizes the organic unity of intelligent multi-objective phased scheduling of reservoirs and engineering applicability.

[0007] A reservoir drought-limited water level phased ecological flow optimization method according to an embodiment of the present invention includes the following steps:

[0008] S1. Construct an initial parameter set and perform preprocessing to form a standardized hydrological time series and generate an ecological window set;

[0009] S2. Use standardized hydrological time series to classify drought, normal water, abundant water, and extreme drought conditions, construct a hydrological state set, and calculate the corresponding hydrological state transition probability matrix;

[0010] S3. Establish a multi-objective optimization problem based on the objectives of minimizing the ecological water shortage rate and minimizing the reservoir water supply risk, and set physical relationship constraints;

[0011] S4. Construct an improved particle swarm optimization model and use period coding to represent the particle position as the reservoir water level control variable and ecological flow control variable corresponding to the ecological window set;

[0012] S5. Embed the hydrological state transition probability matrix into the particle velocity update rule, assign the current hydrological state label to each particle, and completely improve the final construction of the particle swarm optimization model;

[0013] S6. Perform an iterative search on the improved particle swarm optimization model, evaluate the particle fitness based on the multi-objective optimization problem, update the individual particle optimal solution and the global optimal solution of the swarm, adjust the particle position based on the particle's current hydrological state label, and output the optimal particle solution consisting of the reservoir water level control curve and the ecological flow discharge sequence corresponding to the ecological window set;

[0014] S7. Convert the optimal particle solution into executable control signals of gate opening sequence, diversion unit adjustment instructions and real-time reservoir water level setting value that can be recognized by the reservoir monitoring data acquisition and monitoring system, and execute the control signals through the monitoring data acquisition and monitoring system to implement phased reservoir scheduling.

[0015] Optionally, the initial parameter set includes historical water inflow data, real-time water inflow data, reservoir capacity-water level curve, drought-limited water level control line parameters, ecological flow demand data and scheduling year phase division data.

[0016] Optionally, S2 includes the following steps:

[0017] S21. Based on the standardized hydrological time series Q = {q1,q2,...,q T}, where q t represents the standardized inflow flow value at the t-th time step, T represents the total number of time steps, and the hydrological time series is divided into intervals. According to the set hydrological boundary threshold θ={θ1,θ2,θ3}, each q t Mapped to the corresponding hydrological state s t∈S, we get the hydrological state set S = {D1, N1, W1, ED1}, where D1 represents the drought state, N1 represents the normal water state, W1 represents the abundant water state, ED1 represents the extreme drought state, and θ1, θ2, θ3 are the standardized water inflow thresholds used to divide the boundaries of each hydrological state;

[0018] S22. Construct the hydrological state transition frequency matrix M=[m ij ] 4×4 , where m ij Indicates that the hydrological state changes from state s i Transfer to state s j The frequency of s i ,s j ∈S;

[0019] S23. Calculate the hydrological state transition probability matrix P based on the hydrological state transition frequency matrix M = [p ij ] 4×4 , where element p ij Indicates that the hydrological state at time t is s i Under the condition of j probability.

[0020] Optionally, S3 includes the following steps:

[0021] S31. Calculate the ecological water shortage rate based on the ecological window set. The ecological water shortage rate is used to measure the extent to which the actual outflow within the ecological window does not meet the minimum ecological flow requirement within the ecological window;

[0022] S32. Calculate the reservoir water supply risk based on the scheduling year time step. The reservoir water supply risk is used to measure the risk of the reservoir water level at each time step in the scheduling year falling below the drought limit water level control line at the corresponding time step;

[0023] S33. Construct a dual-objective optimization function for the dual-objective optimization problem using weighted normalization based on the ecological water shortage rate and the reservoir water supply risk. Normalize the ecological water shortage rate and the reservoir water supply risk to preset upper and lower bounds, multiply each by a preset weight coefficient, and sum the two weighted results to obtain the value of the dual-objective optimization function.

[0024] S34. For all time steps in the scheduling year, set a constraint that the reservoir water level is not lower than the drought limit water level control line corresponding to that time step, so that the reservoir water level in each time step is greater than or equal to the drought limit water level control line for that time step;

[0025] S35. For all ecological windows in the ecological window set, set a constraint that the actual downstream flow rate is not less than the minimum ecological flow rate requirement of the ecological window, so that the actual downstream flow rate of each ecological window is greater than or equal to the minimum ecological flow rate requirement of the ecological window. The actual downstream flow rate is the optimization variable, and the minimum ecological flow rate requirement of the ecological window is the ecological security benchmark.

[0026] S36. For all time steps in the scheduling year, set a physical relationship constraint between storage capacity and discharge. Based on the storage capacity-water level curve, the reservoir water level at each time step is uniquely mapped to the reservoir storage capacity at that time step. Based on the storage capacity-discharge relationship and the inflow flow at that time step, establish a one-to-one correspondence between the storage capacity, reservoir water level, and discharge flow.

[0027] S37. Integrate the dual-objective optimization function and all constraints to form the multi-objective optimization problem modeling result.

[0028] Optionally, the S4 includes the following steps:

[0029] S41. Initialize the particles using the phased-coupling encoding method, and set the phased decision vector X of each particle i Constructed as a set of ecological windows containing the whole year Ω={ω1,ω2,…,ω K} Each window reservoir water level control variable H i,k and ecological flow control variable Q i,k , and introduce the phased cross-coupling term C i,k Reflecting the dynamic connection between scheduling variables in adjacent ecological windows, the particle code is:

[0030] X i ={[H i,1 ,Q i,1 ,C i,1 ],[H i,2 ,Q i,2 ,C i,2 ],…,[H i,K ,Q i,K ,C i,K ]};

[0031] Among them, H i,k represents the reservoir water level control variable of the i-th particle in the k-th ecological window, Q i,k represents the ecological flow control variable of the i-th particle in the k-th ecological window, C i,k is the periodized cross-coupling term;

[0032] S42. Introduce a phased adaptive inertia weight mechanism, and adjust the particle inertia weight w according to the hydrological state of the current ecological window. i,k Real-time dynamic adjustment:

[0033] w i，k=w base ·(1+α·R k );

[0034] Among them, w base is the basic inertia weight, α is the window sensitivity coefficient, R k represents the historical ecological water shortage risk of the kth ecological window;

[0035] S43. Based on the current ecological flow control variable Q of each ecological window i,k , minimum ecological flow demand and annual water supply safety margin, and the weight of the multi-objective optimization function λ eco,k ,λ sup,k Mapping to periodic variable coefficients:

[0036]

[0037] in, is the minimum ecological flow demand of the kth ecological window, H lim,k is the drought limit water level control line of the kth window;

[0038] S44. Set the population structure to a global-period two-layer collaborative architecture. That is, the particle swarm global optimal solution not only preserves the optimal scheduling sequence for the entire year, but also provides real-time feedback to each ecological window sub-swarm, achieving collaborative dynamic optimization of the annual and periodic solutions. The periodic sub-swarms evolve independently under the guidance of the global particle swarm, optimizing their response to periodic extreme climate events.

[0039] S45. The phased-probability jump operator is introduced in the particle velocity update. The phased-probability jump operator combines the hydrological state transition probability matrix P with the current particle phased decision vector X i When the ecological window is in an extreme drought state, the probability disturbance jump is triggered, and the reservoir water level control variable H i,k and ecological flow control variable Q i,k Introduce the probability disturbance term:

[0040]

[0041] in, is the new velocity of the kth ecological window of the i-th particle, is the old velocity of the kth ecological window of the i-th particle, V i,k is the kth ecological window speed of the ith particle, is the optimal individual history of the particle, is the global historical optimum, γ is the extreme drought disturbance coefficient, η i,k is the random disturbance that obeys the hydrological state transition probability matrix P.

[0042] Optionally, the S5 includes the following steps:

[0043] S51. The hydrological state transition probability matrix P = [p ij ] 4×4 Integrating into the particle velocity update mechanism, for each particle kth ecological window, according to the hydrological state set S and the current hydrological state label s i,k , using the hydrological state transition probability Determine the hydrological state migration of the particle in the next evolution step:

[0044]

[0045] in, represents the updated hydrological state label of the i-th particle in the k-th ecological window, and Categorical(·) is a random sampling function based on the transition probability;

[0046] S52. Update the hydrological status label Mapped to the periodic adaptive inertia weight w i,k , disturbance coefficient γ i,k and the probability disturbance term η i,k The selection rule is that when the hydrological status label is updated When the perturbation coefficient of each stage and the amplitude of the probability perturbation term are increased, the particles can increase their jumping ability in the extreme drought window and maintain ecological and safety constraints;

[0047] S53. The hydrological state transition probability is embedded in the particle velocity update mechanism, and in each iteration, the particle current hydrological state label s is used to update the particle velocity. i,k and its hydrological state transition probability, dynamically adjusting the speed and position of particles;

[0048] S54. An improved particle swarm optimization model is formed, and the improved particle swarm optimization model adaptively adjusts the speed and position of the particles according to the hydrological state transition probability of each ecological window and the current hydrological state label.

[0049] Optionally, the S6 includes the following steps:

[0050] S61. Perform iterative search on the improved particle swarm optimization model. In each iteration, for each particle i and ecological window k, the particle velocity updated based on the hydrological state transition probability matrix P is used. and location

[0051] S62. In each iteration, the current phase decision vector of each particle is calculated based on the multi-objective optimization function and the constraints. Calculate the value of the dual-objective optimization function F i :

[0052]

[0053] Among them, λ eco,k and λ sup,k is the weight of the multi-objective optimization function of the kth ecological window, and are the normalized ecological water shortage rate and water supply risk index of the i-th particle in the k-th ecological window;

[0054] S63. According to the dual objective optimization function value F of each particle i , update the individual historical optimal solution of the particle and the particle swarm historical global optimal solution If the current solution is better than the historical optimal solution, it is updated to the latest value;

[0055] S64. At the end of each iteration, according to the particle's current hydrological state label s i,k Dynamically adjust the position and velocity of particles based on the hydrological state transition probability matrix P;

[0056] S65. Determine whether the improved particle swarm optimization model meets the preset convergence conditions, which can be set based on the global optimal solution change threshold, the maximum iteration number, or the objective function convergence criterion;

[0057] S66. When the convergence conditions are met, output the reservoir water level control curve corresponding to the ecological window set and ecological flow discharge sequence The optimal particle solution composed of

[0058] Optionally, the S7 also includes real-time collection of actual reservoir water level data and actual discharge flow data during the scheduling execution process, generating execution result data, and calculating the execution deviation between the execution result data and the optimal particle solution, correcting the hydrological state transition probability matrix according to the execution deviation and adjusting the improved particle swarm optimization model parameters, and then returning to step S5, performing rolling optimization based on the corrected hydrological state transition probability matrix and the adjusted particle swarm optimization model parameters to achieve continuous closed-loop control of the reservoir drought-limited water level and the phased coordinated scheduling of the ecological flow.

[0059] The beneficial effects of the present invention are:

[0060] (1) The present invention deeply integrates the period coupling coding and Markov chain hydrological state transition mechanism into the particle swarm optimization model, constructs the scheduling decision vector of each particle as the reservoir water level, ecological flow and period continuity variable, and dynamically adjusts the particle inertia weight and velocity disturbance term according to the actual hydrological state. When the ecological window is in an extreme drought or flood mutation state, the particles can automatically generate jump disturbances according to the hydrological state transition probability, thereby greatly improving the model's adaptability and global search capabilities in the face of uncertain water inflow and extreme climate events, reducing premature convergence and local optimal problems, and realizing the coordinated dynamic regulation of the reservoir drought limit water level and ecological flow.

[0061] (2) The present invention introduces a phased adaptive multi-objective weight mapping mechanism. According to the actual discharge flow, minimum ecological flow demand and drought-limited water level constraint of each ecological window, the weight coefficient of the dual-objective optimization function is adjusted in real time and dynamically, and the window sensitivity adjustment is realized by combining it with the historical ecological water shortage risk. Through phased normalization and variable weights, the algorithm can finely balance the dynamic conflict between ecological protection and reservoir water safety, and achieve the optimal matching of phased flow and reservoir water level within the ecological window, thereby improving the ecological protection capacity and subsequent water supply safety under multi-objective phased scheduling, which is significantly better than the existing static weight or global single-objective phased method.

[0062] (3) The present invention directly converts the optimal solution of the particle swarm-Markov chain hybrid model into executable gate opening instructions and real-time reservoir water level settings in stages, and links it with the reservoir automation control system through the SCADA system to form a closed-loop scheduling process from model optimization to project execution. The Bayesian self-learning feedback mechanism is introduced in the scheduling execution process, and the hydrological state transition probability matrix and particle swarm parameters are corrected in real time based on the actual execution deviation. Continuous rolling optimization is carried out, which significantly reduces the risk of ecological flow and reservoir water level imbalance caused by information lag and execution error, improves the robustness of the model to sudden situations and consecutive extreme years, and realizes the organic unity of intelligent multi-objective phased scheduling of reservoirs and engineering applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0064] Figure 1 This is a flow chart of a reservoir drought-limited water level phased ecological flow optimization method proposed by the present invention. DETAILED DESCRIPTION

[0065] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0066] refer to Figure 1 A reservoir drought-limited water level phased ecological flow optimization method comprises the following steps:

[0067] S1. Construct an initial parameter set, perform wavelet denoising and time alignment based on the initial parameter set, form a standardized hydrological time series, and generate an ecological window set;

[0068] S2. Use standardized hydrological time series to classify drought, normal water, abundant water, and extreme drought conditions, construct a hydrological state set, and calculate the corresponding hydrological state transition probability matrix;

[0069] S3. Establish a multi-objective optimization problem based on the objectives of minimizing the ecological water shortage rate and minimizing the reservoir water supply risk. This includes constraints such as the reservoir water level must not fall below the drought limit water level, the ecological window minimum flow constraint, and the physical relationship between reservoir capacity and discharge.

[0070] S4. Build an improved particle swarm optimization model, using phased coding to represent the particle positions as the reservoir water level control variables and ecological flow control variables corresponding to the ecological window set, and set the particle velocity, dynamic inertia weight, learning factor, population size, and global optimal solution parameters;

[0071] S5. Embed the hydrological state transition probability matrix into the particle velocity update rule, assign the current hydrological state label to each particle, and completely improve the final construction of the particle swarm optimization model;

[0072] S6. Perform an iterative search on the improved particle swarm optimization model, evaluate the particle fitness based on the multi-objective optimization problem in step S3, update the individual optimal solution of the particle and the global optimal solution of the swarm, and adjust the particle position using the particle's current hydrological state label to determine whether the improved particle swarm optimization model meets the convergence condition. If so, output the optimal particle solution consisting of the reservoir water level control curve and the ecological flow discharge sequence corresponding to the ecological window set;

[0073] S7. Convert the optimal particle solution into executable control signals of gate opening sequence, diversion unit adjustment instructions and real-time reservoir water level setting value that can be recognized by the reservoir monitoring data acquisition and monitoring system, and implement phased reservoir scheduling through executing the control signals through the monitoring data acquisition and monitoring system.

[0074] In this embodiment, the initial parameter set includes historical water inflow data, real-time water inflow data, reservoir capacity-water level curve, drought limit water level control line parameters, ecological flow demand data and scheduling year phase division data.

[0075] In this embodiment, S2 includes the following steps:

[0076] S21. Based on the standardized hydrological time series Q = {q1,q2,...,qT}, where q t represents the standardized inflow flow value at the t-th time step, T represents the total number of time steps, and the hydrological time series is divided into intervals. According to the set hydrological boundary threshold θ={θ1,θ2,θ3}, each q t Mapped to the corresponding hydrological state s t ∈S, we get the hydrological state set S:

[0077] S = {D1, N1, W1, ED1};

[0078] Among them, D1 represents the drought state, N1 represents the normal water state, W1 represents the abundant water state, ED1 represents the extreme drought state, θ1, θ2, θ3 are the standardized water inflow thresholds, which are used to divide the boundaries of each hydrological state;

[0079] S22. Construct the hydrological state transition frequency matrix M=[m ij ] 4×4 , where m ij Indicates that the hydrological state changes from state s i Transfer to state s j The frequency of s i ,s j ∈S;

[0080] S23. Calculate the hydrological state transition probability matrix P based on the hydrological state transition frequency matrix M = [p ij ] 4×4 , where element p ij Indicates that the hydrological state at time t is s i Under the condition of j probability.

[0081] In this embodiment, S3 includes the following steps:

[0082] S31. Calculate the ecological water shortage rate based on the set of ecological windows. The ecological water shortage rate measures the extent to which the actual outflow within an ecological window falls short of the minimum ecological flow requirement within that window. The ecological water shortage rate is calculated by traversing all ecological windows and, for each window, taking the non-negative part of the difference between the minimum ecological flow requirement within that window and the actual outflow within that window. These results are then accumulated across all ecological windows and averaged. The set of ecological windows is used to divide the ecological scheduling time period. The minimum ecological flow requirement within each window is used to define the water supply guarantee standard for the ecosystem within each ecological window. The actual outflow is the water release result output by the scheduling plan for the corresponding ecological window.

[0083] S32. Calculate the reservoir water supply risk based on the scheduling year time step. The reservoir water supply risk measures the risk of the reservoir water level falling below the drought limit water level control line at each time step during the scheduling year. The reservoir water supply risk is calculated by traversing all time steps in the scheduling year and, for each time step, taking the non-negative part of the difference between the drought limit water level control line and the reservoir water level at that time step. The above results are accumulated across all time steps and averaged. The scheduling year time step is defined as all discrete time points in the annual scheduling. The drought limit water level control line is the minimum water level constraint to ensure safe water supply to the reservoir. The reservoir water level is the actual reservoir water level output by the scheduling plan.

[0084] S33. Construct a dual-objective optimization function for the dual-objective optimization problem using weighted normalization based on the ecological water shortage rate and reservoir water supply risk. The calculation method is to normalize the ecological water shortage rate and the reservoir water supply risk according to preset upper and lower bounds, then multiply each by a preset weight coefficient. Finally, the two weighted results are added together to obtain the dual-objective optimization function value. The dual-objective optimization function value serves as the basis for fitness evaluation of the dual-objective optimization problem. The normalized upper and lower bounds of the ecological water shortage rate and the reservoir water supply risk are used to standardize the dimensions, and the weight coefficient is used to adjust the relative importance of the two objectives in the comprehensive evaluation.

[0085] S34. For all time steps in the scheduling year, set the reservoir water level constraint to not be lower than the drought limit water level control line corresponding to that time step, so that the reservoir water level in each time step is greater than or equal to the drought limit water level control line of that time step.

[0086] Among them, the reservoir water level is the optimization variable, and the drought limit water level control line is the minimum safety guarantee water level;

[0087] S35. For all ecological windows in the ecological window set, set a constraint that the actual downstream flow rate is not less than the minimum ecological flow rate requirement of the ecological window, so that the actual downstream flow rate of each ecological window is greater than or equal to the minimum ecological flow rate requirement of the ecological window. The actual downstream flow rate is the optimization variable, and the minimum ecological flow rate requirement of the ecological window is the ecological security benchmark.

[0088] S36. For all time steps in the scheduling year, set physical constraints on the storage capacity-discharge relationship. Based on the storage capacity-water level curve, the reservoir water level at each time step is uniquely mapped to the reservoir storage capacity at that time step. Then, based on the storage capacity-discharge relationship and the inflow flow at that time step, establish a one-to-one correspondence between storage capacity, reservoir water level, and discharge flow to ensure that scheduling decisions at all time steps meet the physical constraints. The storage capacity-water level curve represents the static functional relationship between storage capacity and reservoir water level, while the storage capacity-discharge relationship represents the dynamic functional relationship between storage capacity and discharge flow. The inflow flow is the inflow from the reservoir upstream.

[0089] S37. Integrate the dual-objective optimization function and all constraints to form the multi-objective optimization problem modeling result.

[0090] In this embodiment, S4 includes the following steps:

[0091] S41. Initialize the particles using the phased-coupling encoding method, and set the phased decision vector X of each particle i Constructed as a set of ecological windows containing the whole year Ω={ω1,ω2,…,ω K} Each window reservoir water level control variable H i,k and ecological flow control variable Q i,k , and introduce the phased cross-coupling term C i,k Reflecting the dynamic connection between scheduling variables in adjacent ecological windows, the particle code is:

[0092] X1={[H i，1 , Q i，1 , C i，1 ],[H i，2 , Q i，2 , C i，2 ], ..., [H i，K , Q i，K , C i，K ]};

[0093] Among them, H i,k represents the reservoir water level control variable of the i-th particle in the k-th ecological window, Q i,k represents the ecological flow control variable of the i-th particle in the k-th ecological window, C i,k is the phased cross-coupling term, which quantifies the continuity of water level and flow scheduling in the front and back windows;

[0094] The phased-coupling encoding scheme proposed by S41 not only explicitly decomposes each particle's decision variables into reservoir water level control variables and ecological flow control variables for each ecological window throughout the year, but also introduces cross-coupling terms within each period to dynamically connect scheduling decision variables between different ecological stages. This formula design enables the scheduling particle encoding to possess both phased adaptive capabilities and flexible control of water level and flow changes between adjacent windows. This structure effectively overcomes the technical bottleneck of traditional particle swarms, which are limited to global single-unit search and lack the ability to accommodate phased ecological needs.

[0095] Furthermore, S41 uses an adaptive inertia weighting mechanism to directly map ecological window risk assessments to particle search weights. This dynamic weight calculation formula ensures the model can adjust its search strategy in real time based on ecological risk. This allows scheduling optimization to take into account the overall nature of each phase while also flexibly responding to extreme situations. This overcomes the limitations of existing technologies, which often involve rigid inertia parameters and poor scheduling robustness, and significantly enhances the flexibility and intelligence of phased scheduling. This formula system of phased-coupling encoding and adaptive weighting represents a significant structural innovation of the particle swarm algorithm in the field of phased ecological scheduling in a watershed.

[0096] S42. Introduce a phased adaptive inertia weight mechanism, and adjust the particle inertia weight w according to the hydrological state of the current ecological window. i,k Real-time dynamic adjustment allows particle inertia weights to adaptively reflect window risks:

[0097] w i,k =w base ·(1+α·R k );

[0098] Among them, w base is the basic inertia weight, α is the window sensitivity coefficient, R k represents the historical ecological water shortage risk of the kth ecological window;

[0099] S43. Implement the particle multi-objective dynamic weight mapping mechanism, based on the current ecological flow control variable Q of each ecological window i,k , minimum ecological flow demand and annual water supply safety margin, and the weight of the multi-objective optimization function λ eco,k ,λ sup,k Mapping to periodic variable coefficients:

[0100]

[0101] in, is the minimum ecological flow demand of the kth ecological window, H lim,k is the drought limit water level control line of the kth window;

[0102] S44. Set the population structure to a global-period two-layer collaborative architecture. That is, the particle swarm global optimal solution not only preserves the optimal scheduling sequence for the entire year, but also provides real-time feedback to each ecological window sub-swarm, achieving collaborative dynamic optimization of the annual and periodic solutions. The periodic sub-swarms evolve independently under the guidance of the global particle swarm, optimizing their response to periodic extreme climate events.

[0103] S45. The phased-probability jump operator is introduced in the particle velocity update. The phased-probability jump operator combines the hydrological state transition probability matrix P with the current particle phased decision vector X i When the ecological window is in an extreme drought state, the probability disturbance jump is triggered, and the reservoir water level control variable H i,k and ecological flow control variable Q i,k Introduce the probability disturbance term:

[0104]

[0105] Among them, V i,k is the kth ecological window speed of the ith particle, is the optimal individual history of the particle, is the global historical optimum, γ is the extreme drought disturbance coefficient, η i,k is the random disturbance that obeys the hydrological state transition probability matrix P.

[0106] In this embodiment, S5 includes the following steps:

[0107] S51. The hydrological state transition probability matrix P = [p ij ] 4×4 Integrating into the particle velocity update mechanism, for each particle kth ecological window, according to the hydrological state set S and the current hydrological state label s i,k , using the hydrological state transition probability Determine the hydrological state migration of the particle in the next evolution step:

[0108]

[0109] in, represents the updated hydrological state label of the i-th particle in the k-th ecological window, and Categorical(·) is a random sampling function based on the transition probability;

[0110] S52. Update the hydrological status label Mapped to the periodic adaptive inertia weight w i,k , disturbance coefficient γ i,k and the probability disturbance term η i,k The selection rule is that when the hydrological status label is updated When the perturbation coefficient of each stage and the amplitude of the probability perturbation term are increased, the particles can increase their jumping ability in the extreme drought window and maintain ecological and safety constraints;

[0111] S53. The hydrological state transition probability is embedded in the particle velocity update mechanism, and in each iteration, the particle current hydrological state label s is used to update the particle velocity. i,k and its hydrological state transition probability, dynamically adjusting the particle speed and position:

[0112]

[0113] in, is the new velocity of the kth ecological window of the i-th particle, w i,k is the current period inertia weight, c1, c2 are learning factors, r1, r2 are uniform random numbers in the interval [0, 1], is the individual historical optimal solution of the kth ecological window of the i-th particle, is the global historical optimal solution of the kth ecological window, γ i,k is the period disturbance coefficient, η i,k is the disturbance randomly generated based on the hydrological state transition probability matrix P;

[0114] Formula S53 defines the core mechanism of particle velocity update in particle swarm optimization during the optimization of reservoir drought-limited water level ecological flow phased scheduling in the present invention. The formula deeply couples the Markov chain hydrological state transition probability matrix P with the traditional particle swarm velocity update mechanism, which is specifically reflected in the following aspects:

[0115] Inertia weight phased adaptation: w i,k It is not a constant value, but is dynamically adjusted according to the current ecological window and its corresponding historical ecological water shortage risk, staged hydrological status, etc., so that particles can adaptively adjust the global search and local convergence capabilities in different ecological windows (such as extreme drought and abundant water).

[0116] Multi-objective phased optimization: retaining the particle's individual historical optimality and the global historical optimal The approaching term ensures the global optimization property of the phased decision.

[0117] Markov probability perturbation is introduced: γ i,k ·η i,k The probabilistic disturbance term is dynamically generated based on the hydrological state label of the particle's ecological window and its Markov state transition probability. It significantly amplifies the disturbance intensity under extreme hydrological conditions (such as extreme drought windows), realizes the probabilistic jumping and multi-path escape of particles, and effectively overcomes the problems of premature convergence and local extreme values.

[0118] Unlike the static parameters and homogeneous perturbations commonly used in prior PSOs, this method directly couples the particle search dynamics with real-world hydrological uncertainties by driving the magnitude and frequency of the perturbation term through a hydrological state transition probability matrix. This fundamentally advances the model from "empirical static optimization" to "probability-driven, phased adaptive optimization," enhancing its global resilience to extreme climate and sudden hydrological scenarios. By closely integrating particle phased encoding with phased weights and phased probabilistic perturbations, the S53 formula accounts for the spatiotemporal differences in ecological windows and variability in hydrological regimes, achieving phased, coordinated optimization that balances ecological and water supply security objectives. The formula demonstrates superior dynamic balance and robustness to traditional methods, particularly under high-risk, low-inflow, or rapid state transitions. The probabilistic perturbation term in the S53 formula not only theoretically increases the probability of obtaining a global optimal solution but also directly corresponds to the flexibility and real-time nature of dispatch instructions in practical engineering applications. This ensures that phased decisions can be adjusted immediately to sudden climate events within the dispatch cycle, significantly reducing the actual engineering risks of ecological dehydration and water supply disruptions, and providing a solid technical foundation for reservoir ecological dispatch and safety assurance.

[0119] S54. An improved particle swarm optimization model is formed. The improved particle swarm optimization model adaptively adjusts the speed and position of particles according to the hydrological state transition probability of each ecological window and the current hydrological state label, thereby improving the global optimization ability and ecological scheduling robustness of particles in extreme hydrological scenarios.

[0120] In this embodiment, S6 includes the following steps:

[0121] S61. Perform iterative search on the improved particle swarm optimization model. In each iteration, for each particle i and ecological window k, the particle velocity updated based on the hydrological state transition probability matrix P is used. and location Update particle positions:

[0122]

[0123] Among them, X i,k is the phased decision variable of the i-th particle in the k-th ecological window, including the reservoir water level control variable H i,k and ecological flow control variable Q i,k ;

[0124] S62. In each iteration, the current phase decision vector of each particle is calculated based on the multi-objective optimization function and the constraints. Calculate the value of the dual-objective optimization function F i :

[0125]

[0126] Among them, λ eco,k and λ sup,k is the weight of the multi-objective optimization function of the kth ecological window, and are the normalized ecological water shortage rate and water supply risk index of the i-th particle in the k-th ecological window;

[0127] The core of S62 is the dual-objective optimization construction for the phased scheduling problem. The formula normalizes the ecological water shortage rate and reservoir water supply risk respectively, and introduces a variable weight λ for each ecological window. eco,k and λ sup,k This approach enables flexible integration of multiple phased objectives and real-time priority adjustment. Dynamic weighting directly binds the particle's decision output in the current window to the degree of achievement of ecological and water supply objectives, avoiding the imbalanced phased scheduling caused by a single global weight in traditional methods.

[0128] The formula system breaks through the conventional group optimization method of single normalized target or static weight configuration, and organically combines phased target weights, phased normalization and dynamic mapping of historical risks, ensuring that the algorithm can take into account dynamic priority adjustment, weigh complex target conflicts and achieve real-time optimal scheduling in different ecological stages.

[0129] S63. According to the dual objective optimization function value F of each particle i , update the individual historical optimal solution of the particle and the particle swarm historical global optimal solution If the current solution is better than the historical optimal solution, it is updated to the latest value;

[0130] S64. At the end of each iteration, according to the particle's current hydrological state label s i,k Dynamically adjust the position and velocity of particles based on the hydrological state transition probability matrix P;

[0131] S65. Determine whether the improved particle swarm optimization model meets the preset convergence conditions, which can be set based on the global optimal solution change threshold, the maximum iteration number, or the objective function convergence criterion;

[0132] S66. When the convergence conditions are met, output the reservoir water level control curve corresponding to the ecological window set and ecological flow discharge sequence The optimal particle solution composed of and They are the optimal reservoir water level and optimal ecological flow control schemes for phased scheduling, respectively, as the optimization results of phased coordinated scheduling of reservoir drought-limited water level and ecological flow.

[0133] In this embodiment, S7 also includes real-time collection of actual reservoir water level data and actual discharge flow data during the scheduling execution process, generating execution result data, and calculating the execution deviation between the execution result data and the optimal particle solution. According to the execution deviation, the hydrological state transition probability matrix is ​​corrected and the improved particle swarm optimization model parameters are adjusted. Then, step S5 is returned to perform rolling optimization based on the corrected hydrological state transition probability matrix and the adjusted particle swarm optimization model parameters to achieve continuous closed-loop control of the reservoir drought-limited water level and the coordinated scheduling of the ecological flow in stages.

[0134] Example 1:

[0135] During a complete dispatching cycle, the dispatching center collected 8760 hourly water inflow data of a reservoir in the previous year, ranging from 15 to 295 m 3 / s, entering the new year, the dispatching system first uses wavelet noise reduction to process the water inflow time series data from the current day to the previous 60 days. The water inflow for five consecutive days is 32, 28, 25, 23, and 19m respectively. 3 / s, standardized and assigned to drought conditions, the automatically generated standardized reservoir capacity-water level curve is: when the reservoir capacity is 860 million cubic meters, the water level is 314.2 meters; when it is 1.03 billion cubic meters, the water level is 316.1 meters; when it is 1.2 billion cubic meters, the water level is 318.3 meters. The resulting stage division is: egg-pregnancy period from January to March, egg-laying period from April to May, growth period from June to October, and wintering period from November to December.

[0136] In a certain scheduling period, the minimum ecological flow requirements of the ecological window are set to 65, 90, 105, and 52m 3 / s. The system automatically detects that the flow rate from the current 5th to the 8th day is less than 65m 3 / s, triggering an ecological water shortage warning. The algorithm recognizes the probability of extreme drought continuing in the current window as 0.68 through the Markov chain state transition probability, so it automatically increases the inertia weight of the phased particles to 1.38, and injects probability disturbances in the particle swarm optimization iteration to adjust the particle position. At this time, the decision output of some particles is adjusted to a water level of 314.8m and a discharge flow of 67m 3 / s, the system automatically issues a water replenishment scheduling instruction, driving the gate opening range to increase by 2%. The real-time reservoir water level changes by 0.12m, and the water shortage rate drops from the original 8.1% to 1.7%.

[0137] As the ecological window switches to the spawning period, the real-time water inflow increases to 89, 96, 101, 104, and 109 m 3 / s, and the normalized mapping is the normal water-flood state. According to the phase-coupling code, the system automatically adjusts the phase coupling term C to the range of 0.81.1 to ensure a smooth transition of water level and flow decision. During the particle swarm optimization process, the multi-objective weight mapping is 0.62 for ecological weight and 0.38 for water supply weight. After the fitness function is normalized and calculated, the optimal particle solution is 316.3m for water level and 93m for downstream flow. 3 / s, the system issued a gate opening adjustment command, and the monitoring station reported that the downstream ecological flow reached 9599m 3 / s fluctuation, the ecological water shortage duration is 0 hours. The discharge flow fluctuation of the traditional dispatching method during the same period is 85-92m 3 / s, ecological water shortage lasts for 4 hours.

[0138] Entering the growth period, the water flow in the reservoir area fluctuated twice, with the lowest flow rate of 42m 3 / s, maximum flow rate 273m 3 / s, rapid switching between extreme drought and flood conditions. The Markov chain hydrological state matrix is ​​automatically updated, the extreme drought probability perturbation coefficient in the particle swarm algorithm is increased to 1.8, and some particles actively increase the water level decision to 317.6m and reduce the downstream flow to 102m. 3 / s, quickly recover to 109m after refilling water 3 / s. Over the entire cycle, the ecological flow compliance rate was 99.1%, the reservoir water level was below the drought limit for only three hours, and the water supply scheduling error was 0.9%. Under traditional scheduling, the compliance rate was 94.7%, the reservoir water level was below the drought limit for 16 hours, and the water supply scheduling error was 3.4%.

[0139] During an extreme climate event, the water level suddenly dropped to 21-23m for two days. 3 / s, standardized as extreme drought. The system recognizes that the phased particles need to respond quickly in this state, and significantly increases the disturbance term η. After the particle speed jumps, the downstream flow rate is immediately adjusted to 54m 3 / s, the ecological water shortage alarm time is less than 20 minutes, and the dry-up duration after water replenishment scheduling is 0 hours. If traditional scheduling is used, due to the lag in the response of the static scheduling curve, the actual dry-up duration is 6 hours, and ecological monitoring points in some downstream river sections will trigger alarms.

[0140] During the annual scheduling summary phase, the model performs Bayesian self-learning corrections on actual execution deviations. For example, during the peak growth period in October, the system detected that the model predicted a discharge flow of 110m 3 / s, the actual monitoring value averages 108.5m 3 The model automatically fine-tunes the Markov state transition matrix and particle swarm weight parameters. Full-cycle statistics show that using this method achieved a 98.9% annual ecological window discharge compliance rate, a 98.4% annual reservoir water level safety margin, zero ecological flow interruptions during the extreme drought window, and a mean scheduling error of 1.3%. Using traditional methods, these indicators were 92.7%, 93.2%, 3, and 5.6%, respectively.

[0141] Table 1 shows some comparative training samples (taking some scheduling cycles as an example):

[0142]

[0143] Comparing the whole year, after adopting the method of the present invention, the discharge flow increased by an average of 6.3%, the ecological water shortage rate decreased by 5.4 percentage points, the water supply safety margin increased by 4.2 percentage points, the risk of ecological flow interruption in the extreme drought window was reduced to 0, and all scheduling operations achieved automatic closed-loop feedback with minimal execution deviation.

[0144] This Example 1 demonstrates the engineering application advantages of the method of the present invention in the key links of extreme hydrological situation identification, phased scheduling adaptation, ecological window continuity control, particle swarm global-phased collaboration and Markov chain dynamic disturbance mechanism. By controlling the reservoir water level and ecological flow in phases, it not only significantly improves the ecological safety and water supply reliability of reservoir scheduling, but also comprehensively overcomes the core defects of traditional scheduling in extreme event response lag and model-execution gap.

[0145] The present invention deeply integrates the period coupling coding and Markov chain hydrological state transition mechanism into the particle swarm optimization model, constructs the scheduling decision vector of each particle as reservoir water level, ecological flow and period continuity variables, and dynamically adjusts the particle inertia weight and velocity disturbance term according to the actual hydrological state. When the ecological window is in an extreme drought or flood mutation state, the particles can automatically generate jump disturbances according to the hydrological state transition probability, thereby greatly improving the model's adaptability and global search capabilities in the face of uncertain water inflow and extreme climate events, reducing premature convergence and local optimal problems, and realizing coordinated dynamic regulation of the reservoir's drought-limited water level and ecological flow.

[0146] The present invention introduces a phased adaptive multi-objective weight mapping mechanism, which dynamically adjusts the weight coefficient of the dual-objective optimization function in real time according to the actual downstream flow, minimum ecological flow demand and drought-limited water level constraint of each ecological window, and combines it with the historical ecological water shortage risk to realize window sensitivity adjustment. Through phased normalization and variable weights, the algorithm can finely balance the dynamic conflict between ecological protection and reservoir water safety, achieve the optimal matching of phased flow and reservoir water level within the ecological window, improve the ecological protection capacity and subsequent water supply safety under multi-objective phased scheduling, and significantly outperform the existing static weight or global single-objective phased method.

[0147] The present invention directly converts the optimal solution of the particle swarm-Markov chain hybrid model into executable gate opening instructions and real-time reservoir water level settings in stages. Through the linkage of the SCADA system and the reservoir automation control system, a closed-loop scheduling process from model optimization to project execution is formed. The Bayesian self-learning feedback mechanism is introduced in the scheduling execution process. The hydrological state transition probability matrix and particle swarm parameters are corrected in real time based on the actual execution deviation, and continuous rolling optimization is carried out. It significantly reduces the risk of ecological flow and reservoir water level imbalance caused by information lag and execution errors, improves the model's robustness to sudden situations and consecutive extreme years, and realizes the organic unity of intelligent multi-objective phased scheduling of reservoirs and engineering applicability.

[0148] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A reservoir drought limit water level phased ecological flow optimization method, characterized in that: The steps include: S1. Construct an initial parameter set and perform preprocessing to form a standardized hydrological time series and generate an ecological window set; S2. Use standardized hydrological time series to classify drought, normal water, abundant water, and extreme drought conditions, construct a hydrological state set, and calculate the corresponding hydrological state transition probability matrix; S3. Establish a multi-objective optimization problem based on the objectives of minimizing the ecological water shortage rate and minimizing the reservoir water supply risk, and set physical relationship constraints; S4. Construct an improved particle swarm optimization model and use period coding to represent the particle position as the reservoir water level control variable and ecological flow control variable corresponding to the ecological window set; S5. Embed the hydrological state transition probability matrix into the particle velocity update rule, assign the current hydrological state label to each particle, and completely improve the final construction of the particle swarm optimization model; S6. Perform an iterative search on the improved particle swarm optimization model, evaluate the particle fitness based on the multi-objective optimization problem, update the individual particle optimal solution and the global optimal solution of the swarm, adjust the particle position based on the particle's current hydrological state label, and output the optimal particle solution consisting of the reservoir water level control curve and the ecological flow discharge sequence corresponding to the ecological window set; S7. Convert the optimal particle solution into executable control signals of gate opening sequence, diversion unit adjustment instructions and real-time reservoir water level setting value that can be recognized by the reservoir monitoring data acquisition and monitoring system, and execute the control signals through the monitoring data acquisition and monitoring system to implement phased reservoir scheduling.

2. The method for optimizing reservoir drought-limited water level by stages according to claim 1 is characterized in that: The initial parameter set includes historical water inflow data, real-time water inflow data, reservoir capacity-water level curve, drought-limited water level control line parameters, ecological flow demand data and scheduling year phase division data.

3. The method for optimizing reservoir drought-limited water level by stages according to claim 2 is characterized in that: The S2 comprises the following steps: S21. Based on the standardized hydrological time series Q = {q1,q2,...,q T }, where q t represents the standardized inflow flow value at the t-th time step, T represents the total number of time steps, and the hydrological time series is divided into intervals. According to the set hydrological boundary threshold θ={θ1,θ2,θ3}, each q t Mapped to the corresponding hydrological state s t ∈S, we get the hydrological state set S = {D1, N1, W1, ED1}, where D1 represents the drought state, N1 represents the normal water state, W1 represents the abundant water state, ED1 represents the extreme drought state, and θ1, θ2, θ3 are the standardized water inflow thresholds used to divide the boundaries of each hydrological state; S22. Construct the hydrological state transition frequency matrix M=[m ij ] 4×4 , where m ij Indicates that the hydrological state changes from state s i Transfer to state s j The frequency of s i ,s j ∈S; S23. Calculate the hydrological state transition probability matrix P based on the hydrological state transition frequency matrix M = [p ij ] 4×4 , where element p ij Indicates that the hydrological state at time t is s i Under the condition of j probability.

4. A reservoir drought-limited water level phased ecological flow optimization method according to claim 3, characterized in that: The S3 includes the following steps: S31. Calculate the ecological water shortage rate based on the ecological window set. The ecological water shortage rate is used to measure the extent to which the actual outflow within the ecological window does not meet the minimum ecological flow requirement within the ecological window; S32. Calculate the reservoir water supply risk based on the scheduling year time step. The reservoir water supply risk is used to measure the risk of the reservoir water level at each time step in the scheduling year falling below the drought limit water level control line at the corresponding time step; S33. Construct a dual-objective optimization function for the dual-objective optimization problem using weighted normalization based on the ecological water shortage rate and the reservoir water supply risk. Normalize the ecological water shortage rate and the reservoir water supply risk to preset upper and lower bounds, multiply each by a preset weight coefficient, and sum the two weighted results to obtain the value of the dual-objective optimization function. S34. For all time steps in the scheduling year, set a constraint that the reservoir water level is not lower than the drought limit water level control line corresponding to that time step, so that the reservoir water level in each time step is greater than or equal to the drought limit water level control line for that time step; S35. For all ecological windows in the ecological window set, set a constraint that the actual downstream flow rate is not less than the minimum ecological flow rate requirement of the ecological window, so that the actual downstream flow rate of each ecological window is greater than or equal to the minimum ecological flow rate requirement of the ecological window. The actual downstream flow rate is the optimization variable, and the minimum ecological flow rate requirement of the ecological window is the ecological security benchmark. S36. For all time steps in the scheduling year, set a physical relationship constraint between storage capacity and discharge. Based on the storage capacity-water level curve, the reservoir water level at each time step is uniquely mapped to the reservoir storage capacity at that time step. Based on the storage capacity-discharge relationship and the inflow flow at that time step, establish a one-to-one correspondence between the storage capacity, reservoir water level, and discharge flow. S37. Integrate the dual-objective optimization function and all constraints to form the multi-objective optimization problem modeling result.

5. A reservoir drought-limited water level phased ecological flow optimization method according to claim 4, characterized in that: The S4 comprises the following steps: S41. Initialize the particles using the phased-coupling encoding method, and set the phased decision vector X of each particle i Constructed as a set of ecological windows containing the whole year Ω={ω1,ω2,…,ω K } Each window reservoir water level control variable H i,k and ecological flow control variable Q i,k , and introduce the phased cross-coupling term C i,k Reflecting the dynamic connection between scheduling variables in adjacent ecological windows, the particle code is: X i ={[H i,1 ,Q i,1 ,C i,1 ],[H i,2 ,Q i,2 ,C i,2 ],…,[H i,K ,Q i,K ,C i,K ]}; Among them, H i,k represents the reservoir water level control variable of the i-th particle in the k-th ecological window, Q i,k represents the ecological flow control variable of the i-th particle in the k-th ecological window, C i,k is the periodized cross-coupling term; S42. Introduce a phased adaptive inertia weight mechanism, and adjust the particle inertia weight w according to the hydrological state of the current ecological window. i,k Real-time dynamic adjustment: w i,k =w base ·(1+α·R k ); Among them, w base is the basic inertia weight, α is the window sensitivity coefficient, R k represents the historical ecological water shortage risk of the kth ecological window; S43. Based on the current ecological flow control variable Q of each ecological window i,k , minimum ecological flow demand and annual water supply safety margin, and the weight of the multi-objective optimization function λ eco,k ,λ sup,k Mapping to periodic variable coefficients: in, is the minimum ecological flow demand of the kth ecological window, H lim,k is the drought limit water level control line of the kth window; S44. Set the population structure to a global-period two-layer collaborative architecture. That is, the particle swarm global optimal solution not only preserves the optimal scheduling sequence for the entire year, but also provides real-time feedback to each ecological window sub-swarm, achieving collaborative dynamic optimization of the annual and periodic solutions. The periodic sub-swarms evolve independently under the guidance of the global particle swarm, optimizing their response to periodic extreme climate events. S45. The phased-probability jump operator is introduced in the particle velocity update. The phased-probability jump operator combines the hydrological state transition probability matrix P with the current particle phased decision vector X i When the ecological window is in an extreme drought state, the probability disturbance jump is triggered, and the reservoir water level control variable H i,k and ecological flow control variable Q i,k Introduce the probability disturbance term: in, is the new velocity of the kth ecological window of the i-th particle, is the old velocity of the kth ecological window of the i-th particle, V i,k is the kth ecological window speed of the ith particle, is the optimal individual history of the particle, is the global historical optimum, γ is the extreme drought disturbance coefficient, η i,k is the random disturbance that obeys the hydrological state transition probability matrix P.

6. A reservoir drought-limited water level phased ecological flow optimization method according to claim 5, characterized in that: The S5 comprises the following steps: S51. The hydrological state transition probability matrix P = [p ij ] 4×4 Integrating into the particle velocity update mechanism, for each particle kth ecological window, according to the hydrological state set S and the current hydrological state label s i,k , using the hydrological state transition probability Determine the hydrological state migration of the particle in the next evolution step: in, represents the updated hydrological state label of the i-th particle in the k-th ecological window, and Categorical(·) is a random sampling function based on the transition probability; S52. Update the hydrological status label Mapped to the periodic adaptive inertia weight w i,k , disturbance coefficient γ i,k and the probability disturbance term η i,k The selection rule is that when the hydrological status label is updated When the perturbation coefficient of each stage and the amplitude of the probability perturbation term are increased, the particles can increase their jumping ability in the extreme drought window and maintain ecological and safety constraints; S53. The hydrological state transition probability is embedded in the particle velocity update mechanism, and in each iteration, the particle current hydrological state label s is used to update the particle velocity. i,k and its hydrological state transition probability, dynamically adjusting the speed and position of particles; S54. An improved particle swarm optimization model is formed, and the improved particle swarm optimization model adaptively adjusts the speed and position of the particles according to the hydrological state transition probability of each ecological window and the current hydrological state label.

7. A reservoir drought-limited water level phased ecological flow optimization method according to claim 6, characterized in that: The S6 comprises the following steps: S61. Perform iterative search on the improved particle swarm optimization model. In each iteration, for each particle i and ecological window k, the particle velocity updated based on the hydrological state transition probability matrix P is used. and location S62. In each iteration, the current phase decision vector of each particle is calculated based on the multi-objective optimization function and the constraints. Calculate the value of the dual-objective optimization function F i : Among them, λ eco,k and λ sup,k is the weight of the multi-objective optimization function of the kth ecological window, and are the normalized ecological water shortage rate and water supply risk index of the i-th particle in the k-th ecological window; S63. According to the dual objective optimization function value F of each particle i , update the individual historical optimal solution of the particle and the particle swarm historical global optimal solution If the current solution is better than the historical optimal solution, it is updated to the latest value; S64. At the end of each iteration, according to the particle's current hydrological state label s i,k Dynamically adjust the position and velocity of particles based on the hydrological state transition probability matrix P; S65. Determine whether the improved particle swarm optimization model meets the preset convergence conditions, which can be set based on the global optimal solution change threshold, the maximum iteration number, or the objective function convergence criterion; S66. When the convergence conditions are met, output the reservoir water level control curve corresponding to the ecological window set and ecological flow discharge sequence The optimal particle solution composed of 8. A reservoir drought-limited water level phased ecological flow optimization method according to claim 7, characterized in that: The S7 also includes real-time collection of actual reservoir water level data and actual downstream flow data during the scheduling execution process, generating execution result data, and calculating the execution deviation between the execution result data and the optimal particle solution. According to the execution deviation, the hydrological state transition probability matrix is ​​corrected and the improved particle swarm optimization model parameters are adjusted. Then, the process returns to step S5, and rolling optimization is performed based on the corrected hydrological state transition probability matrix and the adjusted particle swarm optimization model parameters to achieve continuous closed-loop control of the coordinated scheduling of the reservoir drought limit water level and the ecological flow in stages.

Images