A risk analysis method for flood control operation of reservoir group considering double uncertainties of model structure and hydrological prediction

Abstract

The application discloses a reservoir group flood control scheduling risk analysis method considering double uncertainties of model structure and hydrological prediction, comprising the following steps: establishing a mixed equivalent model and a full set joint scheduling model, proposing a flood risk definition considering model structure uncertainty and a flood risk definition considering double uncertainties of model structure and hydrological prediction; determining the Bayesian network structure of the mixed equivalent model and the full set joint scheduling model respectively; generating random flood samples, performing Bayesian network parameter learning, and establishing a Bayesian network risk analysis model of the mixed equivalent model and the full set joint scheduling model; and performing probability reasoning based on the Bayesian network risk analysis model, and performing risk prediction and diagnosis. The application considers double uncertainties of model structure and hydrological prediction in real-time flood control mixed scheduling of the reservoir group, can quantitatively describe the influence of the scheduling model structure and hydrological prediction error on the flood risk, and provides a basis for flood control risk decision.

Description

Technical Field

[0001] This invention relates to a method for risk analysis of flood control scheduling of reservoir groups, and in particular to a method for risk analysis of flood control scheduling of reservoir groups that considers both the uncertainty of model structure and hydrological forecast. Background Technology

[0002] In real-time flood control scheduling of reservoir groups, a hybrid equivalent model classifies reservoirs into effective and ineffective reservoirs based on their flood control impact. Joint scheduling of effective reservoirs and separate scheduling of ineffective reservoirs can improve decision-making efficiency by reducing the model structure while ensuring scheduling effectiveness. However, this reduction in model structure may lead to a decrease in flood control benefits and an increase in flood risk.

[0003] Real-time flood control scheduling of reservoir groups is typically affected by various uncertainties, such as hydrological forecasting errors, water level-reservoir capacity curve errors, discharge capacity curve errors, and flood evolution errors, leading to uncertainties in reservoir flood control decisions. Over the past few decades, theories and methods related to flood control system risk analysis have developed rapidly, including return period methods, analytical methods, reliability analysis methods, stochastic simulation methods, and Bayesian network methods. With the gradual expansion of flood control engineering systems, basin flood control scheduling has evolved from single-reservoir optimization scheduling to joint river-reservoir optimization scheduling and then to joint reservoir group optimization scheduling. Correspondingly, research on flood control scheduling risk analysis has also evolved from single-reservoir scheduling risk analysis to reservoir group scheduling risk analysis. However, existing research focuses on risk analysis of single reservoirs or fixed topology flood control systems, neglecting the flood risk caused by the uncertainty of model structure in reservoir group scheduling.

[0004] The reservoir system has a large number of reservoirs and a complex topology. The use of a hybrid equivalent model to schedule the reservoirs divides them into effective reservoirs and ineffective reservoirs. The dynamic combination of the two types of reservoirs further increases the difficulty of solving the joint probability distribution. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide a risk analysis method for flood control scheduling of reservoir groups that considers the dual uncertainties of model structure and hydrological forecasting. This method can quantitatively describe the impact of scheduling model structure and hydrological forecasting errors on flood risk, and provide a basis for flood control risk decision-making.

[0006] Technical Solution: The present invention provides a risk analysis method for flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties, comprising the following steps:

[0007] S1. Establish a hybrid equivalent model and a full set joint scheduling model, and propose a flood risk definition that considers the uncertainty of the model structure and a flood risk definition that considers the dual uncertainty of the model structure and hydrological forecast.

[0008] S2. Select random variables of Bayesian network nodes for the full set joint scheduling model and the hybrid equivalent model respectively, and sort the Bayesian network nodes for the full set joint scheduling model and the hybrid equivalent model respectively. Determine the directed edges according to the direction from the cause node to the result node, and determine the Bayesian network structure of the hybrid equivalent model and the full set joint scheduling model respectively.

[0009] S3. Based on the historical flood frequency analysis, random flood samples are generated using the Latin hypercube sampling method. Bayesian network parameter learning is then performed on the hybrid equivalent model and the full set joint scheduling model to establish a Bayesian network risk analysis model for the hybrid equivalent model and the full set joint scheduling model.

[0010] S4. Probabilistic reasoning is performed on Bayesian network risk analysis models based on hybrid equivalent models and full set joint scheduling models respectively to obtain flood risk considering model structure uncertainty and flood risk considering both model structure and hydrological forecast uncertainty. The factors causing downstream flood control point risk and the probability of each factor are analyzed through diagnostic analysis.

[0011] Furthermore, the objective function of the hybrid equivalent model in step S1 is:

[0012] Single-database optimization scheduling adopts the maximum peak shaving criterion, and its objective function is:

[0013]

[0014] Where T is the number of time periods in the scheduling period, and q(t) is the outflow at time t;

[0015] The optimal scheduling of reservoir groups aims to minimize the maximum flow rate at flood control sections, and its objective function is:

[0016]

[0017] Wherein, q'(i,t) represents the process of calculating the outflow from the i-th reservoir to the public flood control point during the t-th time period; q' D (i,t) represents the process of calculating the interval flow from the i-th reservoir to the private flood control point and then to the public flood control point; This refers to the process of calculating the total inter-regional flow from each reservoir to the public flood control point and then applying it to the public flood control point; M * The number of reservoirs participating in the joint dispatch;

[0018] The objective function of the global joint scheduling model is:

[0019]

[0020] Where M represents the total number of reservoirs in the reservoir group;

[0021] The constraints for both models include: water balance constraints, discharge capacity constraints, maximum reservoir water level constraints, reservoir end-of-period water level constraints, and outflow variation constraints.

[0022] Furthermore, the flood risk definitions in step S1 that consider both model structure uncertainties and hydrological forecast uncertainties are as follows:

[0023] S11. The flood risk at public flood control points is defined as the probability that the actual peak flow exceeds the safe discharge capacity.

[0024]

[0025] Where QS represents the safe discharge capacity of public flood control points; QC m The peak flow rate of a flood event at a public flood control point; P(QC) m >QS) represents the probability that the peak flow at a public flood control point exceeds the safe discharge capacity; f(QC) is the probability that the peak flow at the public flood control point exceeds the safe discharge capacity. m ) for QC m The probability density function;

[0026] S12. Flood risk considering model structural uncertainties is defined as the change in flood risk at common flood control points compared to the hybrid equivalent model and the full-set scheduling model, expressed by the following formula:

[0027] △Risk (1) =Risk (1) -Risk (0)

[0028] Among them, Risk (1) The flood risk generated at public flood control points by the hybrid equivalent model scheduling is calculated using the formula in step S11; Risk (0) The flood risk generated at common flood control points is calculated using the formula in step S11 for scheduling the entire set of flood control models; △Risk (1) Flood risk is considered for the uncertainties in the model structure of the hybrid equivalent model;

[0029] S13. Considering the flood peak forecast error, conduct a flood risk analysis under the dual uncertainties of the hybrid equivalent model real-time scheduling model structure and hydrological forecast; if the flood peak forecast error is ε, then:

[0030]

[0031] Among them, Q m This represents the actual peak flood flow. To predict peak flood flow;

[0032] If the hybrid equivalent model considers the flood risk generated at public flood control points due to flood forecasting errors, denoted as Risk, then... (2)Considering the dual uncertainties of model structure and hydrological forecasting, the flood risk is defined as:

[0033] △Risk (2) =Risk (2) -Risk (0) .

[0034] Furthermore, step S2 includes the following steps:

[0035] S21. The flood control system includes M reservoirs and N flood control sections. The inflow rate of reservoir i is QR. i The outflow from reservoir i is qR i The upstream water flow corresponding to section j is QL. j The flood process at section j is QC j ;

[0036] S22. Sort all M reservoirs and N flood control sections according to the principle of upstream first, then downstream, and tributaries first, then main stream;

[0037] S23. Select the following four random variables as nodes in the Bayesian network of the global joint scheduling model: the inflow peak discharge QR of reservoir i in the global joint scheduling model. mi The peak outflow from reservoir i in the full set of joint scheduling models The peak flood discharge QL of the upstream section corresponding to the flood control section j in the full set of joint scheduling models mj The peak flood discharge at flood control section j in the joint scheduling model of the entire set The following four random variables are selected as nodes in the Bayesian network of the hybrid equivalent model: the peak inflow QR of reservoir i in the hybrid equivalent model. mi The outflow peak discharge of reservoir i in the hybrid equivalent model The peak flood discharge QL of the upstream section corresponding to the flood control section j in the hybrid equivalent model mj Peak flood discharge at flood control section j in the hybrid equivalent model

[0038] S24. Analyze the causal relationships of the random variables in the global joint scheduling model and the hybrid equivalent model. According to the order of reservoirs and flood control sections and the principle of "cause first, effect later" for random variables, sort the Bayesian network nodes of the global joint scheduling model and the hybrid equivalent model respectively.

[0039] S25. Based on the causal relationships of the random variables in the global joint scheduling model and the hybrid equivalent model, determine the directed edges in the direction from the cause node to the result node, and determine the Bayesian network structure of the global joint scheduling model and the hybrid equivalent model respectively.

[0040] Furthermore, step S3 includes the following steps:

[0041] S31. Select L typical historical floods, estimate parameters using the linear moment method, and obtain the probability density distribution function of the peak flood discharge of each reservoir; randomly sample the probability density distribution function using the Latin hypercube sampling method to generate Z random peak discharge samples; obtain a random real-time flood sample set containing Z floods from the random peak discharge samples using the same scaling method. i=1,2,…,M, j=1,2,…,N, v=1,2,…,Z;

[0042] S32, For random real-time flood sample sets The Z-field flood was scheduled using both a hybrid equivalent model and a full-set joint scheduling model. The sample set of floods and scheduling results obtained from the hybrid equivalent model is denoted as . The sample set of floods and scheduling results obtained from the full set joint scheduling model is denoted as Where i = 1, 2, ..., M, j = 1, 2, ..., N, v = 1, 2, ..., Z, For the i-th reservoir and the v-th inflow flood process, For the v-th outflow flood process of the i-th reservoir obtained by the hybrid equivalent model, For the v-th outflow flood process of the i-th reservoir obtained from the global joint scheduling model, For the water coming from the j-th flood control point and the v-th interval, For the j-th flood control point and the v-th flood process obtained from the hybrid equivalent model, This refers to the v-th flood process at the j-th flood control point obtained from the global joint scheduling model.

[0043] S33, from Extract the peak flow rate to obtain Extract the peak flow rate to obtain Where i = 1, 2, ..., M, j = 1, 2, ..., N, v = 1, 2, ..., Z, Let the peak flow of the i-th reservoir during the v-th inflow flood process be denoted as . The peak discharge of the i-th reservoir during the v-th flood event is obtained from the hybrid equivalent model. The peak discharge of the i-th reservoir during the v-th flood event, obtained from the joint scheduling model of the entire set, Let V be the peak flow rate of the incoming water at the j-th flood control point in the v-th interval. The peak flood discharge at the j-th flood control point during the v-th flood process is obtained from the hybrid equivalent model. The peak flood discharge of the j-th flood control point during the v-th flood process is obtained from the joint scheduling model of the entire set;

[0044] S34. Using the equal-width interval method, for and The four random variables in the sample are discretized; the maximum likelihood estimation method is used for parameter learning. If there are O observation samples, denoted as X = {X1, X2, ..., X...} O}, then the likelihood function is:

[0045]

[0046] The parameter estimates are:

[0047]

[0048] Bayesian network risk analysis models with hybrid equivalent models and Bayesian network risk analysis models with full set joint scheduling models are obtained respectively.

[0049] Furthermore, step S4 includes the following steps:

[0050] (41) Based on the Bayesian network of the hybrid equivalent model and the Bayesian network of the full set joint scheduling model, predictive inference is performed. Given the predicted peak inflow and the peak inflow of water in the interval, a predictive inference sample set is formed. i = 1, 2, ..., M, j = 1, 2, ..., N, v* = 1, 2, ..., Z*, where v * Z* represents the index of the predicted sample; Z* represents the total number of predicted samples; based on the predicted inference sample set SF... m Inferring the probability of flood risk in the hybrid equivalent model of downstream flood control points. (1) The probability of flooding in the full-set joint scheduling model (Risk) (0) The flood risk △Risk caused by the uncertainty of the model structure is obtained. (1) =Risk (1) -Risk (0) ;

[0051] (42) Considering the flood forecast error as ε, we obtain the prediction inference sample set considering the forecast error. i = 1, 2, ..., M, j = 1, 2, ..., N, v* = 1, 2, ..., Z*, infer the probability of flood risk occurring in the hybrid equivalent model of downstream flood control points. (2) The flood risk △Risk caused by the dual uncertainties in model structure and hydrological forecast is obtained. (2) =Risk (2) -Risk (0) ;

[0052] (43) Assuming that the downstream flood control point exceeds the control flow, reverse the reasoning to derive the probability density function of the peak flow of its parent node, and diagnose the factors that cause this risk and the probability of each factor.

[0053] The present invention provides a reservoir group flood control scheduling risk analysis system considering both model structure and hydrological forecast uncertainties, comprising:

[0054] The Bayesian network structure construction module is used to select random variables of Bayesian network nodes in the full set joint scheduling model and the hybrid equivalent model, sort the Bayesian network nodes of the full set joint scheduling model and the hybrid equivalent model, determine the directed edges according to the direction from the cause node to the result node, and determine the Bayesian network structure of the hybrid equivalent model and the full set joint scheduling model respectively.

[0055] The Bayesian network risk analysis model building module is used to generate random flood samples using the Latin hypercube sampling method based on historical flood frequency analysis, and to learn the Bayesian network parameters of the hybrid equivalent model and the full set joint scheduling model respectively, thereby establishing the Bayesian network risk analysis model of the hybrid equivalent model and the full set joint scheduling model.

[0056] The risk prediction and analysis module is used to perform probabilistic reasoning on Bayesian network risk analysis models based on hybrid equivalent models and full set joint scheduling models, respectively, to obtain flood risks considering model structure uncertainties and flood risks considering both model structure and hydrological forecast uncertainties. It also analyzes the factors that cause risks to downstream flood control points and the probability of each factor through diagnostic analysis.

[0057] An apparatus of the present invention includes a memory and a processor, wherein:

[0058] Memory is used to store computer programs that can run on a processor;

[0059] The processor is configured to execute, while running the computer program, the steps of the above-described method for risk analysis of flood control scheduling of reservoir groups that considers both model structure and hydrological forecast uncertainties.

[0060] The present invention provides a storage medium storing a computer program, which, when executed by at least one processor, implements the steps of the above-described method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties.

[0061] Beneficial effects: Compared with the prior art, the advantages of the present invention are: (1) The hybrid equivalent model can improve the efficiency of real-time flood control scheduling due to the reduction of the model structure, but may cause greater flood risk. The present invention quantitatively describes the flood risk of the hybrid equivalent model in real-time flood control scheduling of reservoir groups considering the uncertainty of the model structure, which can prove the reliability of the hybrid equivalent model; (2) The present invention comprehensively considers the dual uncertainty factors of model structure and hydrological forecast, and predicts the flood risk of the flood control system under different random flood scenarios, which is of great significance for guiding real-time flood control scheduling and risk decision-making; (3) Under the condition that the risk of the downstream flood control point of the reservoir group is known, the present invention diagnoses the cause of the risk and determines the probability corresponding to each cause, which can provide theoretical support for decision-makers to further carry out risk control. Attached Figure Description

[0062] Figure 1 This is a flowchart of the method of the present invention;

[0063] Figure 2 This is a simplified structural diagram of the reservoir group system of the present invention;

[0064] Figure 3 This is a Bayesian network topology diagram of the present invention, wherein (a) is the Bayesian network topology of the whole-set joint scheduling model, and (b) is the Bayesian network topology of the hybrid equivalent model. Detailed Implementation

[0065] The present invention will now be described in detail with reference to the embodiments and accompanying drawings.

[0066] Bayesian networks have a powerful ability to solve joint probability distributions. Therefore, based on Bayesian network theory, this invention proposes a risk analysis method for flood control scheduling of reservoir groups that considers the dual uncertainties of model structure and hydrological forecast. This method quantitatively analyzes the risks caused by the dual uncertainties of model structure and hydrological forecast, clarifies the causes of the risks, and is of great significance for guiding real-time flood control scheduling and risk decision-making.

[0067] like Figure 1 As shown, the present invention provides a method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties, comprising the following steps:

[0068] S1. Establish a hybrid equivalent model and a full set joint scheduling model, and propose a flood risk definition that considers the uncertainty of the model structure and a flood risk definition that considers the dual uncertainty of the model structure and hydrological forecast.

[0069] This embodiment takes the Yellow River mid-lower reaches five reservoirs joint regulation flood control system as an example. The system includes five reservoirs: Sanmenxia, ​​Xiaolangdi, Guxian, Luhun and Hekoucun, and three flood control sections: Heishiguan, Wuzhi and Huayuankou. Among them, Huayuankou is a public flood control point.

[0070] S11. Establish a hybrid equivalent model with the following objective function:

[0071] Single-database optimization scheduling adopts the maximum peak shaving criterion, and its objective function is:

[0072]

[0073] In the formula, T is the number of time periods (h) in the scheduling period, and q(t) is the outflow rate (m³) at time t. 3 / s).

[0074] The optimal scheduling of reservoir groups aims to minimize the maximum flow rate at flood control sections, and its objective function is:

[0075]

[0076] In the formula, q'(i,t) represents the process of calculating the outflow from the i-th reservoir to the public flood control point during the t-th time period (m 3 / s); q' D (i,t) represents the process of calculating the interval flow from the i-th reservoir to the private flood control point and then to the public flood control point (m). 3 / s); The process of calculating the total inter-regional flow from each reservoir to the public flood control point and then back to the public flood control point (m 3 / s); M * The number of reservoirs participating in the joint scheduling.

[0077] Establish a joint scheduling model for the entire set, with the objective function as follows:

[0078]

[0079] In the formula, M represents the total number of reservoirs in the reservoir group, which is 5 in this embodiment.

[0080] The constraints for the two models are as follows:

[0081] Water balance constraints:

[0082]

[0083] In the formula, V(i,t-1) and V(i,t) represent the initial and final water storage (m³) of the i-th reservoir at time t. 3 ); Q(i,t-1) and Q(i,t) are the initial and final inflow rates (m³) of the i-th reservoir during time period t. 3 / s); q(i,t-1) and q(i,t) are the initial and final outflow rates (m³) of the i-th reservoir during time period t. 3 / s); △t is the duration of the time period.

[0084] Discharge capacity constraints:

[0085] q(i,t)≤q(i,Z(i,t)) (5)

[0086] In the formula, q(i,t) is the outflow from the i-th reservoir at time t (m³ / s). 3 / s); q(i,Z(i,t)) is the discharge capacity (m) of the reservoir at time t corresponding to the water level Z(i,t). 3 / s).

[0087] Reservoir maximum water level constraint:

[0088]

[0089] In the formula, Z(i,t) is the water level (m) of the i-th reservoir at time t; Let t be the maximum allowable water level (m) of the i-th reservoir at time t.

[0090] Reservoir end-of-period water level constraints:

[0091] Z i,end =Z i,e (7)

[0092] In the formula, Z i,end Let Z be the reservoir water level (m) calculated at the end of the i-th reservoir scheduling period; i,e Let be the control water level (m) at the end of the i-th reservoir's scheduling period.

[0093] Outbound flow rate variation constraints:

[0094]

[0095] In the formula, |q(i,t)-q(i,t-1)| represents the variation (m) of the outflow from the i-th reservoir during adjacent time periods. 3 / s); The allowable value for the variation in outflow volume between adjacent time periods (m 3 / s).

[0096] S12, the flood risk at the Huayuankou section of the public flood control point is defined as the probability that the actual peak flow exceeds the safe discharge capacity:

[0097]

[0098] Where QS represents the safe discharge capacity of public flood control points; QC m The peak flow rate of a flood event at a public flood control point; P(QC) m >QS) represents the probability that the peak flow at a public flood control point exceeds the safe discharge capacity; f(QC) is the probability that the peak flow at the public flood control point exceeds the safe discharge capacity. m ) for QC m The probability density function;

[0099] S13. Flood risk considering model structural uncertainties is defined as the change in flood risk at common flood control points compared to the hybrid equivalent model and the full-set scheduling model, expressed by the following formula:

[0100] △Risk (1) =Risk (1) -Risk (0) (10)

[0101] Among them, Risk (1) The flood risk generated at public flood control points by the hybrid equivalent model scheduling is calculated using formula (9); Risk (0) The flood risk generated at public flood control points for the whole-set scheduling model is calculated using formula (9); △Risk (1) Flood risk is considered for the uncertainties in the model structure of the hybrid equivalent model;

[0102] S14. Considering the flood peak forecast error, conduct a flood risk analysis under the dual uncertainties of the hybrid equivalent model real-time scheduling model structure and hydrological forecast; if the flood peak forecast error is ε, then:

[0103]

[0104] Among them, Q m This represents the actual peak flood flow. To predict peak flood flow;

[0105] If the hybrid equivalent model considers the flood risk generated at public flood control points due to flood forecasting errors, denoted as Risk, then... (2) Considering the dual uncertainties of model structure and hydrological forecasting, the flood risk is defined as:

[0106] △Risk (2) =Risk (2) -Risk (0) (12).

[0107] S2. Select random variables of Bayesian network nodes for the full set joint scheduling model and the hybrid equivalent model respectively, and sort the Bayesian network nodes for the full set joint scheduling model and the hybrid equivalent model respectively. Determine the directed edges according to the direction from the cause node to the result node, and determine the Bayesian network structure of the hybrid equivalent model and the full set joint scheduling model respectively.

[0108] S21. The flood control system in this embodiment includes 5 reservoirs and 3 flood control sections, such as... Figure 2 As shown, M=5, N=3; the inflow rate of reservoir i is QR. i The outflow from reservoir i is qR i The upstream water flow corresponding to section j is QL. jThe flood process at section j is QC j ;

[0109] S22. All five reservoirs and three flood control sections are sorted according to the principle of upstream first, then downstream, and tributaries first, then main stream.

[0110] S23. Select the following four random variables as nodes in the Bayesian network of the global joint scheduling model: the inflow peak discharge QR of reservoir i in the global joint scheduling model. mi The peak outflow from reservoir i in the full set of joint scheduling models The peak flood discharge QL of the upstream section corresponding to the flood control section j in the full set of joint scheduling models mj The peak flood discharge at flood control section j in the joint scheduling model of the entire set The following four random variables are selected as nodes in the Bayesian network of the hybrid equivalent model: the peak inflow QR of reservoir i in the hybrid equivalent model. mi The outflow peak discharge of reservoir i in the hybrid equivalent model The peak flood discharge QL of the upstream section corresponding to the flood control section j in the hybrid equivalent model mj Peak flood discharge at flood control section j in the hybrid equivalent model

[0111] S24. Analyze the causal relationships of the random variables in the global joint scheduling model and the hybrid equivalent model. According to the order of reservoirs and flood control sections and the principle of "cause first, effect later" for random variables, sort the Bayesian network nodes of the global joint scheduling model and the hybrid equivalent model respectively.

[0112] S25. Based on the causal relationships of the random variables in the global joint scheduling model and the hybrid equivalent model, determine the directed edges according to the direction from the cause node to the result node, and determine the Bayesian network structure of the global joint scheduling model and the hybrid equivalent model respectively, such as... Figure 3 As shown in (a) and (b).

[0113] S3. Based on the historical flood frequency analysis, random flood samples are generated using the Latin hypercube sampling method. Bayesian network parameter learning is then performed on the hybrid equivalent model and the full set joint scheduling model to establish a Bayesian network risk analysis model for the hybrid equivalent model and the full set joint scheduling model.

[0114] S31. In this embodiment, 20 typical historical floods are selected, and the parameters are estimated using the linear moment method to obtain the probability density distribution function of the peak flood discharge of each reservoir. The probability density distribution function is randomly sampled using the Latin hypercube sampling (LHS) method to generate 1000 random peak discharge samples. Based on the same-scale amplification method, a random real-time flood sample set containing 1000 floods is obtained from these random peak discharge samples. i=1,2,…,5, j=1,2,…,3, v=1,2,…,1000;

[0115] S32, For random real-time flood sample sets The 1000 floods were subjected to both hybrid equivalent model scheduling and full set joint scheduling. The sample set of floods and scheduling results obtained from the hybrid equivalent model is denoted as . The sample set of floods and scheduling results obtained from the full set joint scheduling model is denoted as Where i = 1, 2, ..., 5, j = 1, 2, ..., 3, v = 1, 2, ..., 1000, For the i-th reservoir and the v-th inflow flood process, For the v-th outflow flood process of the i-th reservoir obtained by the hybrid equivalent model, For the v-th outflow flood process of the i-th reservoir obtained from the global joint scheduling model, For the water coming from the j-th flood control point and the v-th interval, For the j-th flood control point and the v-th flood process obtained from the hybrid equivalent model, This refers to the v-th flood process at the j-th flood control point obtained from the global joint scheduling model.

[0116] S33, respectively from and Extract the peak flow rate to obtain and Where i = 1, 2, ..., 5, j = 1, 2, ..., 3, v = 1, 2, ..., 1000, Let the peak flow of the i-th reservoir during the v-th inflow flood process be denoted as . The peak discharge of the i-th reservoir during the v-th flood event is obtained from the hybrid equivalent model. The peak discharge of the i-th reservoir during the v-th flood event, obtained from the joint scheduling model of the entire set, Let V be the peak flow rate of the incoming water at the j-th flood control point in the v-th interval. The peak flood discharge at the j-th flood control point during the v-th flood process is obtained from the hybrid equivalent model. The peak flood discharge of the j-th flood control point during the v-th flood process is obtained from the joint scheduling model of the entire set;

[0117] S34. Using the equal-width interval method, the width is set at 200m. 3 / s, for and The four random variables in the sample are discretized; maximum likelihood estimation (MLE) is used for parameter learning. There are 1000 observation samples, denoted as X = {X1, X2, ..., X...}. 1000}, then the likelihood function is:

[0118]

[0119] The parameter estimates are:

[0120]

[0121] The Bayesian network risk analysis models of the hybrid equivalent model and the Bayesian network risk analysis model of the full set joint scheduling model are obtained respectively, including the parameters of the Bayesian network determined by parameter learning and the model structure determined in step S2.

[0122] S4. Probabilistic reasoning is performed on Bayesian network risk analysis models based on hybrid equivalent models and full set joint scheduling models respectively to obtain flood risk considering model structure uncertainty and flood risk considering both model structure and hydrological forecast uncertainty. The factors causing downstream flood control point risk and the probability of each factor are analyzed through diagnostic analysis.

[0123] S41. In this embodiment, prediction and inference are performed based on the Bayesian network of the hybrid equivalent model and the Bayesian network of the full-set joint scheduling model. Given the predicted inflow peak flow and the interval inflow peak flow, a prediction and inference sample set is formed. i = 1, 2, ..., 5, j = 1, 2, ..., 3, v* = 1, 2, ..., 100, where v * This refers to the index of the predicted sample. Based on the predicted inference sample set SF... m Inferring the probability of flood risk in the hybrid equivalent model of downstream flood control points. (1) The probability of flooding in the full-set joint scheduling model (Risk) (0) The flood risk △Risk caused by the uncertainty of the model structure is obtained. (1) =Risk (1) -Risk (0) ;

[0124] S42. Considering a flood forecast error of 10%, obtain the prediction inference sample set considering the forecast error. Given i = 1, 2, ..., 5, j = 1, 2, ..., 3, v* = 1, 2, ..., 100, infer the probability (Risk) of flooding occurring in the hybrid equivalent model of downstream flood control points. (2) The flood risk △Risk caused by the dual uncertainties in model structure and hydrological forecast is obtained. (2) =Risk (2) -Risk (0) ;

[0125] S43. Assuming the downstream flood control point exceeds the control flow, deduce the probability density function of the peak flow of its parent node in reverse, and diagnose the factors causing this risk and the probability of each factor.

[0126] The present invention provides a reservoir group flood control scheduling risk analysis system considering both model structure and hydrological forecast uncertainties, comprising:

[0127] The Bayesian network structure construction module is used to select random variables of Bayesian network nodes in the full set joint scheduling model and the hybrid equivalent model, sort the Bayesian network nodes of the full set joint scheduling model and the hybrid equivalent model, determine the directed edges according to the direction from the cause node to the result node, and determine the Bayesian network structure of the hybrid equivalent model and the full set joint scheduling model respectively.

[0128] The Bayesian network risk analysis model building module is used to generate random flood samples using the Latin hypercube sampling method based on historical flood frequency analysis, and to learn the Bayesian network parameters of the hybrid equivalent model and the full set joint scheduling model respectively, thereby establishing the Bayesian network risk analysis model of the hybrid equivalent model and the full set joint scheduling model.

[0129] The risk prediction and analysis module is used to perform probabilistic reasoning on Bayesian network risk analysis models based on hybrid equivalent models and full set joint scheduling models, respectively, to obtain flood risks considering model structure uncertainties and flood risks considering both model structure and hydrological forecast uncertainties. It also analyzes the factors that cause risks to downstream flood control points and the probability of each factor through diagnostic analysis.

[0130] An apparatus of the present invention includes a memory and a processor, wherein:

[0131] Memory is used to store computer programs that can run on a processor;

[0132] The processor is configured to execute, while running the computer program, the steps of the above-described method for risk analysis of flood control scheduling of reservoir groups that considers both model structure and hydrological forecast uncertainties.

[0133] The present invention provides a storage medium storing a computer program, which, when executed by at least one processor, implements the steps of the above-described method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties, and achieves the same technical effect as the above method.

[0134] In summary, this invention considers the dual uncertainties of model structure and hydrological forecast in real-time flood control hybrid scheduling of reservoir groups, and proposes a flood risk analysis method based on a Bayesian network model. This method can quantitatively describe the impact of scheduling model structure and hydrological forecast errors on flood risk, providing a basis for flood risk decision-making.

Claims

1. A method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties, characterized in that, Includes the following steps: S1. Establish a hybrid equivalent model and a joint scheduling model for the entire set, and propose flood risk definitions considering both model structure uncertainties and hydrological forecast uncertainties; the objective function of the hybrid equivalent model is: Single-database optimization scheduling adopts the maximum peak shaving criterion, and its objective function is: Where T is the number of time periods in the scheduling period. Let t be the outbound flow rate; The optimal scheduling of reservoir groups aims to minimize the maximum flow rate at flood control sections, and its objective function is: ； in, The process of calculating the outflow from reservoir i to public flood control points during time period t; The process of calculating the flow rate from the i-th reservoir to the private flood control point and then to the public flood control point; The process of calculating the total inter-regional flow from each reservoir to the public flood control point and then applying it to the public flood control point; The number of reservoirs participating in the joint scheduling; The objective function of the global joint scheduling model is: ； Where M represents the total number of reservoirs in the reservoir group; The constraints for the two models include: water balance constraints, discharge capacity constraints, maximum reservoir water level constraints, reservoir end-of-period water level constraints, and outflow variation constraints. The flood risk definitions considering model structure uncertainties and the flood risk definitions considering both model structure and hydrological forecast uncertainties are as follows: S11. The flood risk at public flood control points is defined as the probability that the actual peak flow exceeds the safe discharge capacity. ； in, The safe discharge capacity for public flood control points; The peak flow rate of the flood process at public flood control points; The probability that the peak flow at a public flood control point exceeds the safe discharge capacity; for The probability density function; S12. Flood risk considering model structural uncertainties is defined as the change in flood risk at common flood control points compared to the hybrid equivalent model and the full-set joint scheduling model, expressed by the following formula: ; in, The flood risk generated at public flood control points by the hybrid equivalent model scheduling is calculated by the formula in step S11; The flood risk generated at public flood control points is calculated by the formula in step S11 for scheduling by the full set joint scheduling model; Flood risk is considered for the uncertainties in the model structure of the hybrid equivalent model; S13. Considering the flood peak forecast error, conduct a flood risk analysis under the dual uncertainties of the hybrid equivalent model real-time scheduling model structure and hydrological forecast; if the flood peak forecast error is... ,but: ;in, This represents the actual peak flood flow. To predict peak flood flow; If the hybrid equivalent model considers the flood risk generated at public flood control points due to flood forecasting errors, denoted as […], then… Considering the dual uncertainties of model structure and hydrological forecasting, the flood risk is defined as: ; S2. Select random variables of Bayesian network nodes for the full set joint scheduling model and the hybrid equivalent model respectively, and sort the Bayesian network nodes for the full set joint scheduling model and the hybrid equivalent model respectively. Determine the directed edges according to the direction from the cause node to the result node, and determine the Bayesian network structure of the hybrid equivalent model and the full set joint scheduling model respectively. S3. Based on the historical flood frequency analysis, random flood samples are generated using the Latin hypercube sampling method. Bayesian network parameter learning is then performed on the hybrid equivalent model and the full set joint scheduling model to establish a Bayesian network risk analysis model for the hybrid equivalent model and the full set joint scheduling model. S4. Probabilistic reasoning is performed on Bayesian network risk analysis models based on hybrid equivalent models and full set joint scheduling models respectively to obtain flood risk considering model structure uncertainty and flood risk considering both model structure and hydrological forecast uncertainty. The factors causing downstream flood control point risk and the probability of each factor are analyzed through diagnostic analysis.

2. The method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties as described in claim 1, characterized in that, Step S2 includes the following steps: S21. The flood control system includes M reservoirs and N flood control sections. The inflow rate of reservoir i is... The outflow from reservoir i is The upstream water flow corresponding to section j is The flood process at section j is as follows ; S22. Sort all M reservoirs and N flood control sections according to the principle of upstream first, then downstream, and tributaries first, then main stream; S23. Select the following four random variables as nodes in the Bayesian network of the global joint scheduling model: peak inflow of reservoir i in the global joint scheduling model. The peak outflow from reservoir i in the full set of joint scheduling models The peak flow rate of the upstream section corresponding to the flood control section j in the joint scheduling model of the whole set. The peak flood discharge at flood control section j in the joint scheduling model of the entire set The following four random variables are selected as nodes in the Bayesian network of the hybrid equivalent model: peak inflow discharge of reservoir i in the hybrid equivalent model. The outflow peak discharge of reservoir i in the hybrid equivalent model The peak flow rate of the upstream section corresponding to the flood control section j in the hybrid equivalent model. Peak flood discharge at flood control section j in the hybrid equivalent model ; S24. Analyze the causal relationships of the random variables in the global joint scheduling model and the hybrid equivalent model. According to the order of reservoirs and flood control sections and the principle of "cause first, effect later" for random variables, sort the Bayesian network nodes of the global joint scheduling model and the hybrid equivalent model respectively. S25. Based on the causal relationships of the random variables in the global joint scheduling model and the hybrid equivalent model, determine the directed edges in the direction from the cause node to the result node, and determine the Bayesian network structure of the global joint scheduling model and the hybrid equivalent model respectively.

3. The method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties as described in claim 1, characterized in that, Step S3 includes the following steps: S31. Select L typical historical floods, estimate parameters using the linear moment method, and obtain the probability density distribution function of the peak flood discharge of each reservoir inflow; randomly sample the probability density distribution function using the Latin hypercube sampling method to generate Z random peak discharge samples; obtain a random real-time flood sample set containing Z floods from the random peak discharge samples using the same scaling-up method. , , , ; S32, For random real-time flood sample sets The Z-field flood was scheduled using both a hybrid equivalent model and a full-set joint scheduling model. The sample set of floods and scheduling results obtained from the hybrid equivalent model is denoted as . The sample set of floods and scheduling results obtained from the full set joint scheduling model is denoted as . ,in, , , , For the i-th reservoir and the v-th inflow flood process, For the v-th outflow flood process of the i-th reservoir obtained by the hybrid equivalent model, For the v-th outflow flood process of the i-th reservoir obtained from the global joint scheduling model, For the water coming from the j-th flood control point and the v-th interval, For the j-th flood control point and the v-th flood process obtained from the hybrid equivalent model, This refers to the v-th flood process at the j-th flood control point obtained from the global joint scheduling model. S33, from Extract the peak flow rate to obtain ,from Extract the peak flow rate to obtain ,in, , , , Let the peak flow of the i-th reservoir during the v-th inflow flood process be denoted as . The peak discharge of the i-th reservoir during the v-th flood event is obtained from the hybrid equivalent model. The peak discharge of the i-th reservoir during the v-th flood event, obtained from the joint scheduling model of the entire set, Let V be the peak flow rate of the incoming water at the j-th flood control point in the v-th interval. The peak flood discharge at the j-th flood control point during the v-th flood process is obtained from the hybrid equivalent model. The peak flood discharge of the j-th flood control point during the v-th flood process is obtained from the joint scheduling model of the entire set; S34. Using the equal-width interval method, for and The four random variables in the sample are discretized; the maximum likelihood estimation method is used for parameter learning. If there are O observation samples, denoted as... Then the likelihood function is: ； The parameter estimates are: ； Bayesian network risk analysis models with hybrid equivalent models and Bayesian network risk analysis models with full set joint scheduling models are obtained respectively.

4. The method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties as described in claim 1, characterized in that, Step S4 includes the following steps: (41) Based on the Bayesian network of the hybrid equivalent model and the Bayesian network of the full set joint scheduling model, predictive inference is performed. Given the predicted peak inflow and the peak inflow of water in the interval, a predictive inference sample set is formed. , , , , where v * For the predicted sample number, Represents the total number of predicted samples; infers the sample set based on the prediction. Inferring the probability of flood risk in the hybrid equivalent model of downstream flood control points. The probability of flood risk in the full-set joint scheduling model The flood risk caused by model structural uncertainty was obtained. ; (42) Considering the flood forecast error is This yields a prediction inference sample set that takes into account forecast errors. , , , Inferring the probability of flood risk in the hybrid equivalent model of downstream flood control points. The flood risk caused by the dual uncertainties in model structure and hydrological forecasting was obtained. ; (43) Assuming that the downstream flood control point exceeds the control flow, reverse the reasoning to derive the probability density function of the peak flow of its parent node, and diagnose the factors that cause this risk and the probability of each factor.

5. A reservoir group flood control scheduling risk analysis system considering both model structure and hydrological forecast uncertainties, characterized in that, include: The model construction and risk definition module is used to establish a hybrid equivalent model and a full-set joint scheduling model. It proposes flood risk definitions considering both model structure uncertainties and hydrological forecast uncertainties. The objective function of the hybrid equivalent model is: Single-database optimization scheduling adopts the maximum peak shaving criterion, and its objective function is: Where T is the number of time periods in the scheduling period. Let t be the outbound flow rate; The optimal scheduling of reservoir groups aims to minimize the maximum flow rate at flood control sections, and its objective function is: ； in, The process of calculating the outflow from reservoir i to public flood control points during time period t; The process of calculating the flow rate from the i-th reservoir to the private flood control point and then to the public flood control point; The process of calculating the total inter-regional flow from each reservoir to the public flood control point and then applying it to the public flood control point; The number of reservoirs participating in the joint scheduling; The objective function of the global joint scheduling model is: ； Where M represents the total number of reservoirs in the reservoir group; The constraints for the two models include: water balance constraints, discharge capacity constraints, maximum reservoir water level constraints, reservoir end-of-period water level constraints, and outflow variation constraints. The flood risk definitions considering model structure uncertainties and the flood risk definitions considering both model structure and hydrological forecast uncertainties are as follows: S11. The flood risk at public flood control points is defined as the probability that the actual peak flow exceeds the safe discharge capacity. ； in, The safe discharge capacity for public flood control points; The peak flow rate of the flood process at public flood control points; The probability that the peak flow at a public flood control point exceeds the safe discharge capacity; for The probability density function; S12. Flood risk considering model structural uncertainties is defined as the change in flood risk at common flood control points compared to the hybrid equivalent model and the full-set joint scheduling model, expressed by the following formula: ; in, The flood risk generated at public flood control points by the hybrid equivalent model scheduling is calculated by the formula in step S11; The flood risk generated at public flood control points is calculated by the formula in step S11 for scheduling by the full set joint scheduling model; Flood risk is considered for the uncertainties in the model structure of the hybrid equivalent model; S13. Considering the flood peak forecast error, conduct a flood risk analysis under the dual uncertainties of the hybrid equivalent model real-time scheduling model structure and hydrological forecast; if the flood peak forecast error is... ,but: ;in, This represents the actual peak flood flow. To predict peak flood flow; If the hybrid equivalent model considers the flood risk generated at public flood control points due to flood forecasting errors, denoted as […], then… Considering the dual uncertainties of model structure and hydrological forecasting, the flood risk is defined as: ; The Bayesian network structure construction module is used to select random variables of Bayesian network nodes in the full set joint scheduling model and the hybrid equivalent model, respectively, and sort the Bayesian network nodes of the full set joint scheduling model and the hybrid equivalent model, respectively, and determine the directed edges according to the direction from the cause node to the result node, and determine the Bayesian network structure of the hybrid equivalent model and the full set joint scheduling model respectively. The Bayesian network risk analysis model building module is used to generate random flood samples using the Latin hypercube sampling method based on historical flood frequency analysis, and to learn the Bayesian network parameters of the hybrid equivalent model and the full set joint scheduling model respectively, thereby establishing the Bayesian network risk analysis model of the hybrid equivalent model and the full set joint scheduling model. The risk prediction and analysis module is used to perform probabilistic reasoning on Bayesian network risk analysis models based on hybrid equivalent models and full set joint scheduling models, respectively, to obtain flood risks considering model structure uncertainties and flood risks considering both model structure and hydrological forecast uncertainties. It also analyzes the factors that cause risks to downstream flood control points and the probability of each factor through diagnostic analysis.

6. A device, characterized in that, Includes memory and processor, wherein: Memory is used to store computer programs that can run on a processor; A processor, configured to, while running the computer program, perform the steps of a method for risk analysis of flood control scheduling of reservoir groups considering both model structure and hydrological forecast uncertainties as described in any one of claims 1-4.

7. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by at least one processor, implements the steps of the reservoir group flood control scheduling risk analysis method as described in any one of claims 1-4, considering the dual uncertainties of model structure and hydrological forecast.

Images