Flood disaster risk assessment method and system based on interpretable artificial intelligence

Abstract

The invention belongs to the technical field of natural disaster risk assessment, and particularly relates to a flood disaster risk assessment method and system based on interpretable artificial intelligence, and the method comprises the steps: determining impact factors according to historical flood disasters; performing simulation by using an LISFLOOD-FP hydrodynamic model to obtain a disaster-inducing factor index; predicting flood disaster risk levels in different recurrence periods in combination with a gradient boosting decision tree (GBDT) algorithm, and quantitatively identifying the importance of disaster-inducing factors; an LIME model is introduced to carry out local explanation on a GBDT prediction result, and the contribution rate of each disaster-inducing factor is clarified; and constructing an evaluation index system based on the contribution rate and a GBDT prediction result, and performing flood disaster risk evaluation. The system comprises an influence factor determination unit, a hydrodynamic model simulation unit, a GBDT prediction unit, a model interpretation unit, a risk assessment unit, a model verification unit and the like, accurate simulation and intelligent assessment of flood disaster risks are achieved, and reduction of urban flood disaster losses is facilitated.

Description

Technical Field

[0001] The present invention belongs to the technical field of natural disaster risk assessment, and specifically relates to a flood disaster risk assessment method and system based on explainable artificial intelligence. Background Art

[0002] Flood disasters are one of the most common natural disasters in the world, posing a serious threat to people's lives and property. Traditional flood risk assessment methods often rely on historical data and empirical judgments, which are difficult to fully and accurately reflect the actual situation and potential risks of flood disasters. Especially today, with the accelerated urbanization process, the complexity and uncertainty of urban flood disasters have further increased, which has put forward higher requirements for the accuracy and efficiency of flood risk assessment.

[0003] Existing flood disaster risk assessment methods have many shortcomings, such as difficulty in obtaining disaster-causing factors, inaccurate risk assessment models, and poor model interpretability. Therefore, a new method that can accurately identify flood disaster risks, conduct accurate urban flood simulations, and intelligently assess flood risk is urgently needed to reduce urban flood disaster losses. Summary of the invention

[0004] The purpose of the present invention is to provide a flood disaster risk assessment method and system based on explainable artificial intelligence to solve the problems raised in the background technology.

[0005] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0006] In a first aspect, the present invention proposes a flood disaster risk assessment method based on explainable artificial intelligence, the method comprising:

[0007] Determine the influencing factors of flood disasters based on historical flood disasters in the study area;

[0008] By using the influencing factors, combined with the rainstorm intensity formula and the Chicago rainfall pattern, the LISFLOOD-FP hydrodynamic model was constructed and allowed to be simulated to obtain the disaster factor index;

[0009] Combined with the gradient boosting decision tree algorithm, the disaster-causing factor index and influencing factor are used as input to predict the flood disaster risk level corresponding to the study area under different return periods, and quantitatively identify the importance of the disaster-causing factor index;

[0010] The LIME model is introduced to partially explain the prediction results of the GBDT model and clarify the contribution rate of each disaster factor indicator;

[0011] Combining the contribution rate and the prediction results of the GBDT model, a flood disaster risk assessment was conducted in the study area.

[0012] Furthermore, the influencing factors of flood disasters are determined based on the historical flood disasters in the study area, including:

[0013] Collect historical flood disaster data, digital elevation model data, historical rainfall data, and urban infrastructure information in the study area;

[0014] Based on the above data, the influencing factors of flood disasters including rainfall, topography, river system distribution, and urban drainage capacity are screened and determined.

[0015] Furthermore, the influencing factors are used in combination with the rainstorm intensity formula and the Chicago rainfall pattern to construct and run the LISFLOOD-FP hydrodynamic model for simulation to obtain disaster factor indicators, including:

[0016] Based on the rainstorm intensity formula and Chicago rainfall pattern, rainfall scenarios with different return periods are designed to determine flood processes with return periods of 20, 50, and 100 years, including parameters such as rainfall intensity, rainfall duration, and rain peak location;

[0017] Using the rainfall scenario and the LISFLOOD-FP hydrodynamic model, combined with the DEM data, boundary conditions and initial conditions of the study area, flood inundation simulation was performed;

[0018] Disaster factor indicators including flooding depth, flooding duration and flow velocity are extracted from the simulation results.

[0019] Furthermore, the gradient boosting decision tree algorithm is used to predict the flood disaster risk level corresponding to the study area under different return periods, and the importance of the disaster factor index is quantitatively identified, including:

[0020] Preprocess the collected data, including normalization, division into training set and test set, etc.;

[0021] Construct a GBDT model and input the preprocessed data into the GBDT model for training to learn the relationship between the disaster factor index and the flood disaster risk level;

[0022] The trained GBDT model is used to predict the flood disaster risk level of the study area under different return periods;

[0023] The importance of each disaster-causing factor indicator is quantitatively identified through the internal mechanism of the GBDT model.

[0024] Furthermore, the LIME model is introduced to partially explain the prediction results of the GBDT model and clarify the contribution rate of each disaster factor indicator, including:

[0025] For each prediction result of the GBDT model, use LIME to generate perturbation data;

[0026] Use the GBDT model to predict the disturbance data and obtain the prediction results of the disturbance data;

[0027] Based on the perturbed data and its prediction results, train a set interpretable model;

[0028] The parameters of the interpretable model are analyzed to determine the disaster factor index and its contribution rate that have the greatest impact on the prediction results.

[0029] Furthermore, the flood disaster risk assessment of the study area is performed by combining the contribution rate and the prediction results of the GBDT model, including:

[0030] Based on the contribution rate of each disaster factor indicator and the prediction results of the GBDT model, a flood disaster risk assessment index system is constructed;

[0031] Using the evaluation index system to evaluate the flood disaster risk in the study area, and obtaining the evaluation results;

[0032] Based on the assessment results, formulate corresponding disaster prevention and mitigation measures and emergency plans.

[0033] Furthermore, the evaluation method also includes: combining the contribution rate and the prediction results of the GBDT model to conduct a flood disaster risk assessment on the study area, constructing a flood disaster risk assessment model that couples LISFLOOD-FP and an interpretable GBDT algorithm, and verifying the applicability of the model, including:

[0034] Use actual observation data or historical disaster data to verify the flood disaster risk assessment model;

[0035] Compare the model prediction results with actual observations or historical disaster data;

[0036] The applicability and accuracy of flood disaster risk assessment models were evaluated based on the degree of difference.

[0037] In a second aspect, the present invention proposes a flood disaster risk assessment system based on explainable artificial intelligence, which is applied to execute any of the risk assessment methods described above, and the system comprises:

[0038] The impact factor determination unit is used to determine the impact factor of flood disasters based on the historical flood disasters in the study area;

[0039] A hydrodynamic model simulation unit is used to construct and allow the LISFLOOD-FP hydrodynamic model to simulate using the influencing factors, combined with the rainstorm intensity formula and the Chicago rain type, to obtain a disaster factor index;

[0040] The risk assessment and factor importance quantification unit is used to combine the gradient boosting decision tree algorithm, take the disaster factor index and influencing factor as input, predict the flood disaster risk level corresponding to the study area under different return periods, and quantitatively identify the importance of the disaster factor index;

[0041] The model interpretation unit is used to introduce the LIME model, to locally interpret the prediction results of the GBDT model, and to clarify the contribution rate of each disaster factor indicator;

[0042] A flood disaster risk assessment unit is used to conduct a flood disaster risk assessment on the study area by combining the contribution rate and the prediction results of the GBDT model;

[0043] The model verification unit is used to conduct flood disaster risk assessment on the study area in combination with the contribution rate and the prediction results of the GBDT model, to construct a flood disaster risk assessment model coupling LISFLOOD-FP and an interpretable GBDT algorithm, and to verify the applicability of the model.

[0044] The beneficial effects of the present invention are:

[0045] 1. The present invention constructs and uses the LISFLOOD-FP hydrodynamic model for simulation, which can simulate the key disaster factors that are difficult to obtain, such as flood depth, duration and flow rate, and realizes accurate simulation of urban floods. Combined with the gradient boosting decision tree (GBDT) algorithm, it can predict the flood disaster risk level under different return periods, and quantify the importance of identifying disaster factors, thereby improving the accuracy of flood disaster risk assessment.

[0046] 2. This paper introduces the local interpretability model (LIME) to locally explain the prediction results of the GBDT model, clarifies the contribution rate of each disaster factor indicator, enhances the transparency and credibility of the model, and makes the evaluation results easier to understand and accept.

[0047] 3. The present invention constructs a flood disaster risk assessment model (LG-FRAM) that couples LISFLOOD-FP with an interpretable GBDT algorithm, and verifies the applicability of the model. The model can provide a scientific basis for urban flood prevention and disaster reduction, help formulate effective disaster prevention and mitigation measures and emergency plans, reduce urban flood disaster losses, and has high practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the flow of the flood disaster risk assessment method in the present invention;

[0049] Figure 2 This is a system architecture diagram of the flood disaster risk assessment system in the present invention. DETAILED DESCRIPTION

[0050] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0051] Example 1

[0052] like Figure 1 As shown, this embodiment proposes a flood disaster risk assessment method based on explainable artificial intelligence, and the method includes the following steps:

[0053] S1. Determine the influencing factors of flood disasters based on the historical flood disasters in the study area;

[0054] S2. Using the influencing factors, combined with the rainstorm intensity formula and the Chicago rainfall pattern, the LISFLOOD-FP hydrodynamic model is constructed and allowed to be simulated to obtain the disaster factor index;

[0055] S3. Combined with the gradient boosting decision tree algorithm, taking the disaster factor index and influencing factor as input, predict the flood disaster risk level corresponding to the study area under different return periods, and quantitatively identify the importance of the disaster factor index;

[0056] S4. Introduce the LIME model to locally explain the prediction results of the GBDT model and clarify the contribution rate of each disaster factor indicator;

[0057] S5. Combined with the contribution rate and the prediction results of the GBDT model, a flood disaster risk assessment is conducted on the study area.

[0058] Further preferably, determining the influencing factors of flood disasters based on historical flood disasters in the study area includes:

[0059] Collect historical flood disaster data, digital elevation model data, historical rainfall data, and urban infrastructure information in the study area;

[0060] Based on the above data, the influencing factors of flood disasters including rainfall, topography, river system distribution, and urban drainage capacity are screened and determined.

[0061] Further preferably, the influencing factors are used in combination with the rainstorm intensity formula and the Chicago rain type to construct and run the LISFLOOD-FP hydrodynamic model for simulation to obtain the disaster factor index, including:

[0062] Based on the rainstorm intensity formula and Chicago rainfall pattern, rainfall scenarios with different return periods are designed to determine flood processes with return periods of 20, 50, and 100 years, including parameters such as rainfall intensity, rainfall duration, and rain peak location;

[0063] Using the rainfall scenario and the LISFLOOD-FP hydrodynamic model, combined with the DEM data, boundary conditions and initial conditions of the study area, flood inundation simulation was performed;

[0064] Disaster factor indicators including flooding depth, flooding duration and flow velocity are extracted from the simulation results.

[0065] Further preferably, the gradient boosting decision tree algorithm is used to predict the flood disaster risk level corresponding to the study area under different return periods, and the importance of quantitatively identifying the disaster factor index is included:

[0066] Preprocess the collected data, including normalization, division into training set and test set, etc.;

[0067] Construct a GBDT model and input the preprocessed data into the GBDT model for training to learn the relationship between the disaster factor index and the flood disaster risk level;

[0068] The trained GBDT model is used to predict the flood disaster risk level of the study area under different return periods;

[0069] The importance of each disaster-causing factor indicator is quantitatively identified through the internal mechanism of the GBDT model.

[0070] Further preferably, the LIME model is introduced to locally interpret the prediction results of the GBDT model to clarify the contribution rate of each disaster factor indicator, including:

[0071] For each prediction result of the GBDT model, use LIME to generate perturbation data;

[0072] Use the GBDT model to predict the disturbance data and obtain the prediction results of the disturbance data;

[0073] Based on the perturbed data and its prediction results, train a set interpretable model;

[0074] The parameters of the interpretable model are analyzed to determine the disaster factor index and its contribution rate that have the greatest impact on the prediction results.

[0075] Further preferably, the flood disaster risk assessment of the study area in combination with the contribution rate and the prediction result of the GBDT model includes:

[0076] Based on the contribution rate of each disaster factor indicator and the prediction results of the GBDT model, a flood disaster risk assessment index system is constructed;

[0077] Using the evaluation index system to evaluate the flood disaster risk in the study area, and obtaining the evaluation results;

[0078] Based on the assessment results, formulate corresponding disaster prevention and mitigation measures and emergency plans.

[0079] Further preferably, the evaluation method further includes: combining the contribution rate and the prediction results of the GBDT model to conduct a flood disaster risk assessment on the study area, constructing a flood disaster risk assessment model that couples LISFLOOD-FP and an interpretable GBDT algorithm, and verifying the applicability of the model, including:

[0080] Use actual observation data or historical disaster data to verify the flood disaster risk assessment model;

[0081] Compare the model prediction results with actual observations or historical disaster data;

[0082] The applicability and accuracy of flood disaster risk assessment models were evaluated based on the degree of difference.

[0083] According to the above-mentioned embodiment of the present invention, the present invention can simulate and obtain the key disaster-causing factors that are difficult to obtain, such as flood depth, duration and flow rate, etc., by utilizing the LISFLOOD-FP hydrodynamic model. These factors are crucial for assessing the risk of flood disasters. In the process of model construction, combined with the existing data in the region, multiple typical and representative flood processes are selected for simulation, which improves the accuracy and reliability of the simulation. Combined with the GBDT algorithm, the present invention uses disaster-causing factor indicators and influencing factors as input to predict the corresponding flood disaster risk level of the study area under different recurrence periods. The GBDT algorithm can scientifically predict the flood risk level by learning the regular relationship between indicators and risks, and provide strong support for disaster prevention and mitigation. In addition, through the internal mechanism of the GBDT algorithm, the present invention can quantitatively identify the importance of each disaster-causing factor indicator, which helps to identify the factors that have the greatest impact on the risk of flood disasters, and provide a basis for formulating targeted disaster prevention measures.

[0084] like Figure 2 As shown, based on the same inventive concept, the present invention proposes a flood disaster risk assessment system based on explainable artificial intelligence, which is applied to execute the above risk assessment method, and the system includes:

[0085] The impact factor determination unit is used to determine the impact factor of flood disasters based on the historical flood disasters in the study area;

[0086] A hydrodynamic model simulation unit is used to construct and allow the LISFLOOD-FP hydrodynamic model to simulate using the influencing factors, combined with the rainstorm intensity formula and the Chicago rain type, to obtain a disaster factor index;

[0087] The risk assessment and factor importance quantification unit is used to combine the gradient boosting decision tree algorithm, take the disaster factor index and influencing factor as input, predict the flood disaster risk level corresponding to the study area under different return periods, and quantitatively identify the importance of the disaster factor index;

[0088] The model interpretation unit is used to introduce the LIME model, to locally interpret the prediction results of the GBDT model, and to clarify the contribution rate of each disaster factor indicator;

[0089] A flood disaster risk assessment unit is used to conduct a flood disaster risk assessment on the study area by combining the contribution rate and the prediction results of the GBDT model;

[0090] The model verification unit is used to conduct flood disaster risk assessment on the study area by combining the contribution rate and the prediction results of the GBDT model, construct a flood disaster risk assessment model that couples LISFLOOD-FP and interpretable GBDT algorithms, and verify the applicability of the model.

[0091] It should be noted here that each module in the above-mentioned risk assessment system corresponds to steps S1 to S5 in implementing the above-mentioned optimization method, and the instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned embodiment 1.

[0092] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated.

[0093] The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0094] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0095] In addition, each functional module in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0096] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features can be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A flood disaster risk assessment method based on explainable artificial intelligence, characterized in that: The method comprises: Determine the influencing factors of flood disasters based on historical flood disasters in the study area; By using the influencing factors, combined with the rainstorm intensity formula and the Chicago rainfall pattern, the LISFLOOD-FP hydrodynamic model was constructed and allowed to be simulated to obtain the disaster factor index; Combined with the gradient boosting decision tree algorithm, the disaster-causing factor index and influencing factor are used as input to predict the flood disaster risk level corresponding to the study area under different return periods, and quantitatively identify the importance of the disaster-causing factor index; The LIME model is introduced to partially explain the prediction results of the GBDT model and clarify the contribution rate of each disaster factor indicator; Combining the contribution rate and the prediction results of the GBDT model, a flood disaster risk assessment was conducted in the study area.

2. The flood disaster risk assessment method based on explainable artificial intelligence according to claim 1, characterized in that: The influencing factors of flood disasters determined based on historical flood disasters in the study area include: Collect historical flood disaster data, digital elevation model data, historical rainfall data, and urban infrastructure information in the study area; Based on the above data, the influencing factors of flood disasters including rainfall, topography, river system distribution, and urban drainage capacity are screened and determined.

3. The flood disaster risk assessment method based on explainable artificial intelligence according to claim 1 is characterized by: The influencing factors are used in combination with the rainstorm intensity formula and the Chicago rainfall pattern to construct and run the LISFLOOD-FP hydrodynamic model for simulation to obtain the disaster factor indicators, including: Based on the rainstorm intensity formula and Chicago rainfall pattern, rainfall scenarios with different return periods are designed to determine flood processes with return periods of 20, 50, and 100 years, including parameters such as rainfall intensity, rainfall duration, and rain peak location; Using the rainfall scenario and the LISFLOOD-FP hydrodynamic model, combined with the DEM data, boundary conditions and initial conditions of the study area, flood inundation simulation was performed; Disaster factor indicators including flooding depth, flooding duration and flow velocity are extracted from the simulation results.

4. The flood disaster risk assessment method based on explainable artificial intelligence according to claim 1 is characterized in that: The gradient boosting decision tree algorithm is used to predict the flood disaster risk level corresponding to the study area under different return periods, and to quantify the importance of identifying the disaster factor indicators, including: Preprocess the collected data, including normalization, division into training set and test set, etc.; Construct a GBDT model and input the preprocessed data into the GBDT model for training to learn the relationship between the disaster factor index and the flood disaster risk level; The trained GBDT model is used to predict the flood disaster risk level of the study area under different return periods; The importance of each disaster-causing factor indicator is quantitatively identified through the internal mechanism of the GBDT model.

5. The flood disaster risk assessment method based on explainable artificial intelligence according to claim 1 is characterized in that: The LIME model is introduced to partially explain the prediction results of the GBDT model and clarify the contribution rate of each disaster factor indicator, including: For each prediction result of the GBDT model, use LIME to generate perturbation data; Use the GBDT model to predict the disturbance data and obtain the prediction results of the disturbance data; Based on the perturbed data and its prediction results, train a set interpretable model; The parameters of the interpretable model are analyzed to determine the disaster factor index and its contribution rate that have the greatest impact on the prediction results.

6. The flood disaster risk assessment method based on explainable artificial intelligence according to claim 1 is characterized by: The flood disaster risk assessment of the study area is conducted by combining the contribution rate and the prediction results of the GBDT model, including: Based on the contribution rate of each disaster factor indicator and the prediction results of the GBDT model, a flood disaster risk assessment index system is constructed; Using the evaluation index system to evaluate the flood disaster risk in the study area, and obtaining the evaluation results; Based on the assessment results, formulate corresponding disaster prevention and mitigation measures and emergency plans.

7. The flood disaster risk assessment method based on explainable artificial intelligence according to claim 1 is characterized by: The evaluation method further includes: combining the contribution rate and the prediction results of the GBDT model to conduct a flood disaster risk assessment on the study area, constructing a flood disaster risk assessment model that couples LISFLOOD-FP and an interpretable GBDT algorithm, and verifying the applicability of the model, including: Use actual observation data or historical disaster data to verify the flood disaster risk assessment model; Compare the model prediction results with actual observations or historical disaster data; The applicability and accuracy of flood disaster risk assessment models were evaluated based on the degree of difference.

8. A flood disaster risk assessment system based on explainable artificial intelligence, characterized in that: Applied to executing the risk assessment method according to any one of claims 1 to 7, the system comprises: The impact factor determination unit is used to determine the impact factor of flood disasters based on the historical flood disasters in the study area; A hydrodynamic model simulation unit is used to construct and allow the LISFLOOD-FP hydrodynamic model to simulate using the influencing factors, combined with the rainstorm intensity formula and the Chicago rain type, to obtain a disaster factor index; The risk assessment and factor importance quantification unit is used to combine the gradient boosting decision tree algorithm, take the disaster factor index and influencing factor as input, predict the flood disaster risk level corresponding to the study area under different return periods, and quantitatively identify the importance of the disaster factor index; The model interpretation unit is used to introduce the LIME model, to locally interpret the prediction results of the GBDT model, and to clarify the contribution rate of each disaster factor indicator; A flood disaster risk assessment unit is used to conduct a flood disaster risk assessment on the study area by combining the contribution rate and the prediction results of the GBDT model; The model verification unit is used to conduct flood disaster risk assessment on the study area in combination with the contribution rate and the prediction results of the GBDT model, to construct a flood disaster risk assessment model coupling LISFLOOD-FP and an interpretable GBDT algorithm, and to verify the applicability of the model.

Images