Intelligent early warning and automatic triggering type emergency linkage system for small watershed mountain torrents

Abstract

An intelligent early warning and automatic triggering type emergency linkage system for small watershed mountain torrents of the present invention belongs to the technical field of mountain torrent disaster monitoring and early warning, and comprises a data acquisition and processing module, a risk assessment module, an early warning judgment module and an emergency response triggering module. According to the method, early warning can be given out earlier, precious time is won for disaster prevention and reduction, it is ensured that emergency resources are put to the most needed place with the optimal path in the shortest time, the response efficiency is remarkably improved, the system can continuously learn from practice, the early warning and response performance of the system is continuously improved, and the continuous self-evolution ability is achieved.

Description

Technical Field

[0001] This invention belongs to the field of flash flood disaster monitoring and early warning technology, specifically a small watershed flash flood intelligent early warning and automatic triggering emergency linkage system. Background Technology

[0002] In the prevention and control of flash floods, traditional early warning methods mainly rely on setting up fixed hydrological monitoring stations in the watershed to collect single or a few key parameters such as rainfall and water level. When the monitored value exceeds the pre-set static threshold, an early warning is triggered. Emergency response is usually carried out according to a fixed plan, with resources coordinated and dispatched manually.

[0003] The aforementioned traditional technical solutions have significant shortcomings. Flash floods are a complex, non-linear process influenced by a combination of factors, including rainfall intensity, pre-accelerated soil moisture content, surface runoff, and watershed characteristics. Relying solely on a few monitoring parameters and fixed warning thresholds makes it difficult to accurately capture the dynamic evolution of risks, resulting in low accuracy and frequent false alarms, thus weakening the credibility of the warning system. Data processing technology also has limitations. Raw data from different types of sensors, such as rain gauges, water level gauges, and flow meters, have varying physical meanings, dimensions, and numerical ranges. Traditional methods struggle to effectively integrate this heterogeneous data for comprehensive evaluation. Emergency response efficiency is low. After a warning is issued, relying on manual resource allocation decisions not only results in slow response times but also fails to calculate the optimal resource allocation plan in real time under complex conditions. This may lead to rescue forces failing to reach the most needed locations efficiently, wasting valuable rescue time. This lack of dynamic optimization and closed-loop learning capabilities hinders the continuous improvement of the overall prevention and control system's effectiveness.

[0004] In summary, existing technologies, due to their limited data processing methods, static early warning models, and lack of optimized emergency decision-making, are insufficient to achieve accurate and timely early warning and efficient and intelligent response to flash flood risks.

[0005] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds, so as to solve the problems mentioned in the background art.

[0007] The technical solution of the present invention is: including a data acquisition and processing module, used to acquire a set of original monitoring parameters characterizing the status of monitoring points in a small watershed, and based on the original monitoring parameters and a preset standardized model, process them into dimensionless standardized risk factors;

[0008] The risk assessment module is used to combine standardized risk factors with fusion weights that characterize the reliability of monitoring points to generate a comprehensive risk index that quantitatively represents the overall risk of the current small watershed.

[0009] The early warning determination module is used to compare the comprehensive risk index with the early warning threshold that is dynamically adjusted based on historical environmental data, and determine the early warning level that represents the urgency of the emergency.

[0010] The emergency response triggering module is used to generate an emergency response command to initiate the allocation of downstream resources when the warning level reaches the preset triggering conditions, based on the correlation between the warning level and the mutual influence between the monitoring points.

[0011] Preferably, the system also includes a communication assurance module, which is used for:

[0012] The reliability of each data communication link is evaluated in real time, and the overall communication availability of the system is calculated based on this.

[0013] The system's overall communication availability is compared with a preset communication threshold. When the system's overall communication availability is lower than the communication threshold, the preset communication protection strategy is triggered.

[0014] The communication assurance strategy is used to reduce the data acquisition frequency of the data acquisition and processing module or switch the data transmission to a backup communication link.

[0015] Preferably, the specific steps for standardization in the data acquisition and processing module are as follows:

[0016] Collect the corresponding raw measurement values ​​for each monitoring point;

[0017] Obtain the historical statistical parameters of the group corresponding to the category to which the original measurement value belongs. The historical statistical parameters include the historical mean and the historical standard deviation.

[0018] Using a pre-defined Z-score standardization method, dimensionless standardized risk factors are calculated based on the original measured values, historical mean, and historical standard deviation.

[0019] Preferably, the reliability of the monitoring points is quantitatively characterized by the standard deviation of the errors of each monitoring sensor; the risk assessment module uses the following steps to determine the fusion weights:

[0020] Obtain the standard deviation of the error for various types of sensors;

[0021] Based on the preset weighting model, the reciprocal square of the error standard deviation is used to obtain the accuracy measure value;

[0022] The accuracy measurement values ​​are normalized to determine the fusion weights.

[0023] Preferably, the standardized risk factors specifically include standardized rainfall intensity, standardized flow rate of change, and standardized soil saturation; the steps used by the risk assessment module to generate the comprehensive risk index are as follows:

[0024] The risk factor set is composed of standardized rainfall intensity, standardized flow rate change rate, standardized soil saturation, and externally introduced historical disaster factors.

[0025] A comprehensive risk index is generated by weighted summation of the risk factor set through a multi-parameter fusion model.

[0026] Preferably, historical environmental data specifically includes previous impact rainfall and current soil moisture; the steps for the early warning determination module to dynamically adjust the early warning threshold are as follows:

[0027] Obtain the preset baseline threshold;

[0028] Based on the previous rainfall and current soil moisture, the baseline threshold is dynamically adjusted downwards through a dynamic adjustment model to generate an early warning threshold.

[0029] Preferably, the steps for the early warning determination module to determine the early warning level are as follows:

[0030] The risk ratio is obtained by calculating the ratio of the comprehensive risk index to the early warning threshold.

[0031] The risk ratio is matched with a preset multi-level warning range to determine the warning level from multiple preset levels.

[0032] Preferably, the correlation degree is used to quantify the correlation between different monitoring points in the spatiotemporal dimension; the steps for the emergency response triggering module to construct the correlation degree are as follows:

[0033] Obtain the spatial distance and data acquisition time difference between any two monitoring points;

[0034] Based on spatial distance and data acquisition time difference, the correlation degree is calculated and generated through a spatiotemporal correlation matrix model.

[0035] Preferably, the system also includes a response resource intelligent allocation module, which is used for:

[0036] Based on the warning level and in conjunction with the pre-set demand assessment model, determine the specific resource demand.

[0037] With the goal of minimizing the total weighted response time and based on the upper limit of resource supply as a constraint, the optimal resource allocation scheme is obtained.

[0038] The optimal resource allocation plan is taken as the specific execution content, and the emergency response instructions are refined.

[0039] Preferably, the system also includes an evaluation feedback module, which is used for:

[0040] After the emergency response is implemented, obtain the group assessment sub-scores generated after evaluating the effectiveness of the emergency response;

[0041] The evaluation scores include timeliness and coverage.

[0042] Based on the evaluation scores, the model is adjusted through feedback to adaptively adjust the weight coefficients corresponding to the total weighted response time optimization objective in the intelligent resource allocation module.

[0043] This invention provides an improved intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds, which has the following improvements and advantages compared with the prior art:

[0044] 1. By using Z-score standardization, the problem of fusing multi-source heterogeneous data was solved, and a comprehensive risk index combining sensor reliability weights and multi-dimensional risk factors was constructed. The evaluation results are more comprehensive and accurate than traditional methods that rely on a single parameter.

[0045] 2. A dynamic early warning threshold is adopted, which can be adaptively adjusted according to environmental background such as previous rainfall and soil moisture. This makes the system more sensitive under high-risk conditions before a disaster occurs, and can issue early warnings earlier, thus buying valuable time for disaster prevention and mitigation.

[0046] 3. A resource allocation optimization model was introduced, which aims to minimize the total weighted response time and automatically generates the optimal allocation plan. This changes the inefficient traditional model that relies on manual decision-making and ensures that emergency resources are deployed to the most needed locations in the shortest time via the optimal path, significantly improving response efficiency.

[0047] 4. This invention innovatively designs an evaluation feedback module, forming a closed-loop optimization system that can quantify the evaluation results of each emergency response and use them to adjust and optimize the internal decision model parameters. This enables the system to continuously learn from practice, constantly improve its early warning and response performance, and has a continuous self-evolution capability. Attached Figure Description

[0048] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0049] Figure 1 This is a flowchart of the system of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0051] Example 1:

[0052] Please see Figure 1 This invention provides a technical solution for a smart early warning and automatic triggering emergency response system for flash floods in small watersheds, including: a data acquisition and processing module, used to acquire a set of original monitoring parameters characterizing the status of monitoring points in small watersheds, and based on the original monitoring parameters and a preset standardized model, process them into dimensionless standardized risk factors;

[0053] The risk assessment module is used to combine standardized risk factors with fusion weights that characterize the reliability of monitoring points to generate a comprehensive risk index that quantitatively represents the overall risk of the current small watershed.

[0054] The early warning determination module is used to compare the comprehensive risk index with the early warning threshold that is dynamically adjusted based on historical environmental data, and determine the early warning level that represents the urgency of the emergency.

[0055] The emergency response triggering module is used to generate an emergency response command to initiate the allocation of downstream resources when the warning level reaches the preset triggering conditions, based on the correlation between the warning level and the mutual influence between the monitoring points.

[0056] This embodiment provides a smart early warning and automatic triggering emergency response system for flash floods in small watersheds. The data acquisition and processing module obtains raw measurement values ​​from various monitoring points. These raw measurement values ​​come from diverse sources, including sensors for rainfall, water level, and flow velocity, and these data differ in dimensions and numerical ranges. The data acquisition and processing module uses the Z-score standardization method to process the raw measurement values. This method eliminates the differences between data from different types of sensors, and the processed data forms dimensionless standardized risk factors. This provides unified, high-quality input data for subsequent risk assessment.

[0057] Example 2:

[0058] The specific steps for standardization in the data acquisition and processing module are as follows:

[0059] Collect the corresponding raw measurement values ​​for each monitoring point;

[0060] Obtain the historical statistical parameters of the group corresponding to the category to which the original measurement value belongs. The historical statistical parameters include the historical mean and the historical standard deviation.

[0061] Using a pre-defined Z-score standardization method, a dimensionless standardized risk factor is calculated based on the original measured values, historical mean, and historical standard deviation.

[0062] In this embodiment, the standardization process performed by the data acquisition and processing module is further described. The data acquisition and processing module acquires the k-th raw measurement value for the i-th type of sensor; obtains the historical statistical parameters of this type of sensor, which are pre-calculated by analyzing a large amount of historical data, including the historical mean and historical standard deviation; the data acquisition and processing module applies the Z-score standardization method, which calculates the standardized risk factor according to the following formula:

[0063] Where, χ ik X is the standardized result of the k-th measurement by the i-th type of sensor. ik : The raw measurement value of the i-th sensor in the k-th measurement, μ i : The mean of historical data of the i-th type of sensor, s i : Standard deviation of historical data for the i-th type of sensor;

[0064] Through this standardization process, raw monitoring data with different physical meanings and dimensions are transformed into unified, directly comparable, and weighted standardized risk factors, laying the foundation for accurate calculations in the risk assessment module.

[0065] Example 3:

[0066] The reliability of monitoring points is quantitatively characterized by the standard deviation of the errors of each monitoring sensor; the risk assessment module uses the following steps to determine the fusion weights:

[0067] Obtain the standard deviation of the error for various types of sensors;

[0068] Based on the preset weighting model, the reciprocal square of the error standard deviation is used to obtain the accuracy measure value;

[0069] The accuracy measurement values ​​are normalized to determine the fusion weights;

[0070] To clarify the method for obtaining the above parameters, the present invention adopts the following steps:

[0071] For historical statistical parameter μ i and s i Determination: After formal deployment, the system collects raw monitoring data from various sensors for at least one complete hydrological year. Based on this dataset, the historical data mean μ for each type of sensor is calculated. i Standard deviation of historical data s i , which serves as the benchmark parameter for subsequent Z-score standardization;

[0072] Regarding the standard deviation of error σ iDetermination: During the system installation and commissioning phase, each type of sensor is calibrated on-site. Specifically, within the sensor's measurement range, at least five test points are selected. The sensor readings are compared with the readings of a high-precision standard device. The deviations between the two readings are recorded, and the standard deviation of these deviations is calculated. This standard deviation σ is taken as the error standard deviation σ of that type of sensor. i This parameter reflects the actual measurement accuracy of the sensor.

[0073] In this embodiment, the mechanism by which the risk assessment module determines the fusion weights of each sensor's data is revealed. The risk assessment module obtains the standard deviation of the error for each sensor. The standard deviation of the error is a key indicator for measuring the accuracy of a sensor; the higher the accuracy of a sensor, the smaller its standard deviation of error. The risk assessment module calculates the fusion weights based on a preset weighting model. This model measures the accuracy of the sensor by the reciprocal square of the standard deviation of the error and performs normalization processing. The specific calculation formula is as follows:

[0074] Among them, W i : The fusion weights of the i-th type of sensor, σ i : Standard deviation of error of the i-th type of sensor, N: Total number of sensor types participating in fusion;

[0075] The fusion weights calculated by this method allow high-precision sensor data to account for a larger proportion in the subsequent calculation of the comprehensive risk index, which ensures that the final risk index can more accurately reflect the actual risk level.

[0076] Example 4:

[0077] Standardized risk factors specifically include standardized rainfall intensity, standardized flow rate of change, and standardized soil saturation; the steps used by the risk assessment module to generate a comprehensive risk index are as follows:

[0078] The risk factor set is composed of standardized rainfall intensity, standardized flow rate change rate, standardized soil saturation, and externally introduced historical disaster factors.

[0079] A comprehensive risk index is generated by weighted summation of the risk factor set through a multi-parameter fusion model.

[0080] Among them, historical disaster factor H r Quantification was achieved through the following method: Data on officially recorded flash flood events in small watersheds over the past 10 years that caused property damage or displacement were collected and studied. These disasters were categorized into three levels of severity (general, moderate, and severe), assigned scores of 1, 2, and 3 respectively. r H represents the severity score of the most recent disaster event; if there are no records within the last 10 years, then H... r =0;

[0081] The weighting coefficients α, β, γ, and δ are calibrated using a machine learning model. The specific steps are as follows:

[0082] Building the training dataset: Collect historical monitoring data and construct a feature vector for each time point [P] r Q r ,S r H r ], where P r Q r ,S r H is the standardized risk factor corresponding to this time point. r The historical disaster factors at that time; at the same time, based on whether an actual flash flood disaster occurred within a short time window after that point in time, it was marked as a positive sample (1) or a negative sample (0);

[0083] Training a logistic regression model: Use the dataset above to train a logistic regression model; the goal of this model is to predict the probability of flash floods.

[0084] Obtaining weight coefficients: After the model training is completed, the coefficients corresponding to each learned feature are the weight coefficients α, β, γ, δ in this invention; this method ensures the objectivity of weight allocation and the ability to learn from historical patterns.

[0085] In this embodiment, the process of the risk assessment module generating a comprehensive risk index is described in detail. The risk assessment module combines multiple standardized risk factors output by the data acquisition and processing module with an externally introduced historical disaster factor. These factors together constitute a multi-dimensional risk factor set. A multi-parameter fusion model is used to perform a weighted summation of all factors within this set, generating a single, quantitative comprehensive risk index. The calculation formula for this index is defined as follows:

[0086] Φ=αP r +βQ r +γS r +δH r Where Φ: comprehensive risk index, P r Standardized rainfall intensity, Q r Standardized rate of change of flow, S r Standardized soil saturation, H r Historical disaster factors; α, β, γ, δ: weight coefficients corresponding to each factor.

[0087] In this invention, the comprehensive risk index Φ is defined as the linear input part of the activation function in the logistic regression model. It can directly reflect the level of risk. The higher the index Φ is, the higher the probability of flash floods predicted by the model.

[0088] Through this multi-parameter fusion model, the system can comprehensively consider factors such as rainfall, runoff, infiltration, and historical disaster-causing patterns; the generated comprehensive risk index can comprehensively and dynamically characterize the current overall flash flood risk in the watershed.

[0089] Example 5:

[0090] Historical environmental data specifically includes previous impact rainfall and current soil moisture; the steps used by the early warning determination module to dynamically adjust the early warning threshold are as follows:

[0091] Obtain the preset baseline threshold;

[0092] This baseline threshold can be set based on historical disaster statistics or through expert experience, and represents the early warning triggering baseline under normal environmental conditions;

[0093] Based on the previous rainfall and current soil moisture, the baseline threshold is dynamically lowered through a dynamic adjustment model to generate an early warning threshold;

[0094] To determine the adjustment coefficients k1 and k2 in the formula, this invention employs a historical data backtracking analysis method:

[0095] Monitoring data from the 24 hours preceding multiple historical flash flood events were selected. For each event, the change curve of the comprehensive risk index Φ before its occurrence was analyzed, along with the corresponding preceding rainfall P. a And the current soil moisture θ; and the dynamic early warning threshold Φ 1-3 hours before the disaster occurs. th The optimization objective is to ensure that the comprehensive risk index Φ is exceeded. Optimization algorithms such as least squares are used to solve for the optimal k1 and k2 values ​​corresponding to each event. The average of multiple sets of k1 and k2 values ​​obtained from multiple historical events is used as the preset adjustment coefficient for this system. Based on experience, in typical clay and loam watersheds, the value of k1 usually ranges from 0.3 to 0.6, and the value of k2 ranges from 0.4 to 0.7.

[0096] In this embodiment, the mechanism by which the early warning determination module dynamically adjusts the early warning threshold is explained. The early warning determination module does not use a fixed threshold, but rather adaptively adjusts it based on the current environmental background conditions. The module acquires a preset benchmark threshold; it also acquires two key historical environmental data points in real time: previous rainfall and current soil moisture. Based on this data, a dynamic adjustment model is used to lower the benchmark threshold, generating a dynamic early warning threshold that better reflects the current situation. The adjustment formula is as follows:

[0097] Where, Φ th Φ0: Dynamic early warning threshold; P: Baseline threshold a The initial impact on rainfall, Pmax Maximum antecedent rainfall, θ: Current soil moisture, θ s : Saturated soil moisture; k1, k2: Adjustment coefficients;

[0098] This dynamic adjustment mechanism improves the sensitivity of early warnings; when the soil is moist or there has been rainfall in the early stages, the system will trigger an early warning with a lower risk index, thereby achieving earlier and more accurate risk prevention.

[0099] Example 6:

[0100] The steps used by the early warning determination module to determine the early warning level are as follows:

[0101] The risk ratio is obtained by calculating the ratio of the comprehensive risk index to the early warning threshold.

[0102] The risk ratio is matched with a preset multi-level warning range to determine the warning level from multiple preset levels.

[0103] In this embodiment, after generating a dynamic warning threshold, the warning determination module begins to determine the specific warning level. The module divides the real-time comprehensive risk index calculated by the risk assessment module by the dynamic warning threshold it generates. This calculation yields a standardized risk ratio, which intuitively reflects the degree to which the current risk exceeds the warning benchmark. The system matches this risk ratio with a set of preset multi-level warning intervals. For example, the preset intervals may include blue warning (0.6 < Φ / Φ_th ≤ 0.8), yellow warning (0.8 < Φ / Φ_th ≤ 1.0), orange warning (1.0 < Φ / Φ_th ≤ 1.3), and red warning (Φ / Φ_th > 1.3). Through matching, the system determines the most appropriate warning level from these preset levels. This step provides a clear and tiered action basis for subsequent emergency response.

[0104] Example 7:

[0105] Correlation is used to quantify the spatiotemporal correlation between different monitoring points; the steps for constructing correlation in the emergency response triggering module are as follows:

[0106] Obtain the spatial distance and data acquisition time difference between any two monitoring points;

[0107] Based on spatial distance and data acquisition time difference, the correlation degree is calculated and generated through a spatiotemporal correlation matrix model;

[0108] The parameters in the formula are the spatial correlation length l and the time correlation constant T. c Determined based on watershed characteristics and historical data:

[0109] Spatial correlation length l: represents the spatial range of influence of factors such as rainfall, and can be estimated by analyzing the spatial autocorrelation of data from multiple rain gauges in multiple rainfall events within the watershed; l characterizes the characteristic length of the spatial influence range; typically, its value is related to the average slope of the watershed, vegetation cover, etc., and the typical value ranges from 5 km to 20 km.

[0110] Time-dependent constant T c : Represents the rate of decay of monitoring data over time, and is related to the watershed confluence time; it can be estimated by analyzing the autocorrelation of time series data from a single monitoring point; T c It can be taken as the time interval corresponding to when the value of the time autocorrelation function decreases from 1 to 1 / e;

[0111] In this embodiment, before initiating the response, the emergency response triggering module constructs a correlation matrix. This correlation is used to quantify the degree of mutual influence between any two monitoring points in time and space. The module obtains the spatial straight-line distance between the two monitoring points and the time difference between their respective data acquisition times. A preset spatiotemporal correlation matrix model is used to calculate the correlation. This model integrates spatial attenuation effects and time attenuation effects, and the calculation formula is defined as:

[0112] Among them, R ij : The spatiotemporal correlation between monitoring points i and j, d ij Spatial distance between two points, Δt ij : Time difference between two data acquisition points, l: Spatial correlation length, T c Time-dependent constants;

[0113] The calculated correlation is key to determining whether early warning information needs to be disseminated across regions and coordinated with other regions; a high correlation means that the risk status of one point is likely to affect another point, thus requiring the triggering of a coordination mechanism.

[0114] Example 8:

[0115] The system also includes a response resource intelligent allocation module, which is used for:

[0116] Based on the warning level and in conjunction with the pre-set demand assessment model, determine the specific resource demand.

[0117] With the goal of minimizing the total weighted response time and based on the upper limit of resource supply as a constraint, the optimal resource allocation scheme is obtained.

[0118] The optimal resource allocation plan will be used as the specific execution content, and the emergency response instructions will be refined.

[0119] To achieve this intelligent allocation, the specific implementation methods of the relevant modules are as follows:

[0120] Resource demand is directly related to the warning level and the population of the threatened area; the model is: D i =B k ×P i D i Let B be the resource requirement at demand point i. k P is the basic resource coefficient corresponding to the warning level k. i It is the number of permanent residents within the area covered by demand point i;

[0121] By calling the API of a third-party map service and inputting the geographic coordinates of resource point j and demand point i, a driving trip time prediction value based on real-time traffic conditions is obtained, in minutes.

[0122] The complete mathematical expression of this optimization problem is:

[0123]

[0124] Demands satisfy constraints: (The total amount of resources allocated to demand point i must meet its demand).

[0125] Supply ceiling constraints: (The total amount of resources transferred from resource point j cannot exceed its supply limit S) j );

[0126] Nonnegativity constraint: x ij ≥0;

[0127] In this embodiment, when an emergency response is triggered, the intelligent resource allocation module begins operation. Based on the warning level determined by the warning assessment module, this module dynamically calculates the resource demand for each warning area using a preset demand assessment model; the higher the risk, the greater the demand. The module then initiates an optimization model with the objective of minimizing the total weighted response time. The total weighted response time is defined as the sum of the products of resource quantity and travel time. The model is also constrained by the upper limit of available resource quantity in each region. The objective function of this optimization problem is:

[0128] Where, x ij t represents the amount of resources allocated from resource point j to demand point i. ij : Estimated travel time from j to i, m: total number of demand points, n: total number of resource points;

[0129] Solving this model yields an optimal resource allocation scheme that clarifies who will support whom and how much. This scheme is used to refine emergency response instructions, ensuring that rescue resources can be mobilized with the highest efficiency, reflecting the principle of time first in emergency response.

[0130] Example 9:

[0131] The system also includes an evaluation and feedback module, which is used for:

[0132] After the emergency response is implemented, obtain the group assessment sub-scores generated after evaluating the effectiveness of the emergency response;

[0133] The evaluation scores include timeliness and coverage.

[0134] Based on the evaluation scores, the model is adjusted through feedback, and the weight coefficients corresponding to the total weighted response time optimization target in the intelligent resource allocation module are adaptively adjusted.

[0135] The adaptive adjustment here specifically refers to weighting the time term in the objective function of Example 8; the original objective function can be expanded into multi-objective optimization, or simplified to weighting the time cost, i.e. The weight adjustment formula w in this embodiment t′ =w t ·(1+λ t ·(1-E t The adjustment is to the time objective weight w. t (Its initial value can be set to 1);

[0136] The specific definitions of each parameter in the formula are as follows:

[0137] Timeliness score E t : Calculated by comparing the planned and actual response times. Among them, T actual This refers to the actual time from issuing an order to the arrival of the first batch of relief resources at the first point of need; T plan The estimated shortest time is calculated by the optimization model; if the actual time exceeds the plan, the score is less than 1.

[0138] Learning rate factor λ t λ is a hyperparameter used to control the adjustment step size, and its value is set empirically. To ensure the stability of the adjustment process and avoid drastic fluctuations in weights, λ... t Typically, a small value is chosen, such as 0.1; when the score E t At lower values, this formula significantly increases the time weight w. t This will enable future decisions to focus more on shortening response time;

[0139] In this embodiment, the system possesses closed-loop self-optimization capability, with its core being the evaluation feedback module. After an emergency response operation concludes, this module collects the evaluation results of the emergency response, which are presented as a set of evaluation sub-items scores. These scores include at least a timeliness score and a coverage score. The timeliness score reflects the response speed, while the coverage score reflects the completeness of the rescue area. The evaluation feedback module uses these scores, through a feedback adjustment model, to dynamically adjust the weight coefficients in the optimization objectives of the intelligent allocation of response resources module. For example, if the timeliness score is low, it indicates that evacuation or response time is a weakness. The system will increase the weight of evacuation time in the multi-objective function using the following formula, making future decisions more focused on shortening the time:

[0140] Adjusted new weights, original weights, timeliness score, learning rate factor;

[0141] w t′ =w t ·(1+λt(1-E t ), where w t′ The adjusted new weights, w t Original weights, E t Timeliness score, λ t Learning rate factor;

[0142] In this way, the evaluation results are no longer isolated post-event analysis, but are directly used as input to drive the decision-making model to perform iterative optimization; this enables the system to learn and evolve from each emergency event, making future resource allocation decisions more targeted and effective.

[0143] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A smart early warning and automatic triggering emergency response system for flash floods in small watersheds, characterized in that: include: The data acquisition and processing module is used to acquire a set of original monitoring parameters that characterize the status of monitoring points in a small watershed, and based on the original monitoring parameters and a preset standardized model, process them into dimensionless standardized risk factors. The risk assessment module is used to combine standardized risk factors with fusion weights that characterize the reliability of monitoring points to generate a comprehensive risk index that quantitatively represents the overall risk of the current small watershed. The early warning determination module is used to compare the comprehensive risk index with the early warning threshold that is dynamically adjusted based on historical environmental data, and determine the early warning level that represents the urgency of the emergency. The emergency response triggering module is used to generate an emergency response command to initiate the allocation of downstream resources when the warning level reaches the preset triggering conditions, based on the correlation between the warning level and the mutual influence between the monitoring points.

2. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 1, characterized in that, The system also includes a communication assurance module, which is used for: The reliability of each data communication link is evaluated in real time, and the overall communication availability of the system is calculated based on this. The system's overall communication availability is compared with a preset communication threshold. When the system's overall communication availability is lower than the communication threshold, the preset communication protection strategy is triggered. The communication assurance strategy is used to reduce the data acquisition frequency of the data acquisition and processing module or switch the data transmission to a backup communication link.

3. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 1, characterized in that, The specific steps for standardization in the data acquisition and processing module are as follows: Collect the corresponding raw measurement values ​​for each monitoring point; Obtain the historical statistical parameters of the group corresponding to the category to which the original measurement value belongs. The historical statistical parameters include the historical mean and the historical standard deviation. Using a pre-defined Z-score standardization method, dimensionless standardized risk factors are calculated based on the original measured values, historical mean, and historical standard deviation.

4. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 1, characterized in that, The reliability of monitoring points is quantitatively characterized by the standard deviation of the errors of each monitoring sensor; the risk assessment module uses the following steps to determine the fusion weights: Obtain the standard deviation of the error for various types of sensors; Based on the preset weighting model, the reciprocal square of the error standard deviation is used to obtain the accuracy measure value; The accuracy measurement values ​​are normalized to determine the fusion weights.

5. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 1, characterized in that, Standardized risk factors specifically include standardized rainfall intensity, standardized flow rate of change, and standardized soil saturation; the steps used by the risk assessment module to generate a comprehensive risk index are as follows: The risk factor set is composed of standardized rainfall intensity, standardized flow rate change rate, standardized soil saturation, and externally introduced historical disaster factors. A comprehensive risk index is generated by weighted summation of the risk factor set through a multi-parameter fusion model.

6. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 1, characterized in that, Historical environmental data specifically includes previous impact rainfall and current soil moisture; the steps used by the early warning determination module to dynamically adjust the early warning threshold are as follows: Obtain the preset baseline threshold; Based on the previous rainfall and current soil moisture, the baseline threshold is dynamically adjusted downwards through a dynamic adjustment model to generate an early warning threshold.

7. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 5, characterized in that, The steps used by the early warning determination module to determine the early warning level are as follows: The risk ratio is obtained by calculating the ratio of the comprehensive risk index to the early warning threshold. The risk ratio is matched with a preset multi-level warning range to determine the warning level from multiple preset levels.

8. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 1, characterized in that, Correlation is used to quantify the spatiotemporal correlation between different monitoring points; the steps for constructing correlation in the emergency response triggering module are as follows: Obtain the spatial distance and data acquisition time difference between any two monitoring points; Based on spatial distance and data acquisition time difference, the correlation degree is calculated and generated through a spatiotemporal correlation matrix model.

9. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 1, characterized in that, The system also includes a response resource intelligent allocation module, which is used for: Based on the warning level and in conjunction with the pre-set demand assessment model, determine the specific resource demand. With the goal of minimizing the total weighted response time and based on the upper limit of resource supply as a constraint, the optimal resource allocation scheme is obtained. The optimal resource allocation plan is taken as the specific execution content, and the emergency response instructions are refined.

10. The intelligent early warning and automatic triggering emergency response system for flash floods in small watersheds according to claim 9, characterized in that, The system also includes an evaluation and feedback module, which is used for: After the emergency response is implemented, obtain the group assessment sub-scores generated after evaluating the effectiveness of the emergency response; The evaluation scores include timeliness and coverage. Based on the evaluation scores, the model is adjusted through feedback to adaptively adjust the weight coefficients corresponding to the total weighted response time optimization objective in the intelligent resource allocation module.

Images