Karst tunnel seasonal drainage method and system based on artificial intelligence

Abstract

The invention belongs to the technical field of karst tunnel drainage, and discloses a karst tunnel seasonal drainage method and system based on artificial intelligence. The method comprises the following steps: constructing a karst tunnel seasonal drainage system by using an Internet of Things communication technology; based on a karst tunnel seasonal drainage system, constructing a season and groundwater trend prediction model and a drainage control strategy generation model; according to the collected real-time monitoring data, using a season and groundwater trend prediction model to perform season and groundwater trend prediction; generating a drainage control strategy by using a drainage control strategy generation model according to a real-time season and groundwater trend prediction result; and based on the karst tunnel seasonal drainage system, according to the real-time drainage control strategy, karst tunnel seasonal drainage is conducted, and real-time monitoring data collection continues to be conducted. According to the invention, the problems of dependence on artificial experience, lack of real-time performance, limited data processing capability and lack of intelligence in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of karst tunnel drainage, and particularly relates to a seasonal drainage method and system for karst tunnels based on artificial intelligence. Background Technique

[0002] The main source of karst tunnel diseases is karst groundwater. On the one hand, the erosiveness of karst groundwater harms the tunnel lining structure and the waterproof and drainage system. On the other hand, karst groundwater is closely related to surface hydrology. During the rainy season, the karst control level surges suddenly, resulting in groundwater gushing. The traditional drainage system design cannot meet the drainage demand of the instantaneous large water volume during the rainy season, causing the lining structure to bear high external water pressure, thus endangering the safety of the tunnel structure and even causing cracks in the tunnel structure and leakage diseases, seriously affecting the safe operation of the tunnel. Therefore, aiming at the seasonal drainage problem of water-rich karst tunnels, developing a new type of karst tunnel drainage technology that can adapt to the large drainage volume demand during the rainy season has become an urgent tunnel engineering problem to be solved.

[0003] In the technical field of karst tunnel drainage, although some technologies and methods have been applied, they still have many defects, specifically as follows: 1) Relying on manual experience: Traditional drainage technologies often rely on manual experience and static drainage strategies. These strategies are usually based on historical data and general geological conditions and are difficult to adapt to complex and changeable geological environments and climate conditions, resulting in poor drainage effects and even exacerbating disasters in some cases; 2) Lack of real-time performance: Traditional drainage technologies usually lack the ability of real-time monitoring and dynamic adjustment. Based on preset drainage parameters and fixed drainage strategies, they cannot be flexibly adjusted according to real-time geological and climate conditions, resulting in the inability to respond in a timely manner in case of emergencies and increasing the risk of tunnel operation; 3) Limited data processing capacity: Traditional drainage technologies may have bottlenecks in processing a large amount of real-time data, unable to collect and process a large amount of real-time monitoring data, and not applicable to large-scale drainage control scenarios; 4) Lack of intelligence: Existing drainage technologies usually lack intelligent decision-making capabilities and cannot automatically generate optimal drainage control strategies based on real-time monitoring data, requiring manual intervention and adjustment. This not only increases the operation cost but also may lead to decision-making mistakes due to human factors. Summary of the Invention

[0004] In order to solve the problems of relying on manual experience, lack of real-time performance, limited data processing capacity, and lack of intelligence existing in the prior art, the purpose of the present invention is to provide a seasonal drainage method and system for karst tunnels based on artificial intelligence.

[0005] The technical solution adopted by the present invention is as follows: A seasonal drainage method for karst tunnels based on artificial intelligence, comprising the following steps: Based on several drainage control sections in the karst tunnel, use Internet of Things communication technology to build a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture; Based on the seasonal drainage system for karst tunnels, use artificial intelligence algorithms to construct a seasonal and groundwater trend prediction model and a drainage control strategy generation model for karst tunnels; According to the collected real-time monitoring data, use the seasonal and groundwater trend prediction model to conduct seasonal and groundwater trend prediction, and obtain the real-time seasonal and groundwater trend prediction results; According to the real-time seasonal and groundwater trend prediction results, use the drainage control strategy generation model to generate a drainage control strategy, and obtain the real-time drainage control strategy; Based on the seasonal drainage system for karst tunnels, according to the real-time drainage control strategy, conduct seasonal drainage of the karst tunnel and continue to collect real-time monitoring data.

[0006] Furthermore, the real-time monitoring data includes real-time meteorological and hydrological monitoring data and real-time karst tunnel monitoring data; The real-time meteorological and hydrological monitoring data includes real-time ground meteorological monitoring data, real-time upper-air meteorological monitoring data, and real-time ground hydrological monitoring data; The real-time karst tunnel monitoring data includes real-time karst tunnel hydrological monitoring data and real-time karst tunnel surrounding rock stress monitoring data; The real-time seasonal and groundwater trend prediction results include real-time seasonal prediction results and real-time groundwater trend prediction results; The real-time drainage control strategy includes real-time drainage control decisions for each drainage control section.

[0007] Furthermore, based on several drainage control sections in the karst tunnel, use Internet of Things communication technology to build a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture, including the following steps: For each drainage control section in the karst tunnel, set a corresponding Internet of Things base station, karst tunnel monitoring device, drainage control device, and drainage execution device; Connect the karst tunnel monitoring device and the drainage control device located in the same drainage control section to the corresponding Internet of Things base station, and connect the drainage execution device to the corresponding drainage control device; Cascade all the Internet of Things base stations in the karst tunnel, and connect the Internet of Things base station at the head end to the cloud data center; Set meteorological and hydrological monitoring devices for the area where the karst tunnel is located, and connect the meteorological and hydrological monitoring devices to the cloud data center to obtain a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture.

[0008] Furthermore, based on the seasonal drainage system of karst tunnels, an artificial intelligence algorithm is used to construct a groundwater trend prediction model and a drainage control strategy generation model for karst tunnels, including the following steps: Use the karst tunnel monitoring device and meteorological and hydrological monitoring device of the seasonal drainage system of karst tunnels to collect a number of historical monitoring data of karst tunnels in different seasons, and preprocess the number of historical monitoring data to obtain a number of preprocessed historical monitoring data; Use the cloud data center of the seasonal drainage system of karst tunnels. According to a number of preprocessed historical monitoring data, use a deep learning algorithm to construct a season and groundwater trend prediction model, and generate a number of historical season and groundwater trend prediction results; Based on a number of cascaded drainage control devices in karst tunnels, according to a number of historical season and groundwater trend prediction results, use an enhanced reinforcement learning algorithm to construct a drainage control strategy generation model, and generate a number of historical drainage control strategy generation experiences.

[0009] Furthermore, the season and groundwater trend prediction model is constructed based on the RF-Attention-MLP-IFWA -SVM algorithm, and the season and groundwater trend prediction model includes a key feature screening module constructed based on the RF algorithm, a weighted fusion module constructed based on the Attention mechanism, a groundwater trend prediction module constructed based on the MLP algorithm, a prediction result optimization module constructed based on the IFWA algorithm, and a season prediction module constructed based on the SVM algorithm. The weighted fusion module is respectively connected to the key feature screening module, the groundwater trend prediction module, and the season prediction module, and the groundwater trend prediction module is connected to the prediction result optimization module; The drainage control strategy generation model is constructed based on the MPO-MOCMAGRPO algorithm, and the drainage control strategy generation model includes a meta-strategy optimization module constructed based on the MPO algorithm and a drainage control strategy generation module constructed based on the MOCMAGRPO algorithm connected in sequence. The drainage control strategy generation module includes a number of cascaded agents, as well as a general set of objective functions, an Actor network, and an experience replay pool. Each agent is respectively connected to the set of objective functions, the Actor network, and the experience replay pool, and each agent corresponds to a drainage control device. The Actor network is connected to the meta-strategy optimization module.

[0010] Furthermore, based on a number of cascaded drainage control devices in karst tunnels, according to a number of historical season and groundwater trend prediction results, use an enhanced reinforcement learning algorithm to construct a drainage control strategy generation model, and generate a number of historical drainage control strategy generation experiences, including the following steps: Use the MPO-MOCMAGRPO algorithm to construct an initial drainage control strategy generation model; the drainage control strategy generation model includes an initial meta-strategy optimization module and an initial drainage control strategy generation module; the initial drainage control strategy generation module includes several initial agents; Use the prediction results of several historical seasons and groundwater trends to train and optimize the initial meta-strategy optimization module to obtain the final meta-strategy optimization module; Cascade several initial agents in the order of drainage control sections, and use the prediction results of several historical seasons and groundwater trends to train and optimize the initial drainage control strategy generation module to obtain the final drainage control strategy generation module, and generate several historical drainage control strategy generation experiences; Combine the final meta-strategy optimization module and the final drainage control strategy generation module to obtain the final drainage control strategy generation model, and store several historical drainage control strategy generation experiences in the corresponding experience replay pool.

[0011] Furthermore, according to the collected real-time monitoring data, use the season and groundwater trend prediction model to predict the season and groundwater trend to obtain the real-time season and groundwater trend prediction results, including the following steps: Use the karst tunnel monitoring device of the karst tunnel seasonal drainage system to collect the real-time karst tunnel monitoring data of the drainage control section, and upload it to the cloud data center through the cascaded Internet of Things base stations; Use the meteorological and hydrological monitoring device of the karst tunnel seasonal drainage system to collect the real-time meteorological and hydrological monitoring data, and upload it to the cloud data center; Use the cloud data center to combine the real-time meteorological and hydrological monitoring data and all real-time karst tunnel monitoring data to obtain the real-time monitoring data, and perform preprocessing to obtain the preprocessed real-time monitoring data; Use the key feature screening module of the season and groundwater trend prediction model to extract several real-time key data from the preprocessed real-time monitoring data; Use the weighted fusion module of the season and groundwater trend prediction model to perform weighted fusion on several real-time key data according to the preset first attention weight value to obtain the first real-time weighted fusion feature; Use the season prediction module of the season and groundwater trend prediction model to perform season prediction according to the first real-time weighted fusion feature to obtain the real-time season prediction result; Use the weighted fusion module of the season and groundwater trend prediction model to perform weighted fusion on several real-time key data according to the preset second attention weight value to obtain the second real-time weighted fusion feature; The groundwater trend prediction module using the seasonal and groundwater trend prediction model predicts the groundwater trend based on the second real-time weighted fusion feature to obtain the real-time groundwater trend prediction probability distribution; The prediction result optimization module using the seasonal and groundwater trend prediction model optimizes the real-time groundwater trend prediction probability distribution to obtain the corresponding real-time groundwater trend prediction result; Combining the real-time season prediction result and the real-time groundwater trend prediction result to obtain the corresponding real-time season and groundwater trend prediction result.

[0012] Furthermore, according to the real-time season and groundwater trend prediction result, using the drainage control strategy generation model, the drainage control strategy is generated to obtain the real-time drainage control strategy, including the following steps: According to the real-time season and groundwater trend prediction result, using the meta-strategy optimization module of the drainage control strategy generation model, the Actor network of the drainage control strategy generation module is adjusted to obtain the adjusted Actor network; According to the experience generated by randomly extracting several historical drainage control strategies in the experience pool and the real-time season and groundwater trend prediction result, using the first intelligent agent in the drainage control strategy generation module to control the adjusted Actor network to generate the first real-time drainage control decision corresponding to the first drainage control device; According to the experience generated by several historical drainage control strategies, the real-time season and groundwater trend prediction result, and the (M - 1)th real-time drainage control decision, using the Mth intelligent agent in the drainage control strategy generation module to control the adjusted Actor network to generate the Mth real-time drainage control decision corresponding to the Mth drainage control device, where M is the total number of drainage control sections greater than or equal to 2; Integrating all the real-time drainage control decisions from the first to the Mth to obtain the real-time drainage control strategy of the karst tunnel seasonal drainage system.

[0013] Furthermore, based on the karst tunnel seasonal drainage system, according to the real-time drainage control strategy, the karst tunnel seasonal drainage is carried out, and the real-time monitoring data collection continues, including the following steps: Using the cloud data center of the karst tunnel seasonal drainage system, the real-time drainage control strategy is sent to all the Internet of Things base stations of the karst tunnel; Using the Internet of Things base stations to analyze the real-time drainage control strategy to obtain the corresponding real-time drainage control decision, and sending the real-time drainage control decision to the corresponding drainage control device; Using the drainage control device, according to the received real-time drainage control decision, generating a real-time drainage control instruction, and sending the real-time drainage control instruction to the corresponding drainage execution device; Use the drainage execution device to execute the real-time drainage control instruction and perform seasonal drainage in the corresponding drainage control section; Until all the drainage execution devices in the karst tunnel complete the seasonal drainage in the corresponding drainage control section, realize the seasonal drainage of the karst tunnel, and continue to collect real-time monitoring data.

[0014] A seasonal drainage system for karst tunnels based on artificial intelligence is used to realize the seasonal drainage method of karst tunnels. The system includes a cloud data center, a meteorological and hydrological monitoring device, several Internet of Things base stations, several karst tunnel monitoring devices, several drainage control devices, and several drainage execution devices; The cloud data center is respectively communicatively connected with the meteorological and hydrological monitoring device and the Internet of Things base station at the head end in the karst tunnel. Several Internet of Things base stations are cascaded. Each Internet of Things base station is arranged in the corresponding drainage control section in the karst tunnel, and each Internet of Things base station is communicatively connected with a karst tunnel monitoring device and a drainage execution device in the corresponding drainage control section; The cloud data center includes an artificial intelligence model construction unit, a season and groundwater trend prediction unit, and a drainage control strategy generation unit connected in sequence.

[0015] The beneficial effects of the present invention are as follows: A seasonal drainage method and system for karst tunnels based on artificial intelligence provided by the present invention uses artificial intelligence algorithms to construct a season and groundwater trend prediction model and a drainage control strategy generation model, getting rid of the excessive dependence on artificial experience, and being able to make predictions and decisions based on real-time data and scientific algorithms, more accurately coping with complex and changeable geological environments and climate conditions; through the Internet of Things communication technology, real-time monitoring data is collected, and artificial intelligence algorithms are used for real-time analysis and prediction, realizing the real-time and dynamic nature of drainage control, being able to respond to water level changes in a timely manner, and avoiding tunnel safety risks caused by response lags; based on the edge-cloud collaborative architecture of the seasonal drainage system for karst tunnels and the artificial intelligence model, large-scale real-time monitoring data can be efficiently collected, processed, and analyzed, providing a reliable data basis for seasonal drainage control, and ensuring the accuracy of predictions and decisions; the optimal drainage control strategy is automatically generated through the drainage control strategy generation model, avoiding errors and delays caused by manual intervention, ensuring the scientific and timely nature of the drainage strategy, effectively reducing the external water pressure borne by the tunnel lining structure, ensuring the safety of the tunnel structure, and preventing the occurrence of cracking and water leakage diseases.

[0016] Other beneficial effects of the present invention will be further described in the specific implementation manner. Description of the Drawings

[0017] Figure 1 It is a flowchart of the seasonal drainage method for karst tunnels based on artificial intelligence in the present invention.

[0018] Figure 2 It is the structural block diagram of the seasonal drainage system for karst tunnels based on artificial intelligence in the present invention. Specific implementation manners

[0019] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0020] Embodiment 1: As Figure 1 shown, this embodiment provides a seasonal drainage method for karst tunnels based on artificial intelligence, including the following steps: S1: Based on several drainage control sections in the karst tunnel, using Internet of Things communication technology, build a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture, including the following steps: S1-1: For each drainage control section in the karst tunnel, set a corresponding Internet of Things base station, karst tunnel monitoring device, drainage control device, and drainage execution device; S1-2: Connect the karst tunnel monitoring device and the drainage control device located in the same drainage control section to the corresponding Internet of Things base station, and connect the drainage execution device to the corresponding drainage control device; S1-3: Cascade all the Internet of Things base stations in the karst tunnel, and connect the Internet of Things base station at the head end to the cloud data center; S1-4: Set meteorological and hydrological monitoring devices for the area where the karst tunnel is located, and connect the meteorological and hydrological monitoring devices to the cloud data center to obtain a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture; S2: Based on the seasonal drainage system for karst tunnels, use artificial intelligence algorithms to construct a seasonal and groundwater trend prediction model and a drainage control strategy generation model for karst tunnels, including the following steps: S2-1: Use the karst tunnel monitoring device and the meteorological and hydrological monitoring device of the seasonal drainage system for karst tunnels to collect a number of historical monitoring data of the karst tunnel in different seasons, and preprocess the number of historical monitoring data to obtain a number of preprocessed historical monitoring data; The historical monitoring data includes historical meteorological and hydrological monitoring data and historical karst tunnel monitoring data; The historical meteorological and hydrological monitoring data includes historical ground meteorological monitoring data, historical upper-air meteorological monitoring data, and real-time ground hydrological monitoring data; The ground meteorological monitoring data includes ground air temperature, ground air pressure, ground humidity, ground wind speed, ground wind direction, ground precipitation, ground cloud amount, ground visibility, ground weather phenomena (such as rain, snow, fog, etc.), ground snow depth, etc.; The upper-air meteorological monitoring data includes upper-air temperature, upper-air pressure, upper-air wind speed, upper-air wind direction, upper-air dew point temperature, upper-air geopotential height, etc.; Historical karst tunnel monitoring data includes historical karst tunnel hydrogeological monitoring data and historical karst tunnel surrounding rock stress monitoring data; Karst tunnel hydrogeological monitoring data includes groundwater level, karst tunnel temperature and humidity, groundwater pressure, groundwater flow velocity, groundwater flow direction, etc.; Karst tunnel surrounding rock stress monitoring data includes maximum principal stress, minimum principal stress, shear stress, etc.; Preprocessing includes data cleaning, format conversion, magnitude normalization, etc. performed in sequence, which improves data quality and provides data support for subsequent model construction; S2-2: Use the cloud data center of the karst tunnel seasonal drainage system. According to a number of preprocessed historical monitoring data, use deep learning algorithms to construct a seasonal and groundwater trend prediction model and generate a number of historical seasonal and groundwater trend prediction results; The seasonal and groundwater trend prediction model is constructed based on the Random Forest (RF)-Attention-Multilayer Perceptron (MLP)-Improved Fireworks Optimization Algorithm (IFWA)-Support Vector Machine (SVM) algorithm. And the seasonal and groundwater trend prediction model includes a key feature screening module constructed based on the RF algorithm, a weighted fusion module constructed based on the Attention mechanism, a groundwater trend prediction module constructed based on the MLP algorithm, a prediction result optimization module constructed based on the IFWA algorithm, and a seasonal prediction module constructed based on the SVM algorithm. The weighted fusion module is respectively connected to the key feature screening module, the groundwater trend prediction module, and the seasonal prediction module, and the groundwater trend prediction module is connected to the prediction result optimization module; The key feature screening module screens the feature components of the input monitoring data through the internal Classification And Regression Tree (CART). It can handle a large number of feature components, generate the key feature importance scores of each feature component, select the most stable and most discriminatory key feature components according to the key feature importance scores. The trained key feature screening module can directly screen the newly input monitoring data according to the selected key features, obtain a number of key features most relevant to the seasonal and groundwater trend prediction, reduce noise interference by eliminating redundant and irrelevant features, make the model more focused on key information, thereby improving the accuracy of prediction, reducing the dimension of input features, reducing the computational complexity of model training and prediction, accelerating the calculation speed. The selected key features help to better understand the main factors affecting the seasonal and groundwater trend, and enhance the interpretability of the model; Considering the different characteristic considerations between seasonal prediction and groundwater trend prediction, the weighted fusion module assigns different attention weights to different key features, ensuring the accuracy and emphasis of seasonal prediction and groundwater trend prediction, and performing weighted fusion on several key features from the key feature screening module; The groundwater trend prediction module uses the MLP algorithm to model and analyze the weighted fusion key features. It has a powerful non-linear modeling ability and can capture the complex non-linear relationships in the groundwater trend; The prediction result optimization module uses the IFWA algorithm to optimize the output result of the groundwater trend prediction module, which can avoid local optimal solutions, find prediction results closer to the true values, reduce the fluctuations of the prediction results, and improve the stability and reliability of the prediction; The SVM algorithm of the seasonal prediction module has a high accuracy rate in classification and prediction problems, and can accurately judge the current season and season-related characteristics, providing seasonal features for the subsequent generation of drainage control strategies; S2-3: Based on a number of drainage control devices in the karst tunnel in cascade, according to the prediction results of several historical seasons and groundwater trends, use the enhanced reinforcement learning algorithm to construct a drainage control strategy generation model and generate a number of historical drainage control strategy generation experiences; The drainage control strategy generation model is constructed based on the Meta-Policy Optimization (MPO)-Multi-Objective Cascading Multi-Agent Relative Policy Optimization (MOCMAGRPO) algorithm. The drainage control strategy generation model includes a meta-policy optimization module constructed based on the MPO algorithm and a drainage control strategy generation module constructed based on the MOCMAGRPO algorithm connected in sequence. The drainage control strategy generation module includes several agents in cascade, as well as a general set of objective functions, an Actor network, and an experience replay pool. Each agent is respectively connected to the set of objective functions, the Actor network, and the experience replay pool, and each agent corresponds to a drainage control device. The Actor network is connected to the meta-policy optimization module; The meta-policy optimization module is used for the initial network parameters of the Actor network in the drainage control strategy generation module, so that these parameters can quickly adapt to new and unseen seasonal and groundwater trend prediction results, improving the generalization ability of the model. Even under unseen seasonal and groundwater trend prediction results, the Actor network can be updated based on previous learning experiences, improving the adaptability of the drainage control strategy generation model; The objective function set of the drainage control strategy generation module can handle multiple conflicting optimization objectives, such as drainage efficiency, drainage delay, drainage safety, etc., and generate drainage control strategies that balance these objectives; The agent learns historical experiences through the experience replay pool, continuously optimizing its decision-making generation ability. The agent controls the Actor network based on the learned experiences to generate more effective drainage control decisions. The cascaded setting of the agents enables the drainage control decision of the previous agent to affect the generation of the next drainage control decision, strengthening the drainage control linkage in different drainage control sections and avoiding the gradual spread of drainage errors caused by unreasonable previous drainage control decisions, which may affect the final drainage control strategy; The design of the experience replay pool and the agent enables the model to continuously learn and optimize, improving the quality of strategy generation. Since the drainage control strategy generation module adopts the method of population exploration, it can avoid falling into local optimal solutions to a certain extent. The Actor network outputs the distribution probability of actions in a given state, aiming to learn an optimal strategy, that is, to maximize the long-term cumulative reward. In a continuous action space, the Actor network usually outputs a mean value and optional variance parameters to describe the probability distribution of actions. The drainage control strategy generation module directly updates the Actor network through gradients, eliminating the Critic network in traditional reinforcement learning and making the algorithm structure simpler; Based on a number of cascaded drainage control devices in a karst tunnel, according to the prediction results of a number of historical seasons and groundwater trends, using an enhanced reinforcement learning algorithm, a drainage control strategy generation model is constructed, and a number of historical drainage control strategy generation experiences are generated, including the following steps: S2-3-1: Use the MPO-MOCMAGRPO algorithm to construct an initial drainage control strategy generation model; the drainage control strategy generation model includes an initial meta-policy optimization module and an initial drainage control strategy generation module; the initial drainage control strategy generation module includes a number of initial agents; S2-3-2: Use the prediction results of a number of historical seasons and groundwater trends to train and optimize the initial meta-policy optimization module to obtain the final meta-policy optimization module, including the following steps: S2-3-2-1: Use the initial network parameters of the Actor network of the initial drainage control strategy generation module as the output parameters of the initial meta-policy optimization module; S2-3-2-2: Use the problem of generating the initial network parameters of the Actor network as the scenario of the initial meta-policy optimization module, and divide the scenario into a number of sub-scenarios according to different optimization objectives; S2-3-2-3: Based on several sub-scenarios, using several historical seasons and groundwater trend prediction results, train and optimize the initial meta-strategy optimization module to obtain the final meta-strategy optimization module; S2-3-3: Cascade several initial agents in the order of drainage control sections, and use several historical seasons and groundwater trend prediction results to train and optimize the initial drainage control strategy generation module to obtain the final drainage control strategy generation module, and generate several historical drainage control strategy generation experiences, including the following steps: S2-3-3-1: Set a set of objective functions, an experience replay pool, and an Actor network for the initial drainage control strategy generation module; S2-3-3-2: Regard the drainage control strategy generation problem as the simulation environment of the initial drainage control strategy generation module, and regard the drainage control decision generation problem as the simulation sub-environment of the initial agent; S2-3-3-3: Cascade several initial agents in the order of drainage control sections, use the output decision of the previous initial agent as the input information of the next initial agent, and set the corresponding action space and state space for the initial agent according to the input and output information of the drainage control device; S2-3-3-4: Use several historical seasons and groundwater trend prediction results to train and optimize the initial drainage control strategy generation module to obtain the final drainage control strategy generation module, and generate several historical drainage control strategy generation experiences; S2-3-4: Combine the final meta-strategy optimization module and the final drainage control strategy generation module to obtain the final drainage control strategy generation model, and store several historical drainage control strategy generation experiences in the corresponding experience replay pool; S3: According to the collected real-time monitoring data, use the season and groundwater trend prediction model to perform season and groundwater trend prediction to obtain the real-time season and groundwater trend prediction results; The real-time monitoring data includes real-time meteorological and hydrological monitoring data and real-time karst tunnel monitoring data; The real-time meteorological and hydrological monitoring data includes real-time ground meteorological monitoring data, real-time upper-air meteorological monitoring data, and real-time ground hydrological monitoring data; The real-time karst tunnel monitoring data includes real-time karst tunnel hydrological monitoring data and real-time karst tunnel surrounding rock stress monitoring data; The real-time season and groundwater trend prediction results include real-time season prediction results and real-time groundwater trend prediction results; According to the collected real-time monitoring data, use the season and groundwater trend prediction model to perform season and groundwater trend prediction to obtain the real-time season and groundwater trend prediction results, including the following steps: S3-1: The karst tunnel monitoring device using the seasonal drainage system of the karst tunnel collects the real-time karst tunnel monitoring data in the drainage control section and uploads it to the cloud data center through cascaded Internet of Things base stations; S3-2: The meteorological and hydrological monitoring device using the seasonal drainage system of the karst tunnel collects the real-time meteorological and hydrological monitoring data and uploads it to the cloud data center; S3-3: Using the cloud data center, combine the real-time meteorological and hydrological monitoring data and all real-time karst tunnel monitoring data to obtain real-time monitoring data, and perform preprocessing to obtain preprocessed real-time monitoring data; S3-4: Use the key feature screening module of the seasonal and groundwater trend prediction model to extract several real-time key data from the preprocessed real-time monitoring data; S3-5: Use the weighted fusion module of the seasonal and groundwater trend prediction model to perform weighted fusion on several real-time key data according to the preset first attention weight value to obtain the first real-time weighted fusion feature; S3-6: Use the seasonal prediction module of the seasonal and groundwater trend prediction model to perform seasonal prediction based on the first real-time weighted fusion feature to obtain the real-time seasonal prediction result; The real-time seasonal prediction result includes the real-time seasonal classification prediction result (four seasons, rainy season, dry season, flood season, etc.), the real-time seasonal fluctuation law prediction result (seasonal characteristics: analyze the seasonal change law of the groundwater level, such as the water level rising in the rainy season and falling in the dry season; periodic change: identify the periodic fluctuation of the groundwater level, such as the water level change cycle within a year or a quarter; abnormal fluctuation warning: by comparing historical data, mark the abnormal fluctuations that do not conform to the seasonal law, which may be caused by extreme weather or geological activities); S3-7: Use the weighted fusion module of the seasonal and groundwater trend prediction model to perform weighted fusion on several real-time key data according to the preset second attention weight value to obtain the second real-time weighted fusion feature; S3-8: Use the groundwater trend prediction module of the seasonal and groundwater trend prediction model to perform groundwater trend prediction based on the second real-time weighted fusion feature to obtain the real-time groundwater trend prediction probability distribution; S3-9: Use the prediction result optimization module of the seasonal and groundwater trend prediction model to optimize the real-time groundwater trend prediction probability distribution to obtain the corresponding real-time groundwater trend prediction result, including the following steps: S3-9-1: Use the prediction result optimization module of the seasonal and groundwater trend prediction model to encode the real-time probability distribution compensation parameter corresponding to the real-time groundwater trend prediction probability distribution as the individual vector of the IFWA individual in the IFWA algorithm; S3-9-2: Set the fitness function and IFWA population parameters of the IFWA algorithm with the goal of minimizing the mean square error;

[0021] In the formula, is the fitness function; is the mean square error function; S3-9-3: According to the IFWA population parameters and individual vectors, use the Circle chaotic mapping sequence for initialization to generate several initial IFWA individuals of the initial IFWA population; The formula is:

[0022] In the formula, is the initial IFWA individual of the Circle chaotic mapping; is the randomly generated initial IFWA individual; is the IFWA individual indicator; S3-9-4: According to the fitness function, obtain the explosion radius, the number of sparks, and the fitness value of the initial IFWA individuals; The formula is:

[0023] In the formula, is the initial IFWA individual the number of sparks of; is the number constant; is the maximum fitness value in the initialized IFWA population; is the initial IFWA individual the fitness value of; is the infinitesimal constant; is the convergence factor; is a non-zero positive real number;

[0024] In the formula, is the initial IFWA individual the explosion radius of; is the explosion radius adjustment constant; is the minimum fitness value in the initialized IFWA population; is the total number of IFWA individuals;

[0025] In the formula, is the convergence factor; tanh(.) is the hyperbolic tangent function; is the iteration indicator; is the maximum number of iterations; a max and a min are the maximum and minimum values of the convergence factor respectively; λ is the decreasing rate parameter, is the decreasing period parameter, λ = -2 π , = π ; The number of sparks determines the number of sub - fireworks generated after each firework explosion, and the explosion radius determines the distribution range of the sparks generated after the firework explosion in the solution space. In the early stage of iteration, a has a larger value, the number of sparks of the IFWA individuals is smaller, and the explosion radius is larger, which helps to reduce the computational burden and is more widely distributed, helping to explore more solution spaces. In the later stage of iteration, a smaller explosion radius helps to perform fine - grained search in the local area, and a larger number of sparks helps to increase the diversity of the search; S3 - 9 - 5: According to the explosion radius, the number of sparks, and the fitness value, perform firework explosion to obtain a number of updated IFWA individuals of the updated IFWA population; The formula is:

[0026] In the formula, is the updated IFWA individual; is a random number from - 1 to 1; S3 - 9 - 6: Use the Gaussian mutation algorithm to perform Gaussian mutation on the initialized IFWA population to generate a number of Gaussian - mutated IFWA individuals of the Gaussian - mutated IFWA population; The formula is:

[0027] In the formula, is the Gaussian - mutated IFWA individual; is a random number of a Gaussian distribution with both mean and variance equal to 1; S3 - 9 - 7: Obtain the fitness values of each updated IFWA individual, Gaussian - mutated IFWA individual, and reverse IFWA individual, and take the IFWA individual with the minimum fitness value as the optimal individual; S3 - 9 - 8: If the number of iterations reaches the maximum number of iterations or the fitness value of the optimal individual meets the requirements, decode the individual vector of the optimal individual to obtain the optimal real - time probability distribution compensation parameter; S3-9-9: Optimize the real-time groundwater trend prediction probability distribution according to the optimal real-time probability distribution compensation parameters to obtain the corresponding real-time groundwater trend prediction result; The real-time groundwater trend prediction result includes the real-time groundwater level trend prediction result (predicted water level change: predict the rising or falling trend of the groundwater level in the future according to historical monitoring data and seasonal patterns; change range: predict the specific numerical range of the water level change, such as the number of meters the water level may rise or fall in the next month; key time nodes: indicate the time points when the water level reaches the peak or valley value to help formulate countermeasures in advance), the real-time trend change trend prediction result (long-term trend: predict the long-term change trend of the groundwater level, such as whether it shows an annual rising or falling trend; trend inflection point: identify the trend inflection point of the groundwater level, that is, the time point when it changes from rising to falling or from falling to rising; trend stability: evaluate whether the trend of the groundwater level is stable or whether there is a risk of mutation), the real-time groundwater recharge and discharge dynamic prediction result (recharge prediction: predict the groundwater recharge volume and its impact on the water level according to data such as precipitation and surface runoff; discharge prediction: combine the groundwater flow direction and the operation of the drainage system to predict the groundwater discharge volume and the water level decline trend; dynamic balance analysis: evaluate the dynamic balance state of groundwater recharge and discharge to judge whether there is a risk of over-exploitation or insufficient recharge), and the real-time groundwater environment change prediction result (karst tunnel alignment change, groundwater hydro-environment, karst tunnel surrounding rock change), etc.; S3-10: Combine the real-time season prediction result and the real-time groundwater trend prediction result to obtain the corresponding real-time season and groundwater trend prediction result; S4: According to the real-time season and groundwater trend prediction result, use the drainage control strategy generation model to generate a drainage control strategy to obtain the real-time drainage control strategy; The real-time drainage control strategy includes real-time drainage control decisions for each drainage control section; The real-time drainage control decision includes the real-time drainage equipment control actions of the drainage execution device (including the corresponding gate valves, solenoid valves or automatic control valves), real-time priority setting actions, real-time drainage volume control actions, real-time drainage time control actions, and real-time abnormal warning and emergency response actions, etc.; According to the real-time season and groundwater trend prediction result, use the drainage control strategy generation model to generate a drainage control strategy to obtain the real-time drainage control strategy, including the following steps: S4-1: According to the real-time season and groundwater trend prediction result, use the meta-strategy optimization module of the drainage control strategy generation model to adjust the Actor network of the drainage control strategy generation module to obtain the adjusted Actor network; S4-2: Generate experience and real-time seasonal and groundwater trend prediction results based on a number of historical drainage control strategies randomly selected from the experience pool. Use the first agent in the drainage control strategy generation module to control the adjusted Actor network and generate the first real-time drainage control decision corresponding to the first drainage control device, including the following steps: S4-2-1: Randomly select a number of historical drainage control strategies from the experience pool to generate experience. Generate a number of first possible drainage control actions based on the historical drainage control strategies, and adjust the first action space of the first agent according to the first possible drainage control actions to obtain the first adjusted action space; S4-2-2: Analyze the real-time seasonal and groundwater trend prediction results to obtain a number of real-time prediction states, and adjust the first state space of the first agent according to the first real-time prediction states to obtain the first adjusted state space; S4-2-3: Use the first agent in the drainage control strategy generation module to control the adjusted Actor network and generate the probability distribution of all first possible drainage control actions in the first adjusted action space corresponding to each first real-time prediction state in the first adjusted state space; S4-2-4: Take the first possible drainage control action with the highest probability distribution in the first adjusted action space as the executed drainage control action corresponding to the first real-time prediction state; S4-2-5: Integrate the first executed drainage control actions of all first real-time prediction states in the first adjusted state space to obtain the first real-time drainage control decision; S4-3: Generate experience based on a number of historical drainage control strategies, real-time seasonal and groundwater trend prediction results, and the (M - 1)th real-time drainage control decision. Use the Mth agent in the drainage control strategy generation module to control the adjusted Actor network and generate the Mth real-time drainage control decision corresponding to the Mth drainage control device, where M is the total number of drainage control sections greater than or equal to 2; S4-4: Integrate all the real-time drainage control decisions from the first to the Mth to obtain the real-time drainage control strategy for the karst tunnel seasonal drainage system; S5: Based on the karst tunnel seasonal drainage system, perform karst tunnel seasonal drainage according to the real-time drainage control strategy, and continue to collect real-time monitoring data, including the following steps: S5-1: Use the cloud data center of the karst tunnel seasonal drainage system to send the real-time drainage control strategy to all the Internet of Things base stations of the karst tunnel; S5-2: Use the Internet of Things base stations to analyze the real-time drainage control strategy to obtain the corresponding real-time drainage control decision, and send the real-time drainage control decision to the corresponding drainage control device; S5-3: using the drainage control device to generate a real-time drainage control instruction according to the received real-time drainage control decision, and sending the real-time drainage control instruction to the corresponding drainage execution device; S5-4: Use the drainage execution device to execute the real-time drainage control instruction to perform seasonal drainage in the corresponding drainage control section; S5-5: Until all drainage execution devices in the karst tunnel complete the seasonal drainage of the corresponding drainage control section, the seasonal drainage of the karst tunnel is realized, and real-time monitoring data collection continues.

[0028] Embodiment 2: like Figure 2 As shown, this embodiment provides a seasonal drainage system for karst tunnels based on artificial intelligence, which is used to implement a seasonal drainage method for karst tunnels. The system includes a cloud data center, a meteorological and hydrological monitoring device, several Internet of Things base stations, several karst tunnel monitoring devices, several drainage control devices, and several drainage execution devices; The cloud data center is respectively connected to the meteorological and hydrological monitoring devices and the Internet of Things base station located at the head end of the karst tunnel, and several Internet of Things base stations are cascaded. Each Internet of Things base station is set in a corresponding drainage control section in the karst tunnel, and each Internet of Things base station is connected to a karst tunnel monitoring device and a drainage execution device in the corresponding drainage control section. The IoT base station undertakes the edge computing tasks in the edge-cloud collaborative structure, and delegates data collection, preprocessing, and data transmission to the edge nodes, which reduces the computing pressure of the cloud data center and improves the efficiency of drainage control. It is used to preprocess the real-time monitoring data collected by the karst tunnel monitoring device and transmit it to the IoT base station or cloud data center of the previous drainage control section; it receives the real-time drainage control strategy sent by the cloud data center or the IoT base station of the previous drainage control section, and transmits the real-time drainage control strategy to the IoT base station of the next drainage control section. A karst tunnel monitoring device is used to collect real-time karst tunnel monitoring data of the drainage control section and transmit it to the IoT base station of the corresponding drainage control section; A drainage control device, used to receive the real-time drainage control strategy sent by the Internet of Things base station, and control the drainage execution device of the corresponding drainage control section according to the real-time drainage control strategy; Drainage execution device, used to execute real-time drainage control strategy and seasonal drainage of karst tunnels; The IoT base station, karst tunnel monitoring device, drainage control device and drainage execution device realize the regional drainage control of the drainage control section, dividing the overall drainage control of the karst tunnel into smaller control tasks, avoiding the large-scale impact caused by unreasonable drainage control strategy and reducing the cost of drainage control; The cloud data center includes an artificial intelligence model construction unit, a seasonal and groundwater trend prediction unit, and a drainage control strategy generation unit that are connected in sequence; The artificial intelligence model construction unit is used to construct a seasonal and groundwater trend prediction model and a drainage control strategy generation model for the karst tunnel based on the seasonal drainage system of the karst tunnel using artificial intelligence algorithms; The seasonal and groundwater trend prediction unit is used to perform seasonal and groundwater trend prediction according to the collected real-time monitoring data using the seasonal and groundwater trend prediction model to obtain the real-time seasonal and groundwater trend prediction results; The drainage control strategy generation unit is used to generate a drainage control strategy according to the real-time seasonal and groundwater trend prediction results using the drainage control strategy generation model to obtain a real-time drainage control strategy; and send the real-time drainage control strategy to the Internet of Things base station at the head end in the seasonal drainage system of the karst tunnel.

[0029] A method and system for seasonal drainage of karst tunnels based on artificial intelligence provided by the present invention uses artificial intelligence algorithms to construct a seasonal and groundwater trend prediction model and a drainage control strategy generation model, getting rid of the excessive dependence on manual experience, and being able to make predictions and decisions based on real-time data and scientific algorithms, more accurately coping with complex and changeable geological environments and climatic conditions; real-time monitoring data is collected through Internet of Things communication technology, and artificial intelligence algorithms are used for real-time analysis and prediction, realizing the real-time and dynamic nature of drainage control, being able to respond to water level changes in a timely manner, and avoiding tunnel safety risks caused by lagged response; the seasonal drainage system and artificial intelligence model of the karst tunnel based on the edge-cloud collaborative architecture can efficiently collect, process, and analyze large-scale real-time monitoring data, providing a reliable data basis for seasonal drainage control and ensuring the accuracy of predictions and decisions; the optimal drainage control strategy is automatically generated through the drainage control strategy generation model, avoiding errors and delays caused by manual intervention, ensuring the scientific nature and timeliness of the drainage strategy, effectively reducing the external water pressure borne by the tunnel lining structure, ensuring the safety of the tunnel structure, and preventing the occurrence of cracking and leakage diseases.

[0030] The present invention is not limited to the above optional implementation manners, and anyone can obtain other various forms of products under the inspiration of the present invention. The above specific implementation manners should not be construed as limiting the protection scope of the present invention, and the protection scope of the present invention should be defined by the claims, and the description can be used to interpret the claims.

Claims

1. A seasonal drainage method for karst tunnels based on artificial intelligence, characterized in that: It includes the following steps: Based on several drainage control sections in a karst tunnel, use Internet of Things (IoT) communication technology to build a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture; Based on the seasonal drainage system for karst tunnels, use artificial intelligence algorithms to build a seasonal and groundwater trend prediction model and a drainage control strategy generation model for karst tunnels; According to the collected real-time monitoring data, use the seasonal and groundwater trend prediction model to predict the seasonal and groundwater trends and obtain the real-time seasonal and groundwater trend prediction results; According to the real-time seasonal and groundwater trend prediction results, use the drainage control strategy generation model to generate a drainage control strategy and obtain the real-time drainage control strategy; Based on the seasonal drainage system for karst tunnels, according to the real-time drainage control strategy, conduct seasonal drainage of the karst tunnel and continue to collect real-time monitoring data.

2. The karst tunnel seasonal drainage method based on artificial intelligence according to claim 1, characterized in that: The real-time monitoring data includes real-time meteorological and hydrological monitoring data and real-time karst tunnel monitoring data; The real-time meteorological and hydrological monitoring data includes real-time ground meteorological monitoring data, real-time upper-air meteorological monitoring data, and real-time ground hydrological monitoring data; The real-time karst tunnel monitoring data includes real-time karst tunnel hydrological monitoring data and real-time karst tunnel surrounding rock stress monitoring data; The real-time seasonal and groundwater trend prediction results include real-time seasonal prediction results and real-time groundwater trend prediction results; The real-time drainage control strategy includes real-time drainage control decisions for each drainage control section.

3. The seasonal drainage method for karst tunnels based on artificial intelligence according to claim 2, characterized in that: Based on several drainage control sections in a karst tunnel, use Internet of Things (IoT) communication technology to build a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture, including the following steps: For each drainage control section in the karst tunnel, set up a corresponding IoT base station, karst tunnel monitoring device, drainage control device, and drainage execution device; Connect the karst tunnel monitoring device and the drainage control device located in the same drainage control section to the corresponding IoT base station, and connect the drainage execution device to the corresponding drainage control device; Cascade all the IoT base stations in the karst tunnel and connect the IoT base station at the head end to the cloud data center; Set up meteorological and hydrological monitoring devices for the area where the karst tunnel is located and connect the meteorological and hydrological monitoring devices to the cloud data center to obtain a seasonal drainage system for karst tunnels with an edge-cloud collaborative architecture.

4. The karst tunnel seasonal drainage method based on artificial intelligence according to claim 3, characterized in that: Based on the seasonal drainage system for karst tunnels, use artificial intelligence algorithms to build a groundwater trend prediction model and a drainage control strategy generation model for karst tunnels, including the following steps: Use the karst tunnel monitoring device and the meteorological and hydrological monitoring device of the seasonal drainage system for karst tunnels to collect a number of historical monitoring data of the karst tunnel in different seasons and preprocess the number of historical monitoring data to obtain a number of preprocessed historical monitoring data; Use the cloud data center of the seasonal drainage system for karst tunnels, according to the number of preprocessed historical monitoring data, use deep learning algorithms to build a seasonal and groundwater trend prediction model and generate a number of historical seasonal and groundwater trend prediction results; Based on a number of drainage control devices in cascade in karst tunnels, according to the prediction results of a number of historical seasons and groundwater trends, an enhanced reinforcement learning algorithm is used to construct a drainage control strategy generation model and generate a number of historical drainage control strategy generation experiences.

5. The seasonal drainage method for karst tunnels based on artificial intelligence according to claim 4, characterized in that: The season and groundwater trend prediction model is constructed based on the RF-Attention-MLP-IFWA -SVM algorithm, and the season and groundwater trend prediction model includes a key feature screening module constructed based on the RF algorithm, a weighted fusion module constructed based on the Attention mechanism, a groundwater trend prediction module constructed based on the MLP algorithm, a prediction result optimization module constructed based on the IFWA algorithm, and a season prediction module constructed based on the SVM algorithm. The weighted fusion module is respectively connected to the key feature screening module, the groundwater trend prediction module, and the season prediction module. The groundwater trend prediction module is connected to the prediction result optimization module; The drainage control strategy generation model is constructed based on the MPO-MOCMAGRPO algorithm, and the drainage control strategy generation model includes a meta-strategy optimization module constructed based on the MPO algorithm and a drainage control strategy generation module constructed based on the MOCMAGRPO algorithm that are connected in sequence. The drainage control strategy generation module includes a number of cascaded agents, as well as a general set of objective functions, an Actor network, and an experience replay pool. Each of the agents is respectively connected to the set of objective functions, the Actor network, and the experience replay pool, and each agent corresponds to a drainage control device. The Actor network is connected to the meta-strategy optimization module.

6. The seasonal drainage method for karst tunnels based on artificial intelligence according to claim 5, wherein: Based on a number of drainage control devices in cascade in karst tunnels, according to the prediction results of a number of historical seasons and groundwater trends, an enhanced reinforcement learning algorithm is used to construct a drainage control strategy generation model and generate a number of historical drainage control strategy generation experiences, including the following steps: Use the MPO-MOCMAGRPO algorithm to construct an initial drainage control strategy generation model; the drainage control strategy generation model includes an initial meta-strategy optimization module and an initial drainage control strategy generation module; the initial drainage control strategy generation module includes a number of initial agents; Use the prediction results of a number of historical seasons and groundwater trends to train and optimize the initial meta-strategy optimization module to obtain a final meta-strategy optimization module; Cascade a number of initial agents in the order of drainage control sections, and use the prediction results of a number of historical seasons and groundwater trends to train and optimize the initial drainage control strategy generation module to obtain a final drainage control strategy generation module and generate a number of historical drainage control strategy generation experiences; Combine the final meta-strategy optimization module and the final drainage control strategy generation module to obtain a final drainage control strategy generation model, and store a number of historical drainage control strategy generation experiences in the corresponding experience replay pool.

7. A seasonal drainage method for karst tunnels based on artificial intelligence according to claim 6, characterized in that: According to the collected real-time monitoring data, use the seasonal and groundwater trend prediction model to predict the season and groundwater trend, and obtain the real-time season and groundwater trend prediction results, including the following steps: Use the karst tunnel monitoring device of the karst tunnel seasonal drainage system to collect the real-time karst tunnel monitoring data of the drainage control section, and upload it to the cloud data center through the cascaded Internet of Things base stations; Use the meteorological and hydrological monitoring device of the karst tunnel seasonal drainage system to collect the real-time meteorological and hydrological monitoring data, and upload it to the cloud data center; Use the cloud data center to combine the real-time meteorological and hydrological monitoring data with all the real-time karst tunnel monitoring data to obtain the real-time monitoring data, and perform preprocessing to obtain the preprocessed real-time monitoring data; Use the key feature screening module of the seasonal and groundwater trend prediction model to extract several real-time key data from the preprocessed real-time monitoring data; Use the weighted fusion module of the seasonal and groundwater trend prediction model to perform weighted fusion on several real-time key data according to the preset first attention weight value to obtain the first real-time weighted fusion feature; Use the season prediction module of the seasonal and groundwater trend prediction model to perform season prediction according to the first real-time weighted fusion feature to obtain the real-time season prediction result; Use the weighted fusion module of the seasonal and groundwater trend prediction model to perform weighted fusion on several real-time key data according to the preset second attention weight value to obtain the second real-time weighted fusion feature; Use the groundwater trend prediction module of the seasonal and groundwater trend prediction model to perform groundwater trend prediction according to the second real-time weighted fusion feature to obtain the real-time groundwater trend prediction probability distribution; Use the prediction result optimization module of the seasonal and groundwater trend prediction model to optimize the real-time groundwater trend prediction probability distribution to obtain the corresponding real-time groundwater trend prediction result; Combine the real-time season prediction result and the real-time groundwater trend prediction result to obtain the corresponding real-time season and groundwater trend prediction result.

8. A seasonal drainage method for karst tunnels based on artificial intelligence according to claim 7, characterized in that: According to the real-time season and groundwater trend prediction result, use the drainage control strategy generation model to generate the drainage control strategy to obtain the real-time drainage control strategy, including the following steps: According to the real-time season and groundwater trend prediction result, use the meta-strategy optimization module of the drainage control strategy generation model to adjust the Actor network of the drainage control strategy generation module to obtain the adjusted Actor network; According to the experience of generating several historical drainage control strategies randomly selected from the experience pool and the real-time season and groundwater trend prediction result, use the first agent in the drainage control strategy generation module to control the adjusted Actor network to generate the first real-time drainage control decision corresponding to the first drainage control device; According to the experience of generating several historical drainage control strategies, the real-time season and groundwater trend prediction result, and the (M-1)th real-time drainage control decision, use the Mth agent in the drainage control strategy generation module to control the adjusted Actor network to generate the Mth real-time drainage control decision corresponding to the Mth drainage control device, where M is the total number of drainage control sections greater than or equal to 2; Integrate all real-time drainage control decisions from the first to the Mth to obtain the real-time drainage control strategy for the seasonal drainage system of the karst tunnel.

9. The seasonal drainage method for karst tunnels based on artificial intelligence according to claim 8, characterized in that: Based on the seasonal drainage system of the karst tunnel, according to the real-time drainage control strategy, conduct seasonal drainage of the karst tunnel and continue to collect real-time monitoring data, including the following steps: Use the cloud data center of the seasonal drainage system of the karst tunnel to send the real-time drainage control strategy to all Internet of Things base stations in the karst tunnel; Use the Internet of Things base stations to analyze the real-time drainage control strategy to obtain the corresponding real-time drainage control decision, and send the real-time drainage control decision to the corresponding drainage control device; Use the drainage control device to generate a real-time drainage control instruction according to the received real-time drainage control decision, and send the real-time drainage control instruction to the corresponding drainage execution device; Use the drainage execution device to execute the real-time drainage control instruction to conduct seasonal drainage of the corresponding drainage control section; Until all drainage execution devices in the karst tunnel complete the seasonal drainage of the corresponding drainage control section, realize the seasonal drainage of the karst tunnel, and continue to collect real-time monitoring data.

10. An artificial intelligence-based seasonal drainage system for karst tunnels, which is used to implement the seasonal drainage method for karst tunnels described in any one of claims 1-9, and is characterized in that: The described system includes a cloud data center, a meteorological and hydrological monitoring device, a number of Internet of Things base stations, a number of karst tunnel monitoring devices, a number of drainage control devices, and a number of drainage execution devices; The cloud data center is respectively communicatively connected to the meteorological and hydrological monitoring device and the Internet of Things base station at the head end in the karst tunnel. A number of the Internet of Things base stations are cascaded. Each Internet of Things base station is arranged in the corresponding drainage control section in the karst tunnel, and each Internet of Things base station is communicatively connected to a karst tunnel monitoring device and a drainage execution device in the corresponding drainage control section; The cloud data center includes an artificial intelligence model construction unit, a season and groundwater trend prediction unit, and a drainage control strategy generation unit that are connected in sequence.

Images