Hydrological forecast interaction method and system fusing multi-source data

Abstract

The invention relates to the field of hydrological forecasting and scheduling, and discloses a hydrological forecasting interaction method and system fusing multi-source data, and the method comprises the following steps: receiving and preprocessing multi-source hydrological data which comprise real-time data, actual measurement data and forecasting data of hydrology and meteorology; constructing a dynamic weight optimization model based on the multi-source data, and calculating the dynamic weight of each data source by optimizing an objective function; performing dynamic fusion on the multi-source data based on the optimized weight to obtain a target forecasting result; a hydrological scheduling optimization model is constructed based on the target forecast result, and an optimal flood scheduling strategy is calculated in combination with a hydrodynamic model; hydrological forecasting and scheduling result files are generated, and automatic publishing of results is supported. According to the method, the multi-source data weight is dynamically adjusted, the scheduling strategy is optimized, the advantages of real-time, actual measurement and forecast data are fully utilized, the method can effectively adapt to the nonlinear change characteristics of a hydrological system, and flood forecast precision and scheduling scientificity are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological forecasting and scheduling, and in particular to a hydrological forecasting interactive method and system for fusing multi-source data. Background Art

[0002] With the continuous development of hydrological monitoring and flood forecasting and dispatching technologies, traditional flood forecasting and dispatching systems face many challenges in terms of accuracy, adaptability and real-time performance. As an important part of river basin management and flood prevention decision-making, flood forecasting requires processing multi-source hydrological data, including real-time data, measured data and forecast data. These data usually come from different sources, have different characteristics, and are inconsistent in time and space distribution, which brings great difficulties to data fusion and optimization.

[0003] In the existing technology, hydrological forecasts are mostly based on a single data source or a simple linear combination method, which lacks in-depth mining of the spatiotemporal characteristics and dynamic correlations between multi-source data, and it is difficult to fully reflect the true state of complex hydrological systems. At the same time, because traditional methods lack the ability to dynamically adjust data weight allocation, forecast results are prone to deviations in the face of sudden hydrological changes or data source uncertainties. In addition, existing flood scheduling methods usually use fixed rules or empirical parameters for calculations, and the consideration of hydrodynamic models is relatively limited, making it difficult to balance the safety and scientificity of scheduling.

[0004] In addition, in terms of the generation and release of results, the existing system has a low degree of automation, lacks a unified and standardized way of outputting results, and has low efficiency in data release and sharing, which affects the timeliness of flood control decisions. These technical deficiencies make it difficult for traditional flood forecasting and scheduling methods to meet the needs of modern river basin management in terms of accuracy, intelligence and efficiency. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a hydrological forecasting interactive method and system that integrates multi-source data, which solves the problems of inaccurate multi-source data fusion, insufficient dynamic correlation, and low scientificity and automation of flood scheduling in the prior art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a hydrological forecast interactive method integrating multi-source data, comprising the following steps:

[0007] Receiving and preprocessing multi-source hydrological data, wherein the multi-source hydrological data includes real-time data, measured data and forecast data;

[0008] Based on multi-source data, a dynamic weight optimization model is constructed to calculate the dynamic weight of each data source by optimizing the objective function;

[0009] Dynamically fuse multi-source data based on optimized weights to obtain target forecast results;

[0010] Based on the target forecast results, a hydrological scheduling optimization model is constructed, and the optimal flood scheduling strategy is calculated in combination with the hydrodynamic model.

[0011] Preferably, the multi-source data includes at least two types of data:

[0012] Real-time data, derived from real-time water level, flow and rainfall at hydrological stations;

[0013] Measured data refers to data collected on-site by instruments, equipment or sensors during flood events, which is then manually analyzed and calculated, rather than directly collected data.

[0014] Forecast data, including future rainfall data and flow prediction results from weather forecasts.

[0015] Preferably, the step of constructing a dynamic weight optimization model based on multi-source data and calculating the dynamic weight of each data source by optimizing the objective function includes:

[0016] Receive multi-source data, including real-time data, measured data and forecast data, and standardize the data to make them consistent in time step, spatial distribution and physical quantity units;

[0017] Construct the optimization objective function of the dynamic weight optimization model. The objective function is used to minimize the deviation between the target forecast value and the weighted result of multi-source data, while optimizing the smoothness of the weight distribution and ensuring that the optimization result conforms to the laws of hydrological physics.

[0018] Set the objective function as:

[0019]

[0020] Where w(t)=[w 1 (t),w 2 (t),…,w m (t)] is the dynamic weight vector, w i (t) is the dynamic weight of the i-th data source, D i (t) is the value of the ith data source, F(t) is the target forecast value, C(t) is the hydrological physical constraint term, α, β, γ are weight control parameters, t 0 and t f Respectively represent the start and end time of the time interval;

[0021] The dynamic weight w(t) is calculated by optimizing the objective function, so that the deviation between the weighted fusion result of multi-source data and the target forecast value is minimized.

[0022] Preferably, the optimization objective function of the dynamic weight optimization model is solved by the following steps:

[0023] Construct a Lagrangian function based on the optimization objective function to constrain the normalization conditions of the weights:

[0024]

[0025] Among them, L(w,λ) is the Lagrangian function, λ is the Lagrangian multiplier, is the weight normalization constraint;

[0026] According to the variational method, the Lagrange equation is solved and expressed as:

[0027]

[0028] The dynamic weight w of each data source is calculated by this equation i (t);

[0029] Combined with numerical solutions, including gradient descent and finite difference methods, the dynamic weights are iteratively solved to ensure that the weights meet the non-negativity and normalization conditions.

[0030] Preferably, the step of dynamically fusing multi-source data based on optimized weights to obtain target forecast results includes:

[0031] For each data source in the multi-source data, obtain its 0 ,,t f ] constitute the data sequence matrix;

[0032] The dynamic weight vector w(t) is used to perform dynamic weighted fusion on multi-source data, and the target prediction result calculation formula is:

[0033]

[0034] Among them, F′(t) is the forecast result at the target time t, w i (t) represents the dynamic weight of the i-th data source, D i (t) represents the value of the i-th data source at time t.

[0035] Preferably, the hydrological scheduling optimization model is implemented by the following steps:

[0036] Construct optimization objectives to minimize the deviation of inflow and outflow and maintain the water level within a safe range;

[0037] The hydrodynamic model is used as a constraint condition, wherein the constraint includes a water conservation equation, a water level change equation, and a flow non-negative constraint;

[0038] The outflow flow curve and reservoir water level change curve are obtained by numerical optimization method.

[0039] Preferably, the hydrodynamic model comprises:

[0040] Water conservation constraint: the difference between inflow and outflow determines the change of reservoir water level;

[0041] Water level limit constraints, the reservoir water level must not exceed the preset maximum water level or fall below the minimum water level;

[0042] Flow non-negativity constraint: outbound flow cannot be negative.

[0043] Preferably, the method further comprises:

[0044] Generate hydrological forecast and scheduling results files, and support automatic release of results.

[0045] Preferably, the generated hydrological forecast and dispatch results file includes:

[0046] Flood forecast process curve, including rising point, flood peak point and corresponding time;

[0047] Scheduling strategy curve, including outflow curve and water level change curve;

[0048] Data error analysis report shows the deviation and trend between the measured value and the predicted value.

[0049] The present invention also provides a hydrological forecast interactive system integrating multi-source data, comprising:

[0050] Data acquisition module, used to receive and process real-time data, measured data and forecast data;

[0051] Dynamic data fusion module, used to calculate the dynamic weight of multi-source data based on the optimization objective function, and fuse them to obtain the target prediction result;

[0052] The scheduling optimization module is used to build a scheduling optimization model based on the hydrodynamic model and generate the optimal scheduling strategy;

[0053] The user interaction module is used to provide an interactive interface, support users to adjust forecast results and scheduling strategies, and provide real-time feedback on the adjusted optimization results;

[0054] The results generation module is used to generate flood forecasting and scheduling results files, and supports the automatic release of results.

[0055] The present invention provides a hydrological forecasting interactive method and system integrating multi-source data, which has the following beneficial effects:

[0056] 1. The present invention constructs a dynamic weight optimization model, which can dynamically adjust the weight according to the characteristics and real-time changes of multi-source data, effectively integrate real-time data, measured data and forecast data, and make full use of the advantages of various types of data to generate more accurate target forecast results. This method can adapt to the nonlinear change characteristics of the hydrological system and make the forecast results more realistic.

[0057] 2. The present invention constructs a hydrological scheduling optimization model based on the conditions of the hydrodynamic model. By comprehensively considering the water conservation relationship, reservoir water level restrictions, and upper and lower flow limits and other constraints, the optimal outflow flow time series and reservoir water level change process are calculated to ensure that the scheduling plan not only meets safety requirements but also achieves efficient use of water resources.

[0058] 3. The present invention solves the inconsistency problem of different data sources in time, space and accuracy through dynamic weight allocation and weighted fusion, ensuring the full utilization and dynamic association of multi-source data. This dynamic fusion method can respond to sudden hydrological changes in a timely manner and improve the adaptability of forecasting and scheduling.

[0059] 4. The present invention can generate standardized achievement documents from hydrological forecast results and scheduling optimization plans, and support automatic release to relevant units. Through automatic document generation and distribution, the workload of manual operations is greatly reduced, the timeliness and accuracy of information transmission are improved, and it is helpful to make rapid flood control decisions.

[0060] 5. The result file generated by the present invention contains detailed contents such as forecast results, dispatch curves, error analysis and data source description, which is convenient for users to fully grasp the forecast and dispatch process. At the same time, the release log record is automatically generated to ensure the traceability and transparency of the results, which facilitates subsequent analysis and auditing.

[0061] 6. The present invention supports dynamic fusion and optimization of scheduling calculations in different regions, and realizes coordinated management of the entire basin by summarizing the results of each sub-basin. This function is particularly suitable for flood scheduling under complex hydrological conditions in large basins, which can effectively reduce the downstream flood peak pressure and improve flood control capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0063] Figure 2 A schematic diagram of a method flow chart of a preferred embodiment of the present invention;

[0064] Figure 3 A schematic diagram of a method flow chart of another preferred embodiment of the present invention;

[0065] Figure 4 It is a schematic diagram of the system structure of the present invention.

[0066] Among them, 10, data acquisition module; 20, dynamic data fusion module; 30, scheduling optimization module; 40, user interaction module; 50, results generation module. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] Please refer to the attached Figure 1 The embodiment of the present invention provides a hydrological forecasting interactive method for fusing multi-source data, which receives and pre-processes multi-source hydrological data, calculates weights based on a dynamic weight optimization model, dynamically fuses the multi-source data to generate a target forecast result, and constructs a hydrological scheduling optimization model based on the result to calculate the optimal flood scheduling strategy. The various steps of the method of the present invention are described in detail below.

[0069] S1. Receive and preprocess multi-source hydrological data

[0070] This step receives multi-source hydrological data and performs preprocessing to ensure the accuracy, consistency and timeliness of data input, and provides a high-quality data foundation for the subsequent construction of dynamic weight optimization models and forecast calculations. Multi-source hydrological data includes real-time data, measured data and forecast data. Data from different sources may differ in terms of collection methods, time resolution and physical units, so unified preprocessing operations are required to meet the requirements of model calculations.

[0071] In some specific implementations of the present invention, the process of receiving and preprocessing multi-source hydrological data involves data collection, standardization, outlier detection, spatiotemporal alignment, and data matrix construction.

[0072] In this embodiment, the following three types of hydrological data are first received:

[0073] Real-time data: including water level, flow and rainfall data, which are collected in real time by sensors at hydrological stations. For example, the real-time flow rate of a river upstream control station at a certain moment is 500m 3 / s, water level is 10m and rainfall is 15mm.

[0074] Measured data: Data from flood events that are collected on-site using instruments, equipment, or sensors, and then manually analyzed and calculated, rather than directly collected data. For example, flow data calculated by using equipment to measure flow velocity.

[0075] Forecast data: including future rainfall, weather forecast data and flow forecast values ​​from short-term hydrological models. For example, the rainfall forecast for the next 48 hours shows that the upstream catchment area is expected to receive 100 mm of rainfall, corresponding to a flow increase of 300 m 3 / s.

[0076] In a possible implementation, the above data is collected through different sensor interfaces, data storage interfaces, and forecast model interfaces, and stored in a unified data structure.

[0077] In this embodiment, in order to ensure that multi-source data can be used for model calculation, the collected data needs to be preprocessed as follows:

[0078] Generally, the time step and spatial resolution of the data may be inconsistent, and the first step of preprocessing is to standardize the data. Specifically, it includes the following operations:

[0079] The flow data is uniformly converted into units of m 3 / s, rainfall converted to mm / h, water level converted to m;

[0080] The temporal resolution is adjusted to a consistent time step, for example, rainfall data sampled at 10 minutes and water level data sampled at 1 hour are unified to a 1-hour resolution;

[0081] Interpolation is used to fill missing values. For example, if a station is missing 10 minutes of flow data due to sensor failure, linear interpolation can be used to calculate the values ​​in between.

[0082] As an option, in order to improve the reliability of data, outlier detection and cleaning are also required. In this embodiment, if the observed value exceeds the reasonable range, it is determined to be an outlier. For example, when the traffic of a certain site suddenly jumps to 10000m 3 / s, which obviously exceeds the site's historical highest traffic of 2000m 3 / s, it can be marked as an outlier. Specific anomaly detection methods include:

[0083] Compare the historical maximum value with the current data value;

[0084] Statistical methods, such as the 3-times standard deviation principle, were used to eliminate outliers.

[0085] In some embodiments, in order to enhance the applicability of the model, it is also necessary to perform spatiotemporal matching on the data. Specifically:

[0086] According to the geographical distribution of the watershed, real-time observation data and historical record data are mapped to a unified spatial grid. For example, the observation values ​​of multiple rain gauges in a catchment area are weighted averaged to obtain the comprehensive rainfall in the area;

[0087] Align data in time series so that each data source has consistent values ​​in the same time step. For example, the time alignment of real-time flow observations and rainfall forecasts can be accomplished through data interpolation.

[0088] As a specific implementation method, the preprocessed data is finally stored as a multi-source data matrix D(t), which is in the form of:

[0089]

[0090] in:

[0091] D i (t j ) represents the data value of the i-th data source at time j;

[0092] m represents the total number of data sources;

[0093] n represents the total length of the time series.

[0094] In one possible implementation, the data matrix is ​​divided into a real-time observation part, a historical record part, and a prediction part to facilitate the input of the subsequent dynamic weight optimization model.

[0095] In addition, in some embodiments, uncertainty analysis of future forecast data may also be required. Specifically, multiple possible scenarios of future forecast data may be generated by Monte Carlo simulation method, for example:

[0096] Add a ±10% fluctuation range to the rainfall in the next 48 hours to calculate the flow changes under different scenarios;

[0097] Combined with historical flood process data, the probability distribution of future forecast results is estimated.

[0098] Through the above-mentioned preprocessing steps, the present invention ensures the reliability, consistency and timeliness of data input, laying a solid foundation for the subsequent construction of the dynamic weight optimization model and the generation of target forecast results.

[0099] S2. Build a dynamic weight optimization model and calculate weights

[0100] This step constructs an optimization objective function and calculates the dynamic weight of each data source to ensure that the fusion result can reflect the target forecast value as accurately as possible. The dynamic weight optimization model combines mathematical optimization methods with physical constraints to ensure that the weighted fusion results of multi-source data not only have high fitting accuracy, but also meet the requirements of hydrological physical laws.

[0101] In general, the characteristics of multi-source data, such as real-time, historical relevance, and prediction accuracy, vary at different time points, so it is necessary to dynamically adjust the weights to adapt to the time-varying characteristics. In the embodiments of the present invention, by constructing an optimization objective function and combining constraints, the accuracy and dynamics of weight calculation are solved.

[0102] In this embodiment, we first define a multi-source data matrix D(t)=[D 1 (t),D 2 (t),…,D m (t)], where D i (t) represents the value of the i-th data source at time t. The goal is to use the dynamic weight vector w(t) = [w 1 (t),w 2 (t),…,w m (t)], calculate the weighted fusion result F'(t) to make it close to the target prediction value F(t).

[0103] In one possible implementation, the optimization objective function is defined as follows:

[0104]

[0105] Specifically, the above objective function includes the following three parts:

[0106] Part I It is the error term between the weighted fusion result and the target prediction value, which is used to minimize the deviation between the fusion result and the target prediction value;

[0107] Part 2 is the entropy regularization term, which is used to limit the drastic changes in dynamic weights and ensure a smooth weight distribution;

[0108] The third part γC(t) is the physical constraint term, where C(t) includes the physical laws of the hydrological system, such as water conservation conditions and upper and lower flow limits.

[0109] As an option, in order to ensure the rationality of the weights, the dynamic weight vector needs to meet the following constraints:

[0110] Non-negativity of weights:

[0111]

[0112] Non-negativity of weights:

[0113]

[0114] In a possible implementation, in order to solve the above optimization problem, a Lagrangian function L(w,λ) is constructed:

[0115]

[0116] Where λ is the Lagrange multiplier, which is used to constrain the normalization of dynamic weights.

[0117] In general, the Lagrangian function is solved by the variational method, and the extreme value of the dynamic weight vector satisfies the following Euler-Lagrangian equation:

[0118]

[0119] In actual calculations, due to the complexity of the objective function, numerical solutions are usually used to solve the weights. Specifically:

[0120] Use the gradient descent method to iteratively optimize the objective function and update the weight vector w(t);

[0121] In each iteration, the weights are normalized according to the constraints to ensure that the weights are non-negative and sum to 1;

[0122] When the change in the objective function is less than the set convergence threshold, the iteration stops and the final dynamic weight vector is output.

[0123] In some embodiments, in order to improve computational efficiency, the optimization problem can be discretized into a problem with a finite time step. 0 ,,t f ] is discretized into n time steps:

[0124] {t 0 ,t 1 ,t 2 ,…,t n}

[0125] At each time step t j , respectively calculate the corresponding dynamic weight w i (t j ), and update the weighted fusion result.

[0126] In this embodiment, the final output dynamic weight vector w(t) includes the weight distribution of each data source in the entire time interval, which is used for the next step of dynamic fusion calculation. By dynamically adjusting the weights, the present invention can adapt to the time-varying characteristics of multi-source data, for example:

[0127] When the real-time observation value of a data source fluctuates less, its weight is larger;

[0128] When the uncertainty of forecast data increases, its weight automatically decreases.

[0129] The above method realizes the reasonable allocation of dynamic weights by optimizing the calculation of the objective function, and provides theoretical and technical support for the subsequent dynamic fusion of multi-source data and the generation of target forecast values.

[0130] S3. Dynamically integrate multi-source data based on optimized weights

[0131] In this step, the dynamic weight vector calculated in the previous step is used to dynamically weight the multi-source hydrological data to generate the target forecast result. By adjusting the dynamic weight, the characteristics of different data sources are fully utilized, so that the fusion result can reflect the real-time changing hydrological state and provide a high-precision fitting result for the target forecast value. The dynamic fusion process strictly follows the weight distribution of the optimization model and realizes the dynamic association of data in the weighted calculation process.

[0132] In general, multi-source hydrological data have different temporal and spatial distribution characteristics. For example, real-time observation data has high timeliness but may have measurement errors, historical data contributes more to long-term trends but lacks real-time performance, and forecast data has future foresight but may have greater uncertainty. Through dynamic weighted fusion, the weights of each data source can be dynamically adjusted according to the characteristics of time and data to achieve the optimal fit of the target forecast value.

[0133] In this embodiment, the implementation process of dynamic fusion specifically includes the following steps.

[0134] First, the dynamic weight vector w(t)=[w 1 (t),w 2 (t),…,w m (t)], combined with the multi-source data matrix D(t) = [D 1 (t),D 2 (t),…,D m (t)], calculate the target forecast value F'(t). The calculation formula of dynamic fusion is:

[0135]

[0136] in:

[0137] F'(t) represents the dynamic weighted fusion result at the target time t;

[0138] w i (t) represents the dynamic weight of the i-th data source, and its value is determined by the dynamic weight optimization model;

[0139] D i (t) represents the data value of the i-th data source at time t.

[0140] Specifically, the dynamic fusion process is a weighted calculation of each data source, ensuring that high-weight data sources contribute more to the fusion results, while the influence of low-weight data sources is appropriately weakened. For example, during a sudden change in rainfall, the weight of real-time observation data may dominate, while the weight of historical data is relatively reduced.

[0141] As an option, the distribution of dynamic weights can be adjusted based on the accuracy, uncertainty, or relevance of the data to the target value. For example:

[0142] When the historical error of a data source is small or the correlation with the target forecast value is high, its dynamic weight will be relatively increased;

[0143] When the uncertainty of a data source increases (such as greater volatility in forecast data), its dynamic weight will automatically decrease.

[0144] In a possible implementation, in order to improve the stability of the fusion result, the dynamic fusion result can also be smoothed. For example, for some data sources that fluctuate violently in a short period of time, the moving average method can be used to smooth F′(t), and the formula is:

[0145]

[0146] Where k represents the size of the smoothing window, Represents the fusion result after smoothing.

[0147] In some embodiments, in order to further verify the rationality of the dynamic fusion result, F′(t) may be compared with the target prediction value F(t) to calculate the error value ΔF(t), and the formula is:

[0148] ΔF(t)=F'(t)-F(t)

[0149] According to the distribution of error values, the parameters in the dynamic weight optimization model (such as the weight adjustment rate) are adjusted to further improve the fitting accuracy.

[0150] In this embodiment, the dynamic fusion result F'(t) can be output in the form of a time series for input into the subsequent hydrological dispatch optimization model. For example, for a flood forecast in a certain river basin, the dynamic fusion result may show that the flow rate in the next 24 hours will be from 200m 3 / s rises to 800m 3 / s, providing a basis for the formulation of scheduling strategies.

[0151] As a possible extension, in order to adapt to more hydrological forecasting scenarios, regional calculations can also be introduced in the dynamic fusion process. For example, for large-scale basin forecasts, the entire basin can be divided into several sub-basins, and dynamic weighted calculations can be performed on each of them, and the results can be summarized in the entire basin. Assume that the basin is divided into three sub-basins, and their dynamic fusion results are F′ 1 (t), F′ 2 (t) and F′ 3 (t), the fusion result of the whole basin is:

[0152]

[0153] where α k It represents the weight distribution ratio of the sub-basin, which can be calculated based on the area or flow contribution ratio of the sub-basin.

[0154] Through the above-mentioned dynamic fusion method, the present invention realizes the dynamic association and weighted calculation of multi-source data, so that the target forecast results can more accurately reflect the actual hydrological changes, laying a reliable data foundation for subsequent hydrological scheduling optimization.

[0155] S4. Constructing a hydrological dispatch optimization model based on target forecast results

[0156] This step builds a hydrological dispatch optimization model based on the target forecast results generated in step S3, and calculates the optimal flood dispatch strategy by comprehensively considering the reservoir water level, flow and hydrodynamic model conditions. The goal of the hydrological dispatch optimization model is to maintain the safety range of the reservoir water level while smoothing the changes in inflow and outflow, thereby providing a reliable decision-making basis for flood management and water resources dispatch.

[0157] In general, the hydrological dispatching process needs to meet a series of physical constraints, such as water conservation and upper and lower limits of reservoir water levels. In order to achieve the optimization goal, the present invention combines the target forecast results and uses mathematical optimization methods to solve the optimal outflow flow time series and the corresponding water level change process.

[0158] In this embodiment, the objective function of the hydrological dispatch optimization model is defined as:

[0159]

[0160] in:

[0161] q in (t) represents the inflow flow, which is the inflow data in the target forecast result;

[0162] q out (t) represents the outflow flow, which is the variable to be solved for optimization;

[0163] h(t) represents the reservoir water level;

[0164] h safe is the target safe water level of the reservoir;

[0165] λ is the weight factor, which is used to balance the optimization weight between flow variation and stable water level.

[0166] As an option, the optimization objective function can adjust the value of the weight factor λ in different scenarios, for example:

[0167] During the flood peak, priority is given to controlling the reservoir water level safety, so the value of λ is increased;

[0168] In the normal scheduling period, smoothing traffic changes is prioritized, so the value of λ is reduced.

[0169] Specifically, the optimization objective function needs to satisfy the following physical constraints:

[0170] Water conservation relationship:

[0171]

[0172] Where A is the reservoir area, which represents the impact of water volume changes on water level.

[0173] Reservoir water level restrictions:

[0174] h min ≤h(t)≤h max

[0175] where h min and h max They are the lowest and highest water levels of the reservoir, respectively, and are used to ensure safety during the scheduling process.

[0176] Flow non-negative constraint:

[0177] q out (t)≥0

[0178] This constraint ensures that the outbound flow is always positive, which conforms to the laws of physics.

[0179] In a possible implementation, in order to meet the above optimization objectives and constraints, the Lagrange multiplier method is used to construct the constrained optimization problem, and its Lagrange function is expressed as:

[0180]

[0181] in:

[0182] λ h is the Lagrange multiplier for the water conservation constraint;

[0183] λ qis the Lagrange multiplier for the non-negativity constraint of the flow.

[0184] In this embodiment, in order to solve the above optimization problem, a numerical optimization method (such as gradient descent method) is used to iteratively calculate the outflow flow q out (t) and water level change h(t). The optimization process is as follows:

[0185] Initialize the outbound flow time series q out (t) is zero or a constant value;

[0186] Update the flow value according to the gradient calculation formula, for example:

[0187]

[0188] Where η is the learning rate, which is used to control the update step size;

[0189] In each iteration, the water level h(t) is updated in real time according to the water conservation equation;

[0190] When the change of the objective function is less than the set threshold, the iteration stops and the optimal solution is output.

[0191] In some embodiments, in order to adapt to more complex flood scheduling scenarios, the basin can be divided into multiple sub-areas, and the scheduling optimization problem can be solved for each sub-area separately. For example, in a basin with three sub-areas, namely upstream, midstream and downstream, the scheduling optimization results of each sub-area can be summarized according to the downstream superposition principle, and the formula is:

[0192] q total (t) = q upstream (t)+q midstream (t)+q downstream (t)

[0193] where q total (t) represents the total outflow of the entire basin.

[0194] In this embodiment, the scheduling optimization results finally output include:

[0195] Outbound flow time series q out (t);

[0196] Reservoir water level change time series h(t).

[0197] As an extension, the optimization results can also be compared with historical flood processes to verify the feasibility of the scheduling scheme. For example, the scheduling process of a historical flood shows that the peak outflow flow is 800m 3 / s, and this scheduling result reduces the peak value to 600m through smooth adjustment. 3 / s, thereby effectively reducing downstream pressure.

[0198] Through the above optimization process, the present invention realizes dynamic scheduling optimization driven by multi-source data, ensures the safety and stability of flood scheduling, and provides a scientific basis for watershed management and flood control decision-making.

[0199] In general, the present invention receives real-time data, measured data and forecast data, constructs a dynamic weight optimization model, performs dynamic weighted fusion on multi-source data to generate target forecast results, and constructs a hydrological dispatch optimization model in combination with hydrodynamic algorithm conditions to calculate the optimal flood dispatch strategy. Through the combination of dynamic weight adjustment and physical constraints, the present invention can adapt to the nonlinear change characteristics of the hydrological system, provide accurate hydrological forecasts and safe and efficient dispatch solutions, and provide scientific support for basin management and flood control decision-making.

[0200] Please see attached Figure 2 As a preferred embodiment of the present invention, the method further includes: the user adjusts the forecast results and the scheduling strategy through the interface, and the system dynamically corrects the forecast results and the scheduling strategy according to the adjustment.

[0201] This step provides the function of adjusting the forecast results and dispatch strategies generated by the system through the user interaction interface. Users can interactively modify the forecast results and dispatch strategies according to specific management needs or external information (such as field observation data, expert opinions, etc.), and the system responds to user adjustments in real time and re-corrects the forecast results and dispatch plans. This step achieves flexibility and accuracy in forecast dispatch by dynamically updating the optimization model and result feedback, providing more autonomy and adaptability for hydrological management.

[0202] Generally speaking, the data adjusted by the user may destroy the constraints of the original optimization model. Therefore, the system needs to reallocate weights and correct scheduling strategies through real-time calculations to ensure the rationality and physical consistency of the adjustment results.

[0203] In this embodiment, the specific steps for the user to adjust the forecast results and scheduling strategies through the interactive interface include the following:

[0204] First, the system provides a visual user interface to display the fusion results of multi-source data (i.e., target forecast results F'(t) and dispatch optimization results (including inflow flow q int (t), outbound flow q out As an option, users can drag and drop key nodes in the forecast results (such as flood peak value, rising point) or key flow values ​​in the dispatch curve. For example, users can drag the flood peak point on the interface to change it from 800m 3 / s adjusted to 860m 3 / s, or by modifying the outbound flow curve to make the scheduling flow smoother.

[0205] Specifically, the value adjusted by the user is captured by the system in real time and triggers dynamic correction calculation. j The predicted value F′(t j ) is adjusted to the new value The system needs to reallocate the dynamic weight w(t) based on the adjusted results and calculate the new multi-source data weighted results. The adjusted dynamic weight formula is:

[0206]

[0207] in:

[0208] The dynamic weight adjusted by the user must satisfy the non-negativity and normalization constraints;

[0209] D i (t) is the time series value of the i-th data source.

[0210] In general, in order to balance the deviation between user adjustments and the original model results, the redistribution of optimization weights needs to be achieved through the following objective function:

[0211]

[0212] By optimizing and solving the above objective function, the system calculates the adjusted dynamic weight vector

[0213] As a possible implementation method, when the user adjusts the scheduling strategy (such as modifying the outbound flow q out (t)), the system needs to recalculate the water conservation relationship to update the water level change h(t). Specifically, according to the adjusted outflow time series The water conservation equation is updated as follows:

[0214]

[0215] The system iteratively solves the water level change h(t) through numerical methods and re-optimizes the scheduling plan based on the user's adjustment results.

[0216] In some embodiments, in order to prevent the user from making excessive deviations from the physical laws, the system can set a reasonable adjustment range. out (t)=-100m 3 / s, the system will prompt that the adjustment violates the flow non-negativity constraint and automatically reject the adjustment request.

[0217] In this embodiment, the adjusted forecast results and scheduling strategies are fed back to the user interface in real time, and the impact of the adjustment is displayed graphically. For example, the system can overlay the flow curve and water level change curve before and after the adjustment to help users evaluate the rationality of the adjustment.

[0218] Through the above steps, the present invention realizes the organic combination of user interactive adjustment and dynamic correction calculation, enhances the flexibility and adaptability of the system, and ensures that the adjustment results conform to the laws of physics, providing more flexible support for hydrological forecasting and scheduling decisions.

[0219] Please refer to the attached Figure 3 As a preferred embodiment of the present invention, the method further includes: generating hydrological forecast and scheduling results files, and supporting automatic publishing of the results.

[0220] This step organizes the hydrological forecast results and scheduling optimization strategies generated in the previous steps into standardized results files for users to view and use intuitively, and further supports the automatic release of results. By visualizing and structuring the complex hydrological forecast process and scheduling results in the form of files, it is not only convenient for the unified use of basin management departments, but also can realize the rapid transmission of information between multiple departments, improving work efficiency and scientific decision-making.

[0221] In general, the generation of hydrological forecast and dispatch results files needs to include target forecast results, dispatch strategies and related characteristic values, such as flood peak value, rising point, outflow flow curve, etc. In addition, the format of the results file can be generated in PDF, Excel or other standardized file formats according to usage requirements, and pushed to designated units through automated means.

[0222] In this embodiment, the steps of generating the hydrological forecast and dispatching result file specifically include the following:

[0223] First, the target forecast result F'(t) generated in step S3 and the dispatch optimization result calculated in step S4 are sorted out, including the outbound flow time series q out (t) and water level change time series h(t). These results will be formatted into easy-to-read structured data, such as tables, graphs, or regional summary data.

[0224] Specifically, the target forecast result file content may include:

[0225] Flood peak value: the maximum flow rate and its corresponding time during the target forecast process. For example, when the flow rate reaches 800m 3 / s, the time is t = 12h;

[0226] Rising point: the turning point when the flow rate changes from a steady increase to a significant increase;

[0227] Future flow curve: displays the predicted flow change trend in the form of time series.

[0228] The file contents of the scheduling optimization results may include:

[0229] Outflow curve: Displays the changing trend of reservoir outflow in the form of time series;

[0230] Water level change curve: describes the ups and downs of the reservoir water level during the dispatch period;

[0231] Water balance verification results: through the formula

[0232]

[0233] Verify whether the water level changes conform to the law of conservation of water.

[0234] In a possible implementation, the forecast and scheduling result files can be generated in multiple formats:

[0235] Excel file: used for recording and sharing accurate data;

[0236] PDF report: used for decision support or official document release;

[0237] Graphic files (such as PNG): used to visually display the target forecast curve and scheduling result curve.

[0238] As an option, error analysis can be added to the output document. For example, by calculating the error ΔF(t) between the target forecast value F'(t) and the actual value F(t), an error distribution graph is generated to quantify the accuracy of the model prediction.

[0239] In this embodiment, the generated output file also includes necessary metadata information, such as the report generation time, data source and generation method description, to ensure the traceability of the file. For example:

[0240] The report generation time is 20xx-xx-xx, 10:00;

[0241] The data sources are real-time monitoring data from upstream control stations and rainfall forecasts for the next 48 hours.

[0242] After the files are generated, it further supports the automatic release of the results. Specifically:

[0243] Users can preset the receiving unit and file type, and the system will automatically push the result files to the designated department or platform according to the requirements of the task book. For example, send PDF reports via email, or push them to the watershed management system through the API interface;

[0244] Automatically generate release logs, including release content, receiving units, release time, etc., to facilitate traceability and auditing.

[0245] In some embodiments, in order to enhance the reliability of the achievement release, a pre-release review mechanism may be provided. For example, the system generates a release draft before pushing, for the user to check and confirm before sending.

[0246] In one possible implementation, the automated publishing process also supports multi-level unit distribution. For example, files can be pushed to higher-level water conservancy departments and lower-level hydrological stations according to administrative regions, achieving simultaneous publishing across the entire river basin.

[0247] Through the above steps, the present invention can present the complex hydrological forecasting and scheduling process in an intuitive and structured form, and efficiently distribute it to relevant departments or units, ensuring the efficient use of the results and the scientific nature of the decision-making.

[0248] The hydrological forecast interaction system for fusing multi-source data described below and the hydrological forecast interaction method for fusing multi-source data described above can be referenced to each other.

[0249] Please see attached Figure 4 The present invention also provides a hydrological forecast interactive system integrating multi-source data, comprising:

[0250] The data acquisition module 10 is used to receive and process real-time data, measured data and forecast data;

[0251] The dynamic data fusion module 20 is used to calculate the dynamic weights of multi-source data based on the optimization objective function, and fuse them to obtain the target prediction result;

[0252] The scheduling optimization module 30 is used to construct a scheduling optimization model based on a hydrodynamic algorithm to generate an optimal scheduling strategy;

[0253] The user interaction module 40 is used to provide an interactive interface, support users to adjust the forecast results and scheduling strategies, and provide real-time feedback on the adjusted optimization results;

[0254] The result generation module 50 is used to generate flood forecast and dispatch result files and support the automatic release of the results.

[0255] The system of this embodiment can be used to execute the above method embodiments, and its principles and technical effects are similar, which will not be repeated here.

[0256] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A hydrological forecasting interactive method integrating multi-source data, characterized in that: The following steps are involved: Receive and pre-process multi-source hydrological data, including real-time data, measured data and forecast data of hydrology and meteorology; Based on multi-source data, a dynamic weight optimization model is constructed to calculate the dynamic weight of each data source by optimizing the objective function; Dynamically fuse multi-source data based on optimized weights to obtain target forecast results; Based on the target forecast results, a hydrological scheduling optimization model is constructed, and the optimal flood scheduling strategy is calculated in combination with the hydrodynamic model.

2. The hydrological forecast interactive method for fusing multi-source data according to claim 1 is characterized in that: The multi-source data includes at least two types of data: Real-time data, derived from real-time water level, flow and rainfall at hydrological stations; Measured data refers to data collected on-site by instruments, equipment or sensor tools during flood events and then manually analyzed and calculated, rather than directly collected data. Forecast data, including future forecast data of hydrology and meteorology.

3. The hydrological forecast interactive method for fusing multi-source data according to claim 1 is characterized in that: The step of constructing a dynamic weight optimization model based on multi-source data and calculating the dynamic weight of each data source by optimizing the objective function includes: Receive multi-source data, including real-time data, measured data and forecast data, and standardize the data to make them consistent in time step, spatial distribution and physical quantity units; Construct the optimization objective function of the dynamic weight optimization model. The objective function is used to minimize the deviation between the target forecast value and the weighted result of multi-source data, while optimizing the smoothness of the weight distribution and ensuring that the optimization result conforms to the laws of hydrological physics. Set the objective function as: Where w(t)=[w1(t),w2(t),…,w m (t)] is the dynamic weight vector, w i (t) is the dynamic weight of the i-th data source, D i (t) is the value of the ith data source, F(t) is the target forecast value, C(t) is the hydrological physical constraint term, α, β, γ are weight control parameters, t0 and t f Respectively represent the start and end time of the time interval; The dynamic weight w(t) is calculated by optimizing the objective function, so that the deviation between the weighted fusion result of multi-source data and the target forecast value is minimized.

4. The hydrological forecast interactive method for fusing multi-source data according to claim 3 is characterized in that: The optimization objective function of the dynamic weight optimization model is solved by the following steps: Construct a Lagrangian function based on the optimization objective function to constrain the normalization conditions of the weights: Among them, L(w,λ) is the Lagrangian function, λ is the Lagrangian multiplier, is the weight normalization constraint; According to the variational method, the Lagrange equation is solved and expressed as: The dynamic weight w of each data source is calculated by this equation i (t); Combined with numerical solutions, including gradient descent and finite difference methods, the dynamic weights are iteratively solved to ensure that the weights meet the non-negativity and normalization conditions.

5. The hydrological forecast interactive method for fusing multi-source data according to claim 1 is characterized in that: The step of dynamically fusing multi-source data based on optimized weights to obtain target forecast results includes: For each data source in the multi-source data, obtain its time series in the time interval [t0,,t f ] constitute the data sequence matrix; The dynamic weight vector w(t) is used to perform dynamic weighted fusion on multi-source data, and the target prediction result calculation formula is: Among them, F′(t) is the forecast result at the target time t, w i (t) represents the dynamic weight of the i-th data source, D i (t) represents the value of the i-th data source at time t.

6. The hydrological forecast interactive method for fusing multi-source data according to claim 1 is characterized in that: The hydrological regulation optimization model is implemented by the following steps: Construct optimization objectives to minimize the deviation of inflow and outflow and maintain the water level within a safe range; The hydrodynamic model is used as a constraint condition, wherein the constraint includes a water conservation equation, a water level change equation, and a flow non-negative constraint; The outflow flow curve and reservoir water level change curve are obtained by numerical optimization method.

7. The hydrological forecast interactive method for fusing multi-source data according to claim 6 is characterized in that: The hydrodynamic model includes: Water conservation constraint: the difference between inflow and outflow determines the change of reservoir water level; Water level limit constraints, the reservoir water level must not exceed the preset maximum water level or fall below the minimum water level; Flow non-negativity constraint: outbound flow cannot be negative.

8. The hydrological forecast interactive method for fusing multi-source data according to claim 1 is characterized in that: The method further comprises: Generate hydrological forecast and scheduling results files, and support automatic release of results.

9. The hydrological forecast interactive method for fusing multi-source data according to claim 8, characterized in that: The generated hydrological forecast and dispatch results files include: Flood forecast process curve, including rising point, flood peak point and corresponding time; Scheduling strategy curve, including outflow curve and water level change curve; Data error analysis report shows the deviation and trend between the measured value and the predicted value.

10. A hydrological forecast interactive system integrating multi-source data, applied to the method according to any one of claims 1 to 9, characterized in that: include: Data acquisition module, used to receive and process real-time data, measured data and forecast data; Dynamic data fusion module, used to calculate the dynamic weight of multi-source data based on the optimization objective function, and fuse them to obtain the target prediction result; The scheduling optimization module is used to build a scheduling optimization model based on the hydrodynamic model and generate the optimal scheduling strategy; The user interaction module is used to provide an interactive interface, support users to adjust forecast results and scheduling strategies, and provide real-time feedback on the adjusted optimization results; The results generation module is used to generate flood forecasting and scheduling results files, and supports the automatic release of results.

Images