A method and system for regulating discharge of large reservoir groups

Abstract

The present disclosure relates to the field of reservoir dispatching technology, and in particular to a method and system for regulating the discharge of a large-scale reservoir group. The present disclosure is based on artificial intelligence technologies such as knowledge graphs to sort out and mine the data relationships and logic in the multidimensional data information involved in the discharge of a large-scale reservoir group, and effectively integrate the physical mechanism of the operation of the reservoir group, so as to make up for the shortcomings of relying solely on artificial intelligence methods, improve the precise decision-making ability of the joint dispatch of the reservoir group, and help to fully improve the comprehensive benefits of the large-scale reservoir group.

Description

Technical Field

[0001] The present invention relates to the technical field of reservoir regulation, and in particular to a method and system for regulating discharge of a large reservoir group. Background Art

[0002] Large-scale reservoir groups are important national infrastructure for flood prevention and disaster reduction and improving the efficiency of water resource utilization. The operation process involves a large amount of data and information. The mutual influence and constraints between reservoirs and different scheduling needs are complex, which brings great challenges to the precise regulation of reservoir groups.

[0003] With the rapid development of information technology, artificial intelligence technology has shown significant advantages in the rapid mining, identification and learning of information, opening up a new situation for the precise regulation of large-scale reservoir groups. As an important technical branch of artificial intelligence, knowledge graphs have strong information mining capabilities and intuitive display effects compared to traditional data scheduling technologies that rely on experience and automated systems. They can process and feedback information more efficiently, and have great potential in improving the management, understanding and response capabilities of multi-dimensional data information of reservoir groups and promoting intelligent scheduling and decision-making of reservoir groups.

[0004] The key to precise regulation of large-scale reservoir groups is to improve the processing capacity and efficiency of multi-dimensional information involved in the decision-making process of reservoir group dispatching. The environment faced by reservoir group dispatching is complex and needs to be supported by multi-dimensional data, which is mainly reflected in the multi-dimensionality of time (hours, days, ten days, etc.), multi-dimensionality of space (spatial distribution of reservoirs) and multi-dimensionality of factors (hydrology, meteorology, working conditions, etc.). These data are often stored in multiple formats, with poor exchange and sharing, and difficult to query and access quickly. Traditional dispatching methods cannot meet the requirements of precise regulation for the correlation degree, retrieval and feedback speed of reservoir operation and related information.

[0005] In addition, due to the lack of in-depth understanding of data relationship analysis and business logic, knowledge graph technology is rarely studied and applied in the field of reservoir scheduling.

[0006] In summary, there is an urgent need for a technical solution for discharge control of large-scale reservoir groups to meet the requirements of precise control on the correlation degree, retrieval and feedback speed of reservoir operation and related information. Summary of the invention

[0007] In view of the above problems, the present invention provides a method and system for regulating discharge of a large reservoir group.

[0008] In a first aspect, a method for regulating discharge of a large reservoir group is provided, the method comprising:

[0009] Collect historical operation data of large reservoir groups;

[0010] Based on historical operation data, a large-scale reservoir group discharge optimization control model is constructed; through the large-scale reservoir group discharge optimization control model, based on historical water conditions and setting different flood season operating water levels, reservoir discharge control is calculated, and an optimization scheme set for gate discharge and gate opening and closing is established;

[0011] Based on historical operation data and optimization scheme sets, a precise discharge knowledge map of large-scale reservoir groups is constructed;

[0012] When there is a control plan corresponding to the real-time water inflow conditions in the knowledge graph, discharge control is performed according to the corresponding control plan;

[0013] When there is no control plan corresponding to the real-time water inflow conditions in the knowledge graph, a reservoir group discharge simulation control model is constructed based on historical operation data, and based on the water inflow conditions and flood season operating water level conditions, a simulation reference plan is obtained through the trained reservoir group discharge simulation control model. Discharge control is carried out according to the simulation reference plan, and the simulation reference plan is added to the knowledge graph.

[0014] Furthermore, historical operation data includes:

[0015] Time, water levels upstream and downstream of the reservoir, inflow, forecast inflow, hydropower station output, outflow, gate flow, abandoned water flow, gate opening and closing status, opening degree, water level at the flood control point in the basin, name of the dispatcher, gate design parameters, discharge hole design parameters, gate layout, reservoir water level-storage capacity relationship curve, reservoir discharge flow-tailwater level relationship curve, discharge capacity curve, basin control area, characteristic storage capacity, characteristic water level, number of units, installed capacity and reservoir dispatching procedures.

[0016] Furthermore, historical operation data of large reservoir groups are collected, including:

[0017] Collect historical original real-time dispatch data;

[0018] Based on the historical original real-time dispatch data, according to the water balance principle, non-negative constraints and abnormal point identification, the data is checked and corrected to obtain calibration data;

[0019] Based on the calibration data, random forest or principal component analysis is used to identify the key operating factors affecting gate discharge.

[0020] Furthermore, key operating elements include: critical time, critical water level requirements and critical flow requirements.

[0021] Furthermore, a large-scale reservoir group discharge optimization control model is constructed, including:

[0022] The first objective function F is constructed based on the minimum and maximum flow of the downstream flood control point 1 , the formula is as follows:

[0023]

[0024] Where: QF t is the flood flow at the flood control point in the tth period, obtained by Muskingum calculation based on the reservoir discharge; n is the total number of periods.

[0025] Furthermore, the construction of a large-scale reservoir group discharge optimization control model also includes:

[0026] The second objective function F is constructed based on the gate minimization of the number of gate opening and closing times and the amplitude of gate opening change. 2 , the formula is as follows:

[0027]

[0028] Where: is the number of times the jth gate of the kth reservoir is opened and closed in the tth period; is the opening degree of the jth gate of the kth reservoir in the tth period; η is a binary function of the gate opening, which takes a value of 0 when the gate has only two modes, open and closed, and takes a value of 1 when there is a flexible opening choice; n, l, and m are the total number of time periods, the total number of gates of a single reservoir, and the total number of reservoirs, respectively.

[0029] Furthermore, the construction of a large-scale reservoir group discharge optimization control model also includes:

[0030] Water balance constraints, storage capacity constraints, flow constraints and gate discharge restrictions.

[0031] Furthermore, the water balance constraint formula is as follows:

[0032]

[0033] In the formula, and are the initial and final storage capacities of the kth reservoir at the tth period, m 3 ; is the inflow of the kth reservoir in the tth period. For water from non-headwater reservoirs, the inflow is the sum of the interval inflow and the discharge of the upstream reservoir. 3 / s; is the reservoir discharge of the kth reservoir at the tth period, which is the sum of the power generation flow qh and the gate discharge flow qf, m 3 / s; Δt is the unit time interval of scheduling; is the evaporation leakage loss in the current period, m 3 .

[0034] Furthermore, the storage capacity constraint formula is as follows:

[0035]

[0036] In the formula, and are the minimum and maximum storage capacity of the kth reservoir in the tth period respectively.

[0037] Furthermore, the flow constraint formula is as follows:

[0038]

[0039] In the formula, is the upper boundary constraint of the discharge of the kth reservoir at the tth period, m 3 / s; is the lower boundary constraint of the discharge of the kth reservoir at the tth period, m 3 / s.

[0040] Furthermore, the gate discharge limit formula is as follows:

[0041]

[0042] In the formula, is the gate discharge of the kth reservoir at the tth period; and They are respectively the minimum and maximum discharge allowed by the gate of the kth reservoir in the tth period.

[0043] Furthermore, through the optimization and control model of large-scale reservoir discharge, based on historical water conditions and different flood season operating water levels, the reservoir discharge control is calculated, and the optimization scheme set of gate discharge and gate opening and closing combination is established, including:

[0044] Taking the reservoir discharge, gate opening and closing times, and gate opening degree as optimization variables, a multi-objective heuristic intelligent algorithm, including NSGA-2 and PA-DDS, is used to set dynamic adjustment plans for operating water levels in different flood seasons. Combined with historical water inflow conditions, the optimization control model is calculated to form a set of plans for the combination of gate discharge and gate opening and closing.

[0045] Furthermore, a reservoir group discharge simulation control model is constructed, including:

[0046] The physical mechanism of reservoir group discharge is established through water balance constraints, monotonicity constraints, reservoir outflow boundary constraints, gate discharge relationship constraints, and water balance equation constraints.

[0047] A neural network is used to construct a reservoir group discharge simulation control model, and the parameters of the reservoir group discharge simulation control model are mapped with the parameters of the reservoir group discharge simulation control model through mapping rules.

[0048] Furthermore, the water balance constraint formula is as follows:

[0049]

[0050] In the formula, and are the initial and final storage capacities of the kth reservoir at the tth period, m 3 ; is the inflow of the kth reservoir in the tth period, m 3 / s; is the discharge of the kth reservoir at the tth period, m 3 / s; Δt is the unit time interval of scheduling; is the evaporation leakage loss in the current period, m 3 .

[0051] Furthermore, the monotonicity constraint formula is as follows:

[0052]

[0053] In the formula, is the reservoir discharge mapping relationship established by the model based on the input variables; Δq is the increase in the reservoir inflow, m 3 / s.

[0054] Furthermore, the reservoir discharge flow boundary constraint formula is as follows:

[0055]

[0056] In the formula, is the upper boundary constraint of reservoir discharge, m 3 / s; is the lower boundary constraint of reservoir discharge, m 3 / s.

[0057] Furthermore, the gate discharge constraint formula is as follows:

[0058]

[0059] In the formula, is the discharge from the gate of the kth reservoir, m 3 / s; μ, ω, Δh are the orifice discharge coefficient, orifice discharge area and orifice center head respectively.

[0060] Furthermore, the water balance equation is as follows:

[0061] According to the current reservoir water level, the forecast inflow and the current power station load, the decision is made to open the gate to discharge the flood and its corresponding gate opening and closing combination. The goal is to control the reservoir discharge flow QO tTo ensure the safety of downstream flooding, the reservoir discharge flow includes two parts: power generation flow qh and gate discharge flow qf. Among them, qh is calculated by the power station load, and qf is calculated by the water balance equation when the gate opening state is uncertain.

[0062] A differential equation is established based on the water balance equation to describe the quantitative relationship between the reservoir water level, flow, load, and gate opening and closing combination. The calculation expression is as follows:

[0063]

[0064] Where: and are the water levels of the kth reservoir in the tth period and the initial period respectively; and are the gate opening and closing combinations of the kth reservoir in the tth period and the initial period respectively; is the power station load; is the inflow of the kth reservoir in the tth period; is the change in reservoir capacity; and are the gate discharge flows of the kth reservoir in the tth period and the initial period, respectively, which depend on the reservoir water level Z and the gate opening and closing state CO.

[0065] Furthermore, the parameters of the reservoir group discharge physical mechanism are mapped to the parameters of the reservoir group discharge simulation control model through mapping rules, including:

[0066]

[0067] Where g(·) is the reservoir discharge mapping function after training with the fusion of physical mechanism and deep learning algorithm; θ is the network parameter vector of the deep learning algorithm; is the discharge of the kth reservoir at the tth period, m 3 / s; is the gate discharge of the kth reservoir at the tth period; is the inflow of the kth reservoir in the tth period, m 3 / s; is the water level of the kth reservoir at the tth period.

[0068] Furthermore, a neural network is used to construct a reservoir group discharge simulation control model, including:

[0069] The loss function formula of the reservoir group discharge simulation control model is as follows:

[0070]

[0071] loss = λ 1 loss QO+λ 2 loss qf +λ 3 loss balance +λ 4 loss monotonicity +λ 5 loss boundary

[0072] In the formula, loss is the overall loss function. The smaller the value, the higher the simulation accuracy. And loss = f(θ), where θ is the network parameter vector of the deep learning algorithm; loss QO is the root mean square error between the simulated value and the theoretical value of the reservoir discharge; loss qf is the root mean square error between the simulated value and the theoretical value of the gate discharge flow; loss balance is the penalty term for violating the long-term water balance constraint of the reservoir discharge flow; loss monotonicity is the penalty term for violating the monotonicity constraint of the reservoir discharge flow; loss boundary is the penalty term for violating the boundary constraint of the reservoir discharge flow, including the upper boundary constraint and the lower boundary constraint; 1 ,λ 2 ,λ 3 ,λ 4 ,λ 5 are the weight coefficients of the root mean square error of the reservoir discharge, the root mean square error of the gate discharge, the water balance constraint penalty term, the monotonicity constraint penalty term, and the boundary constraint penalty term; λ 3 ,λ 4 ,λ 5 No more than two of them can take the value 0; 3 ,λ 4 ,λ 5 No more than two of them can have the value 0.

[0073] Furthermore, loss QO is the loss function of the reservoir discharge simulation accuracy, and the calculation expression is:

[0074]

[0075] Where: is the simulated value of the reservoir discharge; It is the theoretical value of the reservoir discharge (measured value or optimized calculated value).

[0076] Furthermore, loss qf is the loss function of the gate discharge simulation accuracy, and the calculation expression is:

[0077]

[0078] Where: is the simulated value of the gate discharge flow; It is the theoretical value of the reservoir discharge (measured value or optimized calculated value).

[0079] Furthermore, loss balance As the penalty item for violating the long-term water balance constraint of the reservoir outflow, the long-term water balance is checked by calculating the water volume difference between the inflow and the outflow and the change in storage capacity. The calculation expression is:

[0080]

[0081] Where: n is the length of the time series; m is the number of reservoirs; ΔW k is the water volume change of the kth reservoir, m 3 ; Δt is the unit time interval of scheduling.

[0082] Furthermore, loss monotonicity is the penalty term for violating the monotonicity constraint of the reservoir discharge flow. The monotonicity constraint is that when the time series of other input variables remain unchanged, a small increase in the reservoir discharge flow is given, and the simulation result of the reservoir discharge flow corresponding to the corresponding input sample is not less than the simulation result of the original input sequence. The calculation expression is:

[0083]

[0084] Where: The simulated value of the reservoir discharge flow sequence corresponding to the input sample is constructed by assuming that the other input variable time series remain unchanged and giving a small increase in the inflow flow. 3 / s.

[0085] Furthermore, loss boundary is the penalty term for violating the boundary constraint of the reservoir discharge flow. The flow boundary constraint includes the upper and lower boundaries of the reservoir discharge flow. boundary The expression is:

[0086]

[0087] in,

[0088] Where: is the lower limit of the reservoir discharge, m 3 / s; is the upper limit of the reservoir discharge, m 3 / s.

[0089] Furthermore, the evaluation indicators adopted by the reservoir group discharge simulation control model include: Nash efficiency coefficient NSE and root mean square error RMSE.

[0090] Furthermore, the Nash efficiency coefficient formula is as follows:

[0091]

[0092]

[0093] Where: NSE QO is the simulation accuracy of the reservoir discharge; NSE qf The simulation accuracy of the gate discharge flow; is the average value of theoretical reservoir discharge (arithmetic mean of measured value or optimized calculated value); It is the average value of the theoretical gate discharge flow (arithmetic mean of the measured value or optimized calculated value).

[0094] Furthermore, the root mean square error formula is as follows:

[0095]

[0096]

[0097] Where: RMSE QO RMSE is the root mean square error of reservoir discharge simulation qf is the root mean square error of the gate discharge simulation.

[0098] Furthermore, based on historical operation data and optimization scheme sets, a knowledge graph of accurate discharge of large reservoir groups is constructed, including:

[0099] Based on the sorting of historical operation data, a set of past discharge control schemes for the reservoir group is formed; based on the optimization scheme set of gate discharge and gate opening and closing, an optimized dispatching and operation strategy for the gates of the reservoir group is formed; through the construction of the past discharge control scheme set of the reservoir group and the optimized dispatching and operation strategy for the gates of the reservoir group, an accurate discharge knowledge map of the large reservoir group is obtained;

[0100] During the construction process, the professional knowledge of historical operation data, gate discharge and gate opening and closing combination schemes is formally described through RDF or OWL ontology description language to standardize and constrain the effective expression of knowledge elements in the knowledge base constructed by the data layer.

[0101] Furthermore, when there is a control scheme corresponding to the real-time water inflow condition in the knowledge graph, discharge control is performed according to the corresponding control scheme, including:

[0102] By inputting the real-time water inflow conditions and the boundary conditions of the operating water level during the flood season, similarity measurement is performed with the knowledge in the constructed knowledge base to determine whether there is a matching corresponding regulation plan in the knowledge base. If so, the knowledge base is directly used to query the previous discharge regulation plan set of the reservoir group and the optimized dispatching operation strategy of the reservoir group gate.

[0103] In the second aspect, a large-scale reservoir group discharge control system includes: a data collection unit, a discharge optimization control model construction unit and a discharge knowledge graph construction unit;

[0104] Data collection unit, used to collect historical operation data of large reservoir groups;

[0105] The discharge optimization control model construction unit is used to construct a large-scale reservoir group discharge optimization control model based on historical operation data; through the large-scale reservoir group discharge optimization control model, based on historical water conditions and setting different flood season operating water levels, the reservoir discharge control is calculated, and the optimization scheme set of gate discharge and gate opening and closing combination is established;

[0106] The discharge knowledge graph construction unit is used to construct an accurate discharge knowledge graph for large-scale reservoir groups based on historical operation data and optimization scheme sets;

[0107] The discharge knowledge graph construction unit is also used to control the discharge according to the corresponding control scheme when there is a control scheme corresponding to the real-time water inflow condition in the knowledge graph;

[0108] The discharge knowledge graph construction unit is also used to construct a reservoir group discharge simulation control model based on historical operation data when there is no control plan corresponding to the real-time water inflow conditions in the knowledge graph, and to obtain a simulation reference plan through the trained reservoir group discharge simulation control model based on the water inflow conditions and flood season operating water level conditions, to perform discharge control according to the simulation reference plan, and to add the simulation reference plan to the knowledge graph.

[0109] In a third aspect, an electronic device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0110] a memory storing a computer program;

[0111] The processor is used to implement the above-mentioned method for regulating and controlling discharge of a large-scale reservoir group when executing the computer program stored in the memory.

[0112] In a fourth aspect, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned method for regulating discharge of a large-scale reservoir group.

[0113] The present disclosure includes at least the following beneficial effects:

[0114] Based on artificial intelligence technologies such as knowledge graphs, this paper sorts out and mines the data relationships and logic in the multi-dimensional data information involved in the discharge of large-scale reservoir groups, and effectively integrates the physical mechanism of the operation of the reservoir group, to make up for the shortcomings of relying solely on artificial intelligence methods, and improve the precise decision-making ability of the joint dispatching of reservoir groups, which is conducive to fully improving the comprehensive benefits of large-scale reservoir groups.

[0115] Other features and advantages of the present disclosure will be described in the following description, and partly become apparent from the description, or be understood by implementing the present disclosure. The purpose and other advantages of the present disclosure can be realized and obtained by the structures indicated in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0116] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0117] Figure 1 This is a flow chart of the control method according to an embodiment of the present disclosure;

[0118] Figure 2 This is a schematic diagram of the reservoir discharge control model structure;

[0119] Figure 3 It is a schematic diagram for explaining the knowledge graph of reservoir precise discharge;

[0120] Figure 4 This is a schematic diagram of the graph application and update process;

[0121] Figure 5 This is a schematic diagram of the structure of the control system of the embodiment of the present disclosure;

[0122] Figure 6 A schematic diagram of the electronic device structure. DETAILED DESCRIPTION

[0123] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0124] like Figure 1 As shown, the present disclosure provides a method for regulating discharge of a large reservoir group, the method comprising:

[0125] S101, collect historical operation data of large reservoir groups;

[0126] S102, based on historical operation data, construct a large-scale reservoir group discharge optimization control model; through the large-scale reservoir group discharge optimization control model, based on historical water conditions, set different flood season operating water levels to calculate reservoir discharge control, and establish a set of optimization solutions for gate discharge and gate opening and closing combination;

[0127] S103, constructing a knowledge graph of accurate discharge of large-scale reservoir groups based on historical operation data and optimization scheme sets;

[0128] S104, when there is a control scheme corresponding to the real-time water inflow condition in the knowledge graph, discharge control is performed according to the corresponding control scheme;

[0129] S105, when there is no control plan corresponding to the real-time water inflow conditions in the knowledge graph, a reservoir group discharge simulation control model is constructed based on historical operation data, and according to the water inflow conditions and flood season operating water level conditions, a simulation reference plan is obtained through the trained reservoir group discharge simulation control model, and the discharge is regulated according to the simulation reference plan, and the simulation reference plan is added to the knowledge graph.

[0130] The specific implementation is as follows:

[0131] Historical operation data collection and data processing of large-scale reservoir groups: (1) historical operation data verification and cleaning; (2) identification of factors affecting reservoir group discharge; (3) semantic data conversion of reservoir operation and gate control procedures.

[0132] Construction and solution of the optimization control model for discharge from large reservoir groups: (1) objective function; (2) constraints; (3) model solution.

[0133] Reservoir discharge control model integrating physical mechanism and deep learning algorithm: (1) Analysis of the physical mechanism of reservoir operation and gate control; (2) Reservoir discharge control model based on hydrodynamic simulation; (3) Construction and evaluation of reservoir discharge control model integrating physical mechanism and deep learning algorithm.

[0134] Knowledge graph for accurate discharge of large-scale reservoir groups: (1) knowledge graph construction; (2) graph application and update.

[0135] The functions and logical relationships of each step are as follows: S101 is the compilation of historical actual operation plans; S102 is the optimization plan for calculating reservoir discharge control under different flood season operating water levels according to historical water inflow conditions; the above two steps form a database for training simulation models in S105; S105 provides a simulation reference plan through a trained model based on water inflow conditions and flood season operating water level conditions when no corresponding plan is found in the historical knowledge graph library; the visualization, storage, search and application of the discharge operation plan are realized through the technical means of the knowledge graph.

[0136] Historical operation data collection and data processing of large reservoir groups:

[0137] For each reservoir, the following historical operation data need to be collected, including: time (minute level), upstream and downstream water levels of the reservoir, inflow, forecast inflow, hydropower station output (for power generation reservoirs), outflow, gate flow, abandoned water flow, gate opening and closing status and opening degree, etc. The following static data also need to be collected: water level at the flood control point in the basin, name of the dispatcher, gate design parameters, discharge hole design parameters, gate layout, reservoir water level-storage capacity relationship curve, reservoir discharge flow-tailwater level relationship curve, discharge capacity curve, basin control area, characteristic storage capacity, characteristic water level, number of units, installed capacity, reservoir dispatching procedures, etc.;

[0138] (1) Verification, cleaning and influencing factor analysis of historical real-time dispatch data,

[0139] For the original real-time dispatching data collected above, check and correct the data according to the water balance principle, non-negative constraints, abnormal point identification, etc.;

[0140] (2) Identification of factors affecting reservoir discharge,

[0141] Use big data techniques such as random forest or principal component analysis to identify the main operating factors that affect gate discharge;

[0142] (3) Conversion of semantic data of scheduling operation procedures,

[0143] The key time, key water level requirements, and key flow requirements involved in flood control and benefit-enhancing scheduling related to reservoir operation during flood season in the reservoir scheduling and operation regulations are converted into digital scheduling plans to provide constraint boundaries for the reservoir discharge control model.

[0144] Construction and solution of the optimal regulation model for discharge of large reservoir groups:

[0145] (1) Objective function,

[0146] Two major functions are constructed that take into account both flood control safety and gate operation convenience. Among them, flood control safety mainly considers the minimum and maximum flow rates of downstream flood control points, and gate operation convenience mainly considers minimizing the number of opening and closing times and the amplitude of gate opening changes.

[0147] The objective function expression of flood control safety is:

[0148]

[0149]

[0150] Where: QF t is the flood flow at the flood control point in period t, which can be obtained by Muskingum calculation based on the reservoir discharge; is the number of times the jth gate of the kth reservoir is opened and closed in the current period (normalized); is the opening degree of the jth gate of the kth reservoir in the current period; η is a binary function of the gate opening, which takes a value of 0 when the gate has only two modes, open and closed, and takes a value of 1 when there is a flexible opening choice; n, l, and m are the total number of time periods, the total number of gates of a single reservoir, and the total number of reservoirs, respectively.

[0151] (2) Constraints,

[0152] Water balance constraints, reservoir capacity constraints, flow constraints, gate discharge restrictions, etc.

[0153] a) Reservoir water balance constraints:

[0154]

[0155] In the formula, and are the initial and final storage capacities of the kth reservoir at the tth period, m 3 ; is the inflow of the kth reservoir in the tth period. For water from non-headwater reservoirs, the inflow is the sum of the interval inflow and the discharge of the upstream reservoir. 3 / s; is the reservoir discharge of the kth reservoir at the tth period, and is the power generation flow q p and the gate discharge flow q f The sum, m 3 / s; Δt is the unit time interval of scheduling; is the evaporation leakage loss in the current period, m 3 .

[0156] b) Storage capacity constraints:

[0157]

[0158] In the formula, and are the minimum and maximum storage capacity of the kth reservoir in the tth period respectively.

[0159] c) Reservoir flow constraints:

[0160]

[0161] In the formula, is the upper boundary constraint of the discharge of the kth reservoir at the tth period, m 3 / s; is the lower boundary constraint of the discharge of the kth reservoir at the tth period, m 3 / s.

[0162] d) Gate discharge limit:

[0163]

[0164] In the formula, is the gate discharge of the kth reservoir at the tth period; and They are respectively the minimum and maximum discharge allowed by the gate of the kth reservoir in the tth period.

[0165] (3) Model solution

[0166] Taking the reservoir discharge flow, gate opening and closing times, and gate opening degree as optimization variables, a multi-objective heuristic intelligent algorithm (such as NSGA-2, PA-DDS, etc.) is adopted to set dynamic adjustment plans for operating water levels in different flood seasons. Combined with historical water inflow conditions, the above optimization model is calculated to form a set of plans for gate discharge and gate opening and closing combinations.

[0167] Reservoir discharge control model integrating physical mechanism and deep learning algorithm:

[0168] (1) Analysis of the physical mechanism of reservoir discharge,

[0169] For the construction of the knowledge graph, the model has not yet involved the understanding of the reservoir dispatching process, which easily leads to high-value outflow and poor simulation of the decision interval and negative flow problems in the calculation results, which is difficult to be consistent with the basic principles of reservoir dispatching. Therefore, it is necessary to introduce multiple reservoir dispatching physical mechanisms such as water balance principle, monotonicity constraint, reservoir outflow boundary constraint, gate discharge, etc. into the knowledge graph model to improve the model effect.

[0170] a) Water balance principle:

[0171]

[0172] In the formula, and are the initial and final storage capacities of the kth reservoir at the tth period, m 3 ; is the inflow of the kth reservoir in the tth period, m 3 / s; is the discharge of the kth reservoir at the tth period, m 3 / s; Δt is the unit time interval of scheduling; is the evaporation leakage loss in the current period, m 3 .

[0173] b) Monotonicity constraint:

[0174] The monotonicity constraint is that when the time series of other input variables remain unchanged, a small increase in the inflow is given, and the simulation result of the reservoir discharge flow corresponding to the corresponding sample is not less than the simulation result of the original input sequence.

[0175]

[0176] In the formula, is the reservoir discharge mapping relationship established by the model based on the input variables; Δq is the increase in the reservoir inflow, m 3 / s.

[0177] c) Reservoir discharge flow boundary constraints:

[0178]

[0179] In the formula, is the upper boundary constraint of reservoir discharge, m 3 / s; is the lower boundary constraint of reservoir discharge, m 3 / s.

[0180] d) Gate discharge relationship:

[0181]

[0182] In the formula, is the discharge from the gate of the kth reservoir, m 3 / s; μ, ω, Δh are the orifice discharge coefficient, orifice discharge area and orifice center head respectively.

[0183] (2) Reservoir discharge control model,

[0184] According to the current reservoir water level, the forecast inflow and the current power station load, the decision on the gate discharge flow and the corresponding gate opening and closing combination is made. The goal is to control the reservoir discharge flow. To ensure the safety of downstream flooding. The reservoir discharge flow includes two parts: power generation flow qh and gate discharge flow qf. Among them, qh can be calculated from the power station load, and qf can be calculated through the water balance equation when the gate opening state is uncertain.

[0185] According to the water balance equation, a differential equation is established to describe the quantitative relationship between the reservoir water level, flow, load, and gate opening and closing combination. For any reservoir in the reservoir group, the calculation expression is as follows:

[0186]

[0187] Where: and are the water levels of the kth reservoir in the tth period and the initial period respectively; and are the gate opening and closing combinations of the kth reservoir in the tth period and the initial period respectively; is the power station load; is the inflow of the kth reservoir in the tth period; is the change in reservoir capacity; and are the gate discharge flows of the kth reservoir in the tth period and the initial period, respectively, which depend on the reservoir water level Z and the gate opening and closing state CO.

[0188] (3) Construction and evaluation of reservoir discharge control models integrating physical mechanisms and deep learning algorithms,

[0189] In view of the above-mentioned real-time reservoir control mode, in order to quickly solve the discharge flow value in the entire reservoir water level value space, different loads, and different inflow flows. And to solve the problems of flexible and interactive changes in constraints, and changeable and empirical flood control scheduling requirements of decision makers, multiple reservoir discharge physical mechanisms such as water balance principle, monotonicity constraints, and reservoir outflow boundary constraints are introduced into the loss function of deep learning algorithms such as LSTM and CNN to improve the application effect of the model. The model framework is as follows: Figure 2 shown.

[0190] a) Mapping rules:

[0191]

[0192] where g(·) is the reservoir discharge mapping function after training by integrating physical mechanism and deep learning algorithm; θ is the network parameter vector of deep learning algorithm.

[0193] b) Loss function construction:

[0194] The physical mechanism of reservoir operation is integrated into the loss function of the deep learning algorithm (LSTM / CNN). The specific calculation expression is as follows:

[0195] loss = λ 1 loss Q +λ 2 loss qf +λ 3 loss balance +λ 4 loss monotonicity +λ 5 loss boundary

[0196] loss = λ 1 loss QO +λ 2 loss qf +λ 3 loss balance +λ 4 loss monotonicity +λ 5 loss boundary

[0197] In the formula, loss is the overall loss function. The smaller the value, the higher the simulation accuracy. And loss = f(θ), where θ is the network parameter vector of the deep learning algorithm; loss QO is the root mean square error between the simulated value and the theoretical value of the reservoir discharge; loss qf is the root mean square error between the simulated value and the theoretical value of the gate discharge flow; loss balance is the penalty term for violating the long-term water balance constraint of the reservoir discharge flow; loss monotonicity is the penalty term for violating the monotonicity constraint of the reservoir discharge flow; loss boundary is the penalty term for violating the boundary constraint of the reservoir discharge flow, including the upper boundary constraint and the lower boundary constraint; 1 ,λ 2 ,λ 3 ,λ 4 ,λ 5 are the root mean square error of the reservoir discharge, the root mean square error of the gate discharge, the water balance constraint penalty, the monotonicity constraint penalty, and the weight coefficient of the boundary constraint penalty. The value should make the simulated flow as accurate as possible, which is determined by the trial algorithm (such as the grid search algorithm), and λ 3 ,λ 4 ,λ 5 No more than two of them can have the value 0.

[0198] Now let's introduce the loss functions of each sub-item:

[0199] i)loss QOis the loss function of the reservoir discharge simulation accuracy, and the calculation expression is:

[0200]

[0201] Where: is the simulated value of the reservoir discharge; It is the theoretical value of the reservoir discharge (measured value or optimized calculated value).

[0202] ii) loss qf is the loss function of the gate discharge simulation accuracy, and the calculation expression is:

[0203]

[0204] Where: is the simulated value of the gate discharge flow; It is the theoretical value of the reservoir discharge (measured value or optimized calculated value).

[0205] iii) loss balance It is the penalty term for violating the long-term water balance constraint of the reservoir discharge flow. The long-term water balance is verified by calculating the water volume difference between the inflow and the discharge flow and the change in storage capacity. The calculation expression is:

[0206]

[0207] Where: n is the length of the time series; m is the number of reservoirs; ΔW k is the water volume change of the kth reservoir, m 3 ; Δt is the unit time interval of scheduling.

[0208] iv) loss monotonicity is the penalty term for violating the monotonicity constraint of the reservoir discharge flow. The monotonicity constraint is that when the time series of other input variables remain unchanged, a small increase in the reservoir discharge flow is given, and the reservoir discharge flow simulation result corresponding to the corresponding input sample is not less than the simulation result of the original input sequence. The calculation expression is:

[0209]

[0210] Where: The simulated value of the reservoir discharge flow sequence corresponding to the input sample is constructed by assuming that the other input variable time series remain unchanged and giving a small increase in the inflow flow. 3 / s.

[0211] (v)loss boundary is the penalty term for violating the boundary constraint of the reservoir discharge flow. The flow boundary constraint includes the upper and lower boundaries of the reservoir discharge flow. loss boundary The expression is:

[0212]

[0213] in,

[0214] Where: is the lower limit of the reservoir discharge, m 3 / s; is the upper limit of the reservoir discharge, m 3 / s.

[0215] c) Model verification and evaluation,

[0216] The evaluation indicators of model simulation accuracy include: Nash-Sutcliffe efficiency coefficient (NSE) and root mean square error (RMSE).

[0217] i) Nash efficiency coefficient (NSE),

[0218] The closer the Nash efficiency coefficient is to 1, the higher the model simulation accuracy is. The calculation expression is as follows:

[0219]

[0220]

[0221] Where: NSE QO is the simulation accuracy of the reservoir discharge; NSE qf The simulation accuracy of the gate discharge flow; is the average value of theoretical reservoir discharge (arithmetic mean of measured value or optimized calculated value); It is the average value of the theoretical gate discharge flow (arithmetic mean of the measured value or optimized calculated value).

[0222] ii) Root Mean Square Error (RMSE),

[0223] The root mean square error is used to evaluate the degree of discreteness between the simulated flow process and the observed flow process. The calculation expression is as follows:

[0224]

[0225]

[0226] Where: RMSE QO RMSE is the root mean square error of the reservoir discharge simulation qf is the root mean square error of the gate discharge simulation.

[0227] d) Model implementation:

[0228] In model creation, the Pytorch toolkit is used to define the model structure, determine the network structure, number of inputs, number of hidden layers, number of hidden layer nodes and other hyperparameters for the model (taking the LSTM model as an example: the set hyperparameter search range and search interval are as follows: the time step is 1 to 31, and the interval is 1; the training batch value is set in sequence according to the 2n rule, taking n as 5 to 10, and searching at an interval of 1; the number of hidden layer nodes is 32 to 512, and the interval is 32; the number of iterations is 10 to 300, and the interval is 10; the learning rate is 0.00001 to 0.001, and the interval is 0.00001), set the loss function loss and optimizer optimizer, and complete the model creation; in data generation, read the data of different variables, and stack and convert the data; in model hyperparameter optimization, RMSE is used. QO and RMSE qf While minimizing, NSE QO With NSE qf Four objective functions are maximized at the same time, and the hyperparameters are calibrated within the number of iterations based on the training set and validation set data with the help of the SCE-UA algorithm or the NSGA-2 algorithm. In the model test, the network is trained based on the model training set under the calibrated hyperparameters, and the model effect is tested on the model test set. Finally, the model and results are stored.

[0229] Construction and application of knowledge graph of precise discharge of large reservoir groups:

[0230] (1) Knowledge graph construction,

[0231] Considering the actual needs of precise discharge decision-making, the precise discharge expertise is formally described through ontology description languages ​​such as RDF (Resource Description Framework) or OWL (Web Ontology Language) to regulate and constrain the effective expression of knowledge elements in the knowledge base constructed at the data layer. The reservoir precise discharge knowledge graph is explained as follows: Figure 3 According to the knowledge graph, based on the sorting of S101 historical operation data, a set of previous discharge control schemes for the reservoir group is formed; according to the knowledge graph, based on the optimized dispatch of S102, a set of reference schemes for discharge control for the reservoir group and an optimized dispatching operation strategy for the gates of the reservoir group are formed.

[0232] (2) Graph application and update,

[0233] Taking a certain cascade reservoir as the research object, by collecting and organizing basic data and fully sorting out the physical mechanism of dispatching, the real-time regulation rules of large-scale reservoir groups are excavated, and the samples of real-time regulation schemes of large-scale reservoir groups are enriched with the optimized dispatching data during the flood season. The proposed precise discharge knowledge graph and its implementation steps are used to form a knowledge graph library of discharge of large-scale reservoir groups. In actual applications, dispatchers can input real-time water inflow conditions, flood season operating water level boundary conditions, etc., first measure the similarity with the knowledge in the constructed knowledge base, and use manual experience to determine whether there is matching knowledge in the knowledge base. If so, directly use the knowledge base to query the reservoir group's previous discharge control scheme set (reservoir discharge, gate discharge and opening and closing information) and the reservoir group discharge control reference scheme set (the best combination of gate opening and closing and the best operating time, reservoir discharge, etc.); if there is no matching relevant knowledge, the reservoir discharge, gate discharge and opening and closing scheme of each reservoir is calculated through the real-time control model that integrates physical mechanisms and deep learning algorithms. Then, based on the optimized dispatching operation strategy of the reservoir group gate, the knowledge can be supplemented into the existing knowledge base after quality evaluation to enrich and expand the existing knowledge base. The application process is as follows Figure 4 shown.

[0234] like Figure 5 As shown, the present disclosure provides a large-scale reservoir group discharge control system, including: a data collection unit 501, a discharge optimization control model construction unit 502 and a discharge knowledge graph construction unit 503;

[0235] Data collection unit 501, used to collect historical operation data of large reservoir groups;

[0236] The discharge optimization control model construction unit 502 is used to construct a large-scale reservoir group discharge optimization control model based on historical operation data; through the large-scale reservoir group discharge optimization control model, based on historical water conditions, set different flood season operating water levels to calculate reservoir discharge control, and establish an optimization solution set for gate discharge and gate opening and closing combination;

[0237] A discharge knowledge graph construction unit 503 is used to construct an accurate discharge knowledge graph of a large reservoir group based on historical operation data and an optimization solution set;

[0238] The discharge knowledge graph construction unit 503 is also used to perform discharge control according to the corresponding control scheme when there is a control scheme corresponding to the real-time water inflow condition in the knowledge graph;

[0239] The discharge knowledge graph construction unit 503 is also used to construct a reservoir group discharge simulation control model based on historical operation data when there is no control plan corresponding to the real-time water inflow conditions in the knowledge graph, and to obtain a simulation reference plan through the trained reservoir group discharge simulation control model based on the water inflow conditions and the flood season operating water level conditions, to perform discharge control according to the simulation reference plan, and to add the simulation reference plan to the knowledge graph.

[0240] like Figure 6 As shown, the present disclosure provides an electronic device, including a processor 601, a communication interface 602, a memory 603 and a communication bus 604, wherein the processor 601, the communication interface 602 and the memory 603 communicate with each other through the communication bus 604;

[0241] Memory 603, storing computer programs;

[0242] The processor 601 is configured to implement the above method when executing the computer program stored in the memory 603 .

[0243] The present disclosure provides a computer-readable storage medium storing a computer program, wherein the computer program implements the above method when executed by a processor.

[0244] The computer-readable storage medium may be included in the device / apparatus described in the above embodiment; or it may exist independently without being assembled into the device / apparatus. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0245] According to an embodiment of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0246] Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A method for regulating discharge of a large reservoir group, characterized in that: The method comprises: Collect historical operation data of large reservoir groups; Based on historical operation data, a large-scale reservoir group discharge optimization control model is constructed; through the large-scale reservoir group discharge optimization control model, based on historical water conditions and setting different flood season operating water levels, reservoir discharge control is calculated, and an optimization scheme set for gate discharge and gate opening and closing is established; Based on historical operation data and optimization scheme sets, a precise discharge knowledge map of large-scale reservoir groups is constructed; When there is a control plan corresponding to the real-time water inflow conditions in the knowledge graph, discharge control is performed according to the corresponding control plan; When there is no control scheme corresponding to the real-time water inflow conditions in the knowledge graph, a reservoir group discharge simulation control model is constructed based on historical operation data, and a simulation reference scheme is obtained through the trained reservoir group discharge simulation control model according to the water inflow conditions and flood season operation water level conditions. The discharge is regulated according to the simulation reference scheme, and the simulation reference scheme is added to the knowledge graph; Construct a large-scale reservoir group discharge optimization control model, including: The first objective function F1 is constructed based on the minimum and maximum flow of the downstream flood control point, and the formula is as follows: Where: QF t is the flood flow at the flood control point in the tth period, which is obtained by Muskingum calculation based on the reservoir discharge; n is the total number of periods; The second objective function F2 is constructed based on the gate minimization opening and closing times and the amplitude of gate opening change. The formula is as follows: Where: is the number of times the jth gate of the kth reservoir is opened and closed in the tth period; is the opening degree of the jth gate of the kth reservoir in the tth period; η is a binary function of gate opening, which takes a value of 0 when the gate has only two modes, open and closed, and takes a value of 1 when there is a flexible opening choice; n, l, and m are the total number of time periods, the total number of gates of a single reservoir, and the total number of reservoirs, respectively; Construct a reservoir group discharge simulation control model, including: The physical mechanism of reservoir group discharge is established through water balance constraints, monotonicity constraints, reservoir outflow boundary constraints, gate discharge relationship constraints, and water balance equation constraints. A neural network is used to construct a reservoir group discharge simulation control model, and the parameters of the reservoir group discharge simulation control model are mapped with the parameters of the reservoir group discharge simulation control model through mapping rules.

2. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: Historical operation data, including: Time, water levels upstream and downstream of the reservoir, inflow, forecast inflow, hydropower station output, outflow, gate flow, abandoned water flow, gate opening and closing status, opening degree, water level at the flood control point in the basin, name of the dispatcher, gate design parameters, discharge hole design parameters, gate layout, reservoir water level-storage capacity relationship curve, reservoir discharge flow-tailwater level relationship curve, discharge capacity curve, basin control area, characteristic storage capacity, characteristic water level, number of units, installed capacity and reservoir dispatching procedures.

3. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: Collect historical operation data of large reservoir groups, including: Collect historical original real-time dispatch data; Based on the historical original real-time dispatch data, according to the water balance principle, non-negative constraints and abnormal point identification, the data is checked and corrected to obtain calibration data; Based on the calibration data, random forest or principal component analysis is used to identify the key operating factors affecting gate discharge.

4. A method for regulating discharge of a large reservoir group according to claim 3, characterized in that: Key operating elements include: critical time, critical water level requirements and critical flow requirements.

5. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: Constructing a large-scale reservoir group discharge optimization control model, including: Water balance constraints, storage capacity constraints, flow constraints and gate discharge restrictions.

6. A method for regulating discharge of a large reservoir group according to claim 5, characterized in that: The water balance constraint formula is as follows: In the formula, and V t k are the initial and final storage capacities of the kth reservoir at the tth period, m 3 ; is the inflow of the kth reservoir in the tth period. For water from non-headwater reservoirs, the inflow is the sum of the interval inflow and the discharge of the upstream reservoir. 3 / s; is the reservoir discharge of the kth reservoir at the tth period, which is the sum of the power generation flow qh and the gate discharge flow qf, m 3 / s; Δt is the unit time interval of scheduling; is the evaporation leakage loss in the current period, m 3 .

7. A method for regulating discharge of a large-scale reservoir group according to claim 5, characterized in that: The storage capacity constraint formula is as follows: In the formula, and are the minimum and maximum storage capacity of the kth reservoir in the tth period respectively.

8. A method for regulating discharge of a large reservoir group according to claim 5, characterized in that: The flow constraint formula is as follows: In the formula, is the upper boundary constraint of the discharge of the kth reservoir at the tth period, m 3 / s; is the lower boundary constraint of the discharge of the kth reservoir at the tth period, m 3 / s.

9. A method for regulating discharge of a large reservoir group according to claim 5, characterized in that: The gate discharge limit formula is as follows: In the formula, is the gate discharge of the kth reservoir at the tth period; and They are respectively the minimum and maximum discharge allowed by the gate of the kth reservoir in the tth period.

10. A method for regulating discharge of a large-scale reservoir group according to claim 1, characterized in that: Through the large-scale reservoir group discharge optimization control model, based on historical water conditions and different flood season operating water levels, the reservoir discharge control is calculated, and the optimization scheme set of gate discharge and gate opening and closing combination is established, including: Taking the reservoir discharge, gate opening and closing times and gate opening degree as optimization variables, a multi-objective heuristic intelligent algorithm, including NSGA-2 and PA-DDS, is used to set dynamic adjustment plans for operating water levels in different flood seasons. Combined with historical water inflow conditions, the optimization control model is calculated to form a set of plans for gate discharge and gate opening and closing combinations.

11. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: The water balance constraint formula is as follows: In the formula, and V t k are the initial and final storage capacities of the kth reservoir at the tth period, m 3 ; is the inflow of the kth reservoir in the tth period, m 3 / s; is the discharge of the kth reservoir at the tth period, m 3 / s; Δt is the unit time interval of scheduling; is the evaporation leakage loss in the current period, m 3 .

12. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: The monotonicity constraint formula is as follows: In the formula, is the reservoir discharge mapping relationship established by the model based on the input variables; Δq is the increase in the reservoir inflow, m 3 / s.

13. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: The boundary constraint formula of reservoir discharge flow is as follows: In the formula, is the upper boundary constraint of reservoir discharge, m 3 / s; is the lower boundary constraint of reservoir discharge, m 3 / s.

14. A method for regulating discharge of a large-scale reservoir group according to claim 1, characterized in that: The gate discharge constraint formula is as follows: In the formula, qf t k is the discharge from the gate of the kth reservoir, m 3 / s; μ, ω, Δh are the orifice discharge coefficient, orifice discharge area and orifice center head respectively.

15. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: The water balance equation is as follows: A differential equation is established based on the water balance equation to describe the quantitative relationship between the reservoir water level, flow, load, and gate opening and closing combination. The calculation expression is as follows: Where: and are the water levels of the kth reservoir in the tth period and the initial period respectively; and are the gate opening and closing combinations of the kth reservoir in the tth period and the initial period respectively; is the power station load; is the inflow of the kth reservoir in the tth period; is the change in reservoir capacity; is the gate discharge flow of the kth reservoir in the tth period, which depends on the reservoir water level Z and the gate opening and closing state CO, is the gate discharge flow of the kth reservoir in the initial period, which depends on the reservoir water level Z and the gate opening and closing state CO.

16. A method for regulating discharge of a large reservoir group according to claim 1, characterized in that: The parameters of the reservoir group discharge physical mechanism are mapped with the parameters of the reservoir group discharge simulation control model through mapping rules, including: Where g(·) is the reservoir discharge mapping function after training with the fusion of physical mechanism and deep learning algorithm; θ is the network parameter vector of the deep learning algorithm; is the discharge of the kth reservoir at the tth period, m 3 / s;qf t k is the gate discharge of the kth reservoir at the tth period; is the inflow of the kth reservoir in the tth period, m 3 / s; is the water level of the kth reservoir at the tth period.

17. A method for regulating discharge of a large-scale reservoir group according to claim 1, characterized in that: A reservoir group discharge simulation control model is constructed using a neural network, including: The loss function formula of the reservoir group discharge simulation control model is as follows: loss=λ1loss QO +λ2loss qf +λ3loss balance +λ4loss monotonicity +λ5loss boundary In the formula, loss is the overall loss function. The smaller the value, the higher the simulation accuracy. And loss = f(θ), where θ is the network parameter vector of the deep learning algorithm; loss QO is the root mean square error between the simulated value and the theoretical value of the reservoir discharge; loss qf is the root mean square error between the simulated value and the theoretical value of the gate discharge flow; loss balance is the penalty term for violating the long-term water balance constraint of the reservoir discharge flow; loss monotonicity is the penalty term for violating the monotonicity constraint of the reservoir discharge flow; loss boundary is the penalty item for violating the boundary constraint of the reservoir discharge flow, including the upper boundary constraint and the lower boundary constraint; λ1, λ2, λ3, λ4, λ5 are the weight coefficients of the root mean square error of the reservoir discharge flow, the root mean square error of the gate discharge flow, the water balance constraint penalty item, the monotonicity constraint penalty item, and the boundary constraint penalty item respectively; no more than two of λ3, λ4, λ5 can take values ​​of 0.

18. A method for regulating discharge of a large reservoir group according to claim 17, characterized in that: loss QO is the loss function of the reservoir discharge simulation accuracy, and the calculation expression is: Where: is the simulated value of the reservoir discharge; is the theoretical value of the reservoir discharge.

19. A method for regulating discharge of a large reservoir group according to claim 17, characterized in that: loss qf is the loss function of the gate discharge flow simulation accuracy, and the calculation expression is: Where: qf t k,sim (θ) is the simulated value of the gate discharge flow; qftk,obs is the theoretical value of the gate discharge flow.

20. A method for regulating discharge of a large-scale reservoir group according to claim 17, characterized in that: loss balance As the penalty item for violating the long-term water balance constraint of the reservoir outflow, the long-term water balance is checked by calculating the water volume difference between the inflow and the outflow and the change in storage capacity. The calculation expression is: Where: n is the length of the time series; m is the number of reservoirs; ΔW k is the water volume change of the kth reservoir, m 3 ; Δt is the unit time interval of scheduling.

21. A method for regulating discharge of a large reservoir group according to claim 17, characterized in that: loss monotonicity is the penalty term for violating the monotonicity constraint of the reservoir discharge flow, and its calculation expression is: Where: The simulated value of the reservoir discharge flow sequence corresponding to the input sample is constructed by assuming that the other input variable time series remain unchanged and giving a small increase in the inflow flow. 3 / s.

22. A method for regulating discharge of a large reservoir group according to claim 17, characterized in that: loss boundary is the penalty term for violating the boundary constraint of the reservoir discharge flow. The flow boundary constraint includes the upper and lower boundaries of the reservoir discharge flow. boundary The expression is: in, Where: is the lower limit of the reservoir discharge, m 3 / s; is the upper limit of the reservoir discharge, m 3 / s.

23. A method for regulating discharge of a large reservoir group according to claim 17, characterized in that: The evaluation indicators used in the reservoir group discharge simulation control model include: Nash efficiency coefficient NSE and root mean square error RMSE.

24. A method for regulating discharge of a large reservoir group according to claim 23, characterized in that: The Nash efficiency coefficient formula is as follows: Where: NSE QO is the simulation accuracy of the reservoir discharge; NSE qf The simulation accuracy of the gate discharge flow; is the average value of theoretical reservoir discharge; It is the average value of theoretical gate discharge flow.

25. A method for regulating discharge of a large reservoir group according to claim 23, characterized in that: The root mean square error formula is as follows: Where: RMSE QO RMSE is the root mean square error of reservoir discharge simulation qf is the root mean square error of the gate discharge simulation.

26. A method for regulating discharge of a large-scale reservoir group according to claim 1, characterized in that: Based on historical operation data and optimization scheme sets, a precise discharge knowledge graph of large reservoir groups is constructed, including: Based on the sorting of historical operation data, a set of past discharge control schemes for the reservoir group is formed; based on the optimization scheme set of gate discharge and gate opening and closing, an optimized dispatching and operation strategy for the gates of the reservoir group is formed; through the construction of the past discharge control scheme set of the reservoir group and the optimized dispatching and operation strategy for the gates of the reservoir group, an accurate discharge knowledge map of the large reservoir group is obtained; During the construction process, the professional knowledge of historical operation data, gate discharge and gate opening and closing combination schemes is formally described through RDF or OWL ontology description language to standardize and constrain the effective expression of knowledge elements in the knowledge base constructed by the data layer.

27. A method for regulating discharge of a large reservoir group according to claim 26, characterized in that: When there is a control scheme corresponding to the real-time water inflow condition in the knowledge graph, discharge control is performed according to the corresponding control scheme, including: By inputting the real-time water inflow conditions and the boundary conditions of the operating water level during the flood season, similarity measurement is performed with the knowledge in the constructed knowledge base to determine whether there is a matching corresponding regulation plan in the knowledge base. If so, the knowledge base is directly used to query the previous discharge regulation plan set of the reservoir group and the optimized dispatching operation strategy of the reservoir group gate.

28. A large-scale reservoir group discharge control system, characterized in that: include: Data collection unit, discharge optimization and control model construction unit and discharge knowledge graph construction unit; Data collection unit, used to collect historical operation data of large reservoir groups; The discharge optimization control model construction unit is used to construct a large-scale reservoir group discharge optimization control model based on historical operation data; through the large-scale reservoir group discharge optimization control model, based on historical water conditions and setting different flood season operating water levels, the reservoir discharge control is calculated, and the optimization scheme set of gate discharge and gate opening and closing combination is established; The discharge knowledge graph construction unit is used to construct an accurate discharge knowledge graph for large-scale reservoir groups based on historical operation data and optimization scheme sets; The discharge knowledge graph construction unit is also used to control the discharge according to the corresponding control scheme when there is a control scheme corresponding to the real-time water inflow condition in the knowledge graph; The discharge knowledge graph construction unit is also used to construct a reservoir group discharge simulation control model based on historical operation data when there is no control scheme corresponding to the real-time water inflow conditions in the knowledge graph, and obtain a simulation reference scheme through the trained reservoir group discharge simulation control model based on the water inflow conditions and the flood season operation water level conditions, perform discharge control according to the simulation reference scheme, and add the simulation reference scheme to the knowledge graph; Construct a large-scale reservoir group discharge optimization control model, including: The first objective function F1 is constructed based on the minimum and maximum flow of the downstream flood control point, and the formula is as follows: Where: QF t is the flood flow at the flood control point in the tth period, which is obtained by Muskingum calculation based on the reservoir discharge; n is the total number of periods; The second objective function F2 is constructed based on the gate minimization opening and closing times and the amplitude of gate opening change. The formula is as follows: Where: is the number of times the jth gate of the kth reservoir is opened and closed in the tth period; is the opening degree of the jth gate of the kth reservoir in the tth period; η is a binary function of gate opening, which takes a value of 0 when the gate has only two modes, open and closed, and takes a value of 1 when there is a flexible opening choice; n, l, and m are the total number of time periods, the total number of gates of a single reservoir, and the total number of reservoirs, respectively; Construct a reservoir group discharge simulation control model, including: The physical mechanism of reservoir group discharge is established through water balance constraints, monotonicity constraints, reservoir outflow boundary constraints, gate discharge relationship constraints, and water balance equation constraints. A neural network is used to construct a reservoir group discharge simulation control model, and the parameters of the reservoir group discharge simulation control model are mapped with the parameters of the reservoir group discharge simulation control model through mapping rules.

29. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; a memory storing a computer program; The processor is used to implement a large-scale reservoir group discharge control method as described in any one of claims 1-26 when executing a computer program stored in the memory.

30. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a method for regulating discharge of a large reservoir group according to any one of claims 1 to 26 is implemented.

Images