Reservoir flood discharge flow simulation calculation method based on LSTM model

Abstract

The invention discloses a reservoir flood discharge flow simulation calculation method based on an LSTM model. The reservoir flood discharge flow simulation calculation method comprises the steps of collecting reservoir historical operation characteristic data; establishing a reservoir flood discharge scheduling model: taking flow matching, operation simplification and operation continuity as optimization targets, introducing constraint conditions, and solving flood discharge parameters of various flood discharge buildings; creating a flood discharge operation label: associating the water level of the reservoir and the inflow with the flood discharge operation; the LSTM model learns a flood discharge operation rule; simulation verification period flood discharge scheduling: simulating reservoir flood discharge operations under different reservoir water levels and inflow conditions in the verification period; and in combination with a weir / hole flood discharge flow formula, simulating by using an LSTM model to obtain an optimal flood discharge parameter and calculating the flood discharge flow. According to the method, an optimization algorithm and a deep learning technology are combined, so that reservoir scheduling is converted into data-driven intelligent scheduling from a traditional rule decision mode, and intelligent, precise and real-time scheduling optimization is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of water conservancy engineering and intelligent scheduling, and specifically relates to a reservoir flood discharge simulation calculation method based on an LSTM model. Background Art

[0002] Reservoir flood discharge scheduling is a core issue in water resources management, involving multiple goals such as reservoir safety, downstream flood control, water supply and ecological protection. Traditional scheduling methods mainly rely on empirical rules or rule bases, but in complex river basin environments, manual scheduling is often difficult to achieve optimal control, and the response to sudden floods is delayed, making it difficult to balance safety and economic benefits. In recent years, with the improvement of hydrological data acquisition capabilities and the development of artificial intelligence technology, intelligent scheduling methods based on deep learning have gradually become a research hotspot. As a neural network suitable for time series data, the long short-term memory (LSTM) model can effectively learn the historical operation mode of reservoir scheduling, providing new ideas for intelligent and automated flood discharge scheduling. Summary of the invention

[0003] The present invention aims to improve the accuracy and intelligence level of reservoir flood discharge prediction, and proposes a reservoir flood discharge flow simulation calculation method based on LSTM model to improve the simulation accuracy of reservoir flood discharge flow.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: A reservoir flood discharge simulation calculation method based on LSTM model, including: S1. Collect historical reservoir operation characteristic data: specifically including reservoir water level, inflow, flood discharge and basic information related to flood discharge scheduling, and pre-process the data, including data cleaning, removing outliers and filling missing data; S2. Establishing a reservoir flood discharge dispatching model: constructing a reservoir flood discharge dispatching model with flow matching, operation simplification and operation continuity as optimization objectives, introducing constraints in combination with reservoir operation characteristics, substituting the pre-processed data in step S1 into the reservoir flood discharge dispatching model and using an optimization algorithm to solve the flood discharge parameters of the flood discharge structure; S3. Create flood discharge operation "labels": Based on the optimized flood discharge parameters, the reservoir water level and inflow are associated with the flood discharge operation, a "label" system for flood discharge operation is constructed, and a data set is constructed, which is divided into a rate period and a verification period; S4. Construct an LSTM model to learn flood discharge operation rules: During the rate period, use reservoir water level and inflow as input to train the LSTM model to learn the mapping relationship between the flood discharge operation "label" and reservoir scheduling; S5. Simulate flood discharge scheduling during the verification period: Based on the trained LSTM model, simulate the reservoir flood discharge operation under different reservoir water levels and inflow conditions during the verification period to obtain the optimal flood discharge parameters; S6. Calculation of flood discharge: Combined with the weir / hole flood discharge formula, the flood discharge is calculated using the optimal flood discharge parameters obtained in step S5.

[0005] Furthermore, the reservoir historical operation characteristic data collected in step S1 are daily-scale data: reservoir water level, inflow, and flood discharge, to ensure the temporal resolution and data consistency of the model. The basic information related to flood discharge scheduling determines the limitations of the constraints.

[0006] Furthermore, the objective functions of the reservoir flood discharge scheduling model in step S2 include flow matching, operation simplification and operation continuity, and the formulas of each objective function are as follows: Traffic matching target f 1: Ensure the simulated daily flood discharge Q total Compared with the measured daily flood discharge Q measured Minimum error: ; Among them, minimize(•) means to minimize an objective function; Q measured To measure the daily flood discharge, and It is the first flood discharge structure i The spillway and the j The gate status of each leak hole, the open state is 1, and the closed state is 0; and It is the first flood discharge structure i The spillway and the j The outflow rate corresponding to the leakage hole is: i and j represent the i-th spillway weir and the j-th spillway hole respectively, m and n represent the total number of weirs and spillways respectively; Operational simplification goals f 2: Control the number of weirs / holes opened every day. When the reservoir has multiple discharge weirs / holes, the fewer weirs / holes are opened, the easier the operation is. Therefore, the "weir / hole number optimization item" is defined as follows: ; Operational Continuity Objectives f3: Ensure the stability of daily flood discharge operations to reduce frequent opening / closing operations. The operation continuity optimization item is introduced by calculating the change value of the weir / hole opening and closing state between the current calculation period and the previous period: ; In the formula, and Respectively t Moment and t-1 Moment i The open and closed status of a spillway weir; and Respectively t Moment and t-1 Moment j The open and closed state of a leak hole.

[0007] The opening of holes on two consecutive days should be kept continuous as much as possible, and the operation of the later day should be based on the operation of the previous day, rather than closing one hole and opening another one.

[0008] Furthermore, the objective functions of flow matching, operation simplification, and operation continuity in step S2 are normalized to obtain , the final objective function f of the reservoir flood discharge scheduling model is the weighted sum of the three normalized objective functions: ; in, and It is a weight coefficient used to balance the relative importance of flow matching and operation simplification and operation continuity objectives, and is determined based on the basic information related to flood discharge scheduling obtained.

[0009] Furthermore, the constraints in step S2 include water balance constraints, water level constraints, discharge capacity constraints, operation sequence constraints and weir / hole opening constraints: Water balance constraints:

[0010] in, V ( t +1) and V ( t ) are t +1 and t The amount of water stored in the reservoir at the time; I ( t )and O ( t ) is the amount of water entering the reservoir (e.g., rainfall, water from upstream, etc.) and the amount of water leaving the reservoir (e.g., power generation, water use downstream, etc.) at time t; Water level constraints: Flood discharge operations should ensure that the water level is within a reasonable range: ; in, H min is the elevation of the top of the weir / hole, or the lowest water level at which the spillway opens. H max Refers to the maximum water level allowed to be reached by the reservoir during flood discharge; H refers to the reservoir water level during flood discharge; Discharge capacity constraints: Avoid excessive flood discharge in a short period of time: ; In the formula, express t The total leakage flow at time, express t The maximum flood discharge allowed at any time; Operation order constraints: Operation order constraints comply with the priority of reservoir operation.

[0011] Opening constraint: The opening of the hole should be within the opening range allowed by the project.

[0012] Furthermore, water balance constraints, water level constraints and discharge capacity constraints are necessary constraints for each reservoir building. Whether the operation sequence constraints and opening constraints are necessary depends on the basic information related to flood discharge scheduling obtained.

[0013] The opening and closing of gates of many reservoir structures have priorities, and most reservoir structures have corresponding dispatching rules, which determine the constraints, which are related to the basic information related to flood discharge dispatch obtained.

[0014] Furthermore, the optimization algorithm used to solve the flood discharge parameters in step S2 is a genetic algorithm, which mainly includes the following steps: Initialization: randomly generate a set of initial solutions of weir / hole opening and closing states and opening degrees as a population; Fitness calculation: Calculate the fitness of each individual according to the objective function. The lower the fitness value, the better the individual. The fitness function is:

[0015] in, f is the final objective function of the reservoir flood discharge operation model; Selection: select individuals to enter the next generation based on their fitness (roulette wheel selection, tournament selection, etc. can be used); Crossover: Perform a crossover operation on the offspring to generate new solutions and select some variables for exchange; Mutation: randomly changing the opening state or flow coefficient of a hole with a small probability to increase population diversity; Termination condition: reaching the preset algebraic or convergence standard and outputting the solution with the best current fitness.

[0016] Furthermore, the “label” of the flood discharge operation in step S3 not only includes the opening and closing status of different flood discharge structures in each calculation period, but also includes their opening settings. This information together constitutes an important feature dataset for training the LSTM model, with 70% of the dataset length as the rate period and the remaining 30% as the validation period.

[0017] Furthermore, the flood discharge formula of a single weir / hole in step S6 is:

[0018] in, C w and C c are the outflow coefficients of the spillway weir and the spillway hole, respectively, which are related to the type of flood discharge structure. B is the width of the spillway weir crest, A is the water flow area of ​​the leakage hole, H d is the water head on the hole, Q w and Q c They are the discharge flows of a single spillway weir and a single spillway hole in the flood discharge structure respectively.

[0019] Beneficial results of the present invention: The present invention combines optimization algorithms and deep learning technology to transform reservoir scheduling from the traditional rule-based decision-making model to data-driven intelligent scheduling. It can automatically learn the optimal flood discharge parameters according to different water levels and inflows, realize adaptive decision-making, reduce the need for human intervention, and improve scheduling efficiency and intelligence.

[0020] The present invention breaks through the limitations of traditional reservoir flood discharge scheduling methods and realizes intelligent, precise and real-time scheduling optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 Schematic diagram of flood discharge structure discharge, where (a) is weir flow and (b) is hole flow; Figure 3 The present invention studies the simulation results of the reservoir flood discharge process line; Figure 4 This is a scatter plot of reservoir flood discharge simulation for the present invention. DETAILED DESCRIPTION

[0022] The technical method of the present invention is further described below in conjunction with the accompanying drawings and specific examples.

[0023] Example

[0024] Taking the Ertan Reservoir in the Yalong River Basin as the research reservoir and the period from January 1, 2002 to December 31, 2018 as the research period, a reservoir flood discharge simulation calculation method based on the LSTM model is as follows: Figure 1 As shown, the specific implementation includes the following steps: Step 1: Collect historical reservoir operation characteristic data: including reservoir water level, inflow, flood discharge and basic information related to flood discharge scheduling, and pre-process the data, including data cleaning, outlier removal and missing data filling; The reservoir water level, inflow and flood discharge data are all daily scale data to ensure the temporal resolution and data consistency of the model. The basic information related to flood discharge scheduling determines the limitations of the constraints.

[0025] The basic information related to flood discharge scheduling includes the record table of flood discharge scheduling (including inflow, outflow, water level and other information), as well as the operating principles of flood discharge structures, including the opening sequence of each hole and operating requirements and other scheduling procedures.

[0026] Step 2: Establish a reservoir flood discharge scheduling model: Taking flow matching, operation simplification, and operation continuity as optimization goals, combined with reservoir operation characteristics, introduce a variety of constraints, and use optimization algorithms to solve the flood discharge parameters of various flood discharge structures.

[0027] The objective function of the reservoir flood discharge operation model mainly includes flow matching, operation simplification, and operation continuity. The weir flow and orifice flow are shown in the figure below. Figure 2 As shown in (a) and (b), H d is the water head on the hole, Q w and Q c They are the flood discharge of a single spillway weir and a single spillway hole respectively.

[0028] The ultimate objective function of the reservoir flood discharge scheduling model is a weighted combination of flow matching, operation simplification, and operation continuity.

[0029] Different flood discharge structures have different focuses. Some buildings tend to emphasize operational continuity, which can reduce hydraulic fluctuations and lower the risk of equipment failure, thus prioritizing safety. Some buildings tend to emphasize operational simplification, reducing the number of operations to save labor costs, but a certain switching frequency must be accepted. Therefore, the specific weights can be determined based on the actual situation of the building.

[0030] Water balance constraint, water level constraint and discharge capacity constraint are necessary constraints for each reservoir structure. Whether the operation sequence constraint and opening constraint are established depends on the operation principle of the flood discharge structure. The operation rules of Ertan Reservoir include that the surface holes must be operated symmetrically, the middle holes are not allowed to operate with partial opening, and the priority of opening and closing the surface holes is higher than that of the middle holes. This embodiment requires the operation sequence constraint and opening constraint.

[0031] Operation order constraints: Operation order constraints comply with the priority of reservoir operation; Opening constraint: opening of the hole (flow coefficient) C i (H) Should comply with the opening range allowed by the project:

[0032] In the formula, C min , C max Respectively represent the minimum / maximum opening range of the hole.

[0033] Genetic algorithm is used to solve the flood discharge parameters of various flood discharge structures. The flood discharge structures of Ertan Reservoir include 7 surface holes, 6 middle holes and 2 flood discharge tunnels.

[0034] Step 3: Create a “label” for flood discharge operations: Based on the optimized flood discharge parameters, associate the reservoir water level and inflow with the flood discharge operation, and build a “label” system for flood discharge operations, dividing the data set into a rate period and a verification period.

[0035] The flood discharge operation “labels” include not only the opening and closing status of different flood discharge structures in each calculation period, but also their opening settings. This information together constitutes an important feature dataset for training the LSTM model, with 70% of the dataset length as the rate period and the remaining 30% as the validation period.

[0036] Step 4: Construct an LSTM model to learn flood discharge operation rules: During the rate period, use the reservoir water level and inflow as input to train the LSTM (Long Short-Term Memory) model to learn the mapping relationship between the flood discharge operation "label" and reservoir scheduling.

[0037] Step 5: Simulate flood discharge scheduling during the verification period: Based on the trained LSTM model, simulate the reservoir flood discharge operation under different water levels and inflow conditions during the verification period to obtain the optimal flood discharge parameters; Step 6: Calculation of flood discharge: The flood discharge formula for a single weir / hole is:

[0038] in, Cw and C c are the outflow coefficients of the weir and the orifice, B is the width of the spillway weir crest, A is the water flow area of ​​the leakage hole, H d is the water head on the hole, Q w and Q c They are the discharge flows of a single spillway weir and a single spillway hole in the flood discharge structure respectively.

[0039] The simulated daily flood discharge of each weir / hole is summed up to obtain the total flood discharge of the reservoir on that day.

[0040] The simulation results are as follows Figure 3 and Figure 4 As shown, Figure 3 NSE refers to the Nash efficiency coefficient, RE refers to the relative error, Figure 4 CalBIAS and Cal R refer to the rate period bias and correlation, respectively, while Val BIAS and Val R refer to the validation period bias and correlation, respectively. The model performs well in the rate period, with a strong trend capture capability, NSE of 0.94, and the simulated values ​​are highly consistent with the observed values. However, the model effect declined in the validation period, showing a large systematic deviation, but the trend consistency between the measured and simulated values ​​is still good.

[0041] The above is only a preferred embodiment of the present invention. The protection scope of the present invention is not limited to the above embodiment. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A reservoir flood discharge simulation calculation method based on LSTM model, characterized in that: include: S1. Collect historical reservoir operation characteristic data: specifically including reservoir water level, inflow, flood discharge and basic information related to flood discharge scheduling, and pre-process the data, including data cleaning, removing outliers and filling missing data; S2. Establishing a reservoir flood discharge dispatching model: constructing a reservoir flood discharge dispatching model with flow matching, operation simplification and operation continuity as optimization objectives, introducing constraints in combination with reservoir operation characteristics, substituting the pre-processed data in step S1 into the reservoir flood discharge dispatching model and using an optimization algorithm to solve the flood discharge parameters of the flood discharge structure; S3. Create flood discharge operation "labels": Based on the optimized flood discharge parameters, the reservoir water level and inflow are associated with the flood discharge operation, a "label" system for flood discharge operation is constructed, and a data set is constructed, which is divided into a rate period and a verification period; S4. Construct an LSTM model to learn flood discharge operation rules: During the rate period, use the reservoir water level and inflow as input to train the LSTM model so that it can learn the mapping relationship between the flood discharge operation "label" and reservoir scheduling; S5. Simulate flood discharge scheduling during the verification period: Based on the trained LSTM model, simulate the reservoir flood discharge operation under different water levels and inflow conditions during the verification period to obtain the optimal flood discharge parameters; S6. Calculation of flood discharge: Calculate the flood discharge flow using the optimal flood discharge parameters obtained in step S5 in combination with the weir / hole flood discharge flow formula.

2. The method for simulating and calculating the flood discharge of a reservoir based on the LSTM model according to claim 1 is characterized in that: The reservoir historical operation characteristic data collected in step S1 are daily scale data.

3. The reservoir flood discharge simulation calculation method based on the LSTM model according to claim 1 is characterized in that: The objective functions of the reservoir flood discharge scheduling model in step S2 include flow matching, operation simplification and operation continuity. The formulas of each objective function are as follows: Traffic matching target f 1: Ensure the simulated daily flood discharge Q total Compared with the measured daily flood discharge Q measured Minimum error: ； Among them, minimize(•) means to minimize an objective function. Q measured To measure the daily flood discharge, and It is the first flood discharge structure i The spillway and the j The gate status of each leak hole, the open state is 1, and the closed state is 0; and It is the first flood discharge structure i The spillway and the j The outflow rate corresponding to the leakage hole is: i and j represent the i-th spillway weir and the j-th spillway hole respectively, m and n represent the total number of weirs and spillways respectively; Operational simplification goals f 2: Control the number of weirs / holes opened every day. When the reservoir has multiple discharge weirs / holes, the fewer weirs / holes are opened, the easier the operation is. Therefore, the "weir / hole number optimization item" is defined as follows: ； Operational Continuity Objectives f 3: Ensure the stability of daily flood discharge operations to reduce frequent opening / closing operations. The operation continuity optimization item is introduced by calculating the change value of the weir / hole opening and closing state between the current calculation period and the previous period: ； In the formula, and Respectively t Moment and t-1 Moment i The open and closed status of a spillway weir; and Respectively t Moment and t-1 Moment j The open and closed state of a leak hole.

4. The method for simulating and calculating the reservoir flood discharge based on the LSTM model according to claim 3 is characterized in that: The objective functions of flow matching, operation simplification, and operation continuity in step S2 are normalized to obtain , the final objective function f of the reservoir flood discharge scheduling model is the weighted sum of the three normalized objective functions: ； in, and It is a weight coefficient used to balance the relative importance of flow matching and operation simplification and operation continuity objectives, and is determined based on the basic information related to flood discharge scheduling obtained.

5. The method for simulating and calculating the reservoir flood discharge based on the LSTM model according to claim 1 is characterized in that: The constraints in step S2 include water balance constraints, water level constraints, discharge capacity constraints, operation sequence constraints, and weir / hole opening constraints: Water balance constraints: ; in, V ( t +1) and V ( t ) are t +1 and t The amount of water stored in the reservoir at the time; I ( t )and O ( t ) is the amount of water entering and leaving the reservoir at time t; Water level constraints: Flood discharge operations should ensure that the water level is within a reasonable range: ； in, H min is the elevation of the top of the weir / hole, or the lowest water level at which the spillway opens. H max It refers to the maximum water level that the reservoir is allowed to reach during flood discharge; H refers to the reservoir water level during flood discharge; Discharge capacity constraints: Avoid excessive flood discharge in a short period of time: ； In the formula, express t The total leakage flow at time, express t The maximum flood discharge allowed at any time; Operation order constraints: Operation order constraints comply with the priority of reservoir operation; Opening constraint: The opening of the hole should be within the opening range allowed by the project.

6. The method for simulating and calculating the flood discharge of a reservoir based on the LSTM model according to claim 5 is characterized in that: Water balance constraints, water level constraints and discharge capacity constraints are necessary constraints for each reservoir building. Whether the operation sequence constraints and opening constraints are necessary depends on the basic information related to flood discharge scheduling obtained.

7. The method for simulating and calculating the reservoir flood discharge based on the LSTM model according to claim 1 is characterized in that: The optimization algorithm used to solve the flood discharge parameters in step S2 is a genetic algorithm, which mainly includes the following steps: Initialization: randomly generate a set of initial solutions of weir / hole opening and closing states and opening degrees as a population; Fitness calculation: Calculate the fitness of each individual according to the objective function. The lower the fitness value, the better the individual. The fitness function is: ； in, f is the final objective function of the reservoir flood discharge operation model; Selection: Selecting individuals to enter the next generation based on their fitness; Crossover: Perform a crossover operation on the offspring to generate new solutions and select some variables for exchange; Mutation: randomly changing the opening state or flow coefficient of a hole with a small probability to increase population diversity; Termination condition: reaching the preset algebraic or convergence standard and outputting the solution with the best current fitness.

8. The method for simulating and calculating the flood discharge of a reservoir based on the LSTM model according to claim 1 is characterized in that: The flood discharge operation "label" in step S3 not only includes the opening and closing status of different flood discharge structures in each calculation period, but also includes their opening settings. This information together constitutes an important feature data set for training the LSTM model, with 70% of the data set length as the rate period and the remaining 30% as the validation period.

9. The method for simulating and calculating the reservoir flood discharge based on the LSTM model according to claim 1, characterized in that: The flood discharge formulas for a single weir / hole in step S6 are: ； in, C w and C c are the outflow coefficients of the weir and the orifice, B is the width of the overflow weir crest, A is the water flow area of ​​the leakage hole, H d is the water head on the hole, Q w and Q c They are the discharge flows of a single spillway weir and a single spillway hole in the flood discharge structure respectively.

Images