Mineral development and ecological protection collaborative zoning method and device based on deep reinforcement learning, electronic equipment and storage medium

Abstract

The present application belongs to the technical field of mineral development and ecological protection, and provides a mineral development and ecological protection collaborative zoning method and device based on deep reinforcement learning, an electronic device and a storage medium, comprising: determining the ecological sensitivity index and the resource development potential index of the target area; dividing according to the ecological sensitivity index and the resource development potential index to obtain the initial partition; using Markov decision process to determine the state transition probability equation of the target area; using the agent and deep reinforcement learning to predict according to the objective function, the reward function and the global collaborative reward function to obtain the development intensity prediction result; using long-term and short-term time scale optimization to process according to the development intensity prediction result to obtain the development and ecological protection collaborative zoning result. The present application has the beneficial effect of realizing the dynamic coordination management of the development and environmental protection of the mining area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and the technical field of mineral development and environmental protection, and particularly relates to a mineral development and ecological protection collaborative zoning method and device based on deep reinforcement learning, an electronic device and a storage medium. BACKGROUND

[0002] The contradiction between mineral resource development and ecological environment protection is a key problem restricting sustainable development. Traditional static zoning methods cannot cope with the dynamic nature of mine development and the hysteresis of pollution, and cannot efficiently coordinate large-scale collaborative decision-making in multiple mining areas. SUMMARY

[0003] To at least solve one of the technical problems in the prior art, the present application provides a mineral development and ecological protection collaborative zoning method and device based on deep reinforcement learning, which realizes dynamic coordination and management of mine development and environmental protection.

[0004] One aspect of the present application provides a mineral development and ecological protection collaborative zoning method based on deep reinforcement learning, comprising:

[0005] According to the ecological index and the resource development index of the target area, the ecological sensitivity index and the resource development potential index of the target area are determined;

[0006] According to the ecological sensitivity index and the resource development potential index, the initial zoning is obtained by division, wherein the initial zoning includes zoning constraints and mine zoning types;

[0007] According to the dynamic environmental characteristics and the control factors of the target area, the state transition probability matrix of the target area is determined by using Markov decision process;

[0008] According to the mine zoning type, the zoning constraint and the state transition probability equation of the target area, the reward function and the global collaborative reward function are determined, and according to the objective function, the reward function and the global collaborative reward function, the agent and the deep reinforcement learning are used for prediction to obtain the development intensity prediction result;

[0009] According to the development intensity prediction result, the long-short term time scale optimization is used for processing to obtain the development and ecological protection collaborative zoning result.

[0010] According to the mineral development and ecological protection collaborative zoning method based on deep reinforcement learning, wherein the target area is divided according to the ecological sensitivity index and the resource development potential index to obtain the initial zoning, comprising:

[0011] The ecological sensitivity index is determined according to the ecological index and the weight of each ecological index is:

[0012] ;

[0013] Among them, the ecological index For vegetation coverage, Soil erosion modulus, pH value of the water body Air Quality Index. For species richness index, As an ecological vulnerability index;

[0014] The resource development potential index is determined based on the resource development index and the weight of each resource development index. for:

[0015] ;

[0016] Among them, the resource development index The remaining recoverable reserves, For ore grade, The depth of the ore deposit. To improve the convenience of transportation in the mining area.

[0017] According to the aforementioned deep reinforcement learning-based collaborative zoning method for mineral development and ecological protection, the initial zoning is determined using discriminant analysis based on the ecological sensitivity index and resource development potential index, including:

[0018] Based on the resource development potential index and ecological sensitivity index, the target area is divided using discriminant analysis to obtain initial zoning. The initial zoning includes core protection zones, restricted development zones, moderate development zones, and key development zones. Zoning constraints include development intensity and protection strength.

[0019] According to the aforementioned deep reinforcement learning-based collaborative zoning method for mineral development and ecological protection, the state transition probability matrix of the target area is determined using a Markov decision process based on the dynamic environmental characteristics and regulatory factors of the target area, including:

[0020] Acquire the dynamic environmental characteristics of the target area, including the current status, historical time series data, surrounding mining area correlation data, and mining area zoning type;

[0021] Obtain the regulatory factors for the target area, including the adjustment amount of development intensity, ecological input for protection, and parameters for correction of zoning boundaries;

[0022] Using dynamic environmental characteristics as the state space and control factors as the action space, the state transition probabilities are determined using a Gaussian distribution:

[0023] ;

[0024] in, Let be the transition probability function. This indicates that after taking action 'a', the state transitions to a new state. The probability distribution; It follows a Gaussian distribution; This indicates that after taking action a in state s, the user transitions to a new state. The mean, Obtained through training with historical data; This is a preset value, representing the variance of the new state obtained after taking action a in state s;

[0025] Adjust actions based on input development intensity Determine the state transition probability matrix based on the state transition probabilities. for:

[0026] ;

[0027] in, Remaining recoverable reserves Given the current development intensity, For time step;

[0028] And, determine the state transition probability equation for the pollution value, wherein the state transition probability equation for the pollution value is... for:

[0029] ;

[0030] in The cumulative pollution value, To develop pollution coefficients, The governance efficiency coefficient. It is noise, and , Investment in ecological protection measures;

[0031] The state transition probability equation of the target area is determined based on the state transition probability matrix of the remaining mineral reserves and the state transition probability equation of the pollution value.

[0032] According to the aforementioned deep reinforcement learning-based collaborative zoning method for mineral development and ecological protection, the reward function and global collaborative reward function are determined based on the mineral zoning type, zoning constraints, and state transition probability equation of the target area. Based on the target function, reward function, and global collaborative reward function, an agent and deep reinforcement learning are used for prediction to obtain the development intensity prediction result, including:

[0033] The global reward function at time t is determined based on the state transition probability matrix. for:

[0034] ;

[0035] Where n is the number of mining areas of the same type. This represents the mineral resource development revenue at time t for the i-th mining area of ​​the same type. Let be the development cost of the i-th similar mining area at time t. Let be the ecological environment restoration cost of the i-th mining area of ​​the same type at time t. The historical cumulative pollution impact cost of the i-th type of mining area at time t. For the i-th mining area of ​​the same type, the historical influence weighting coefficient is determined according to the mining area zoning type;

[0036] The reward function for each mining area type is determined based on the global reward function, the revenue from mineral resource development, the cost of ecological environment restoration, and the cost of historical cumulative pollution impact.

[0037] Based on the objective function, reward function, and global collaborative reward function, agents are assigned to different mining area zoning types. These agents process historical time-series data using a deep recursive Q-network, and then a global coordination layer performs predictions to obtain the development intensity prediction results for each mining area zoning type. The objective function is:

[0038] ;

[0039] Where T is the decision-making period.

[0040] According to the aforementioned deep reinforcement learning-based collaborative zoning method for mineral development and ecological protection, the training of the agent includes:

[0041] Based on the mining area zoning type, the agents assigned to each mining area zoning type are trained separately using their own historical data. After completing the individual training, they are trained collaboratively through the global coordination layer using global rewards and optimized collaborative strategies.

[0042] According to the deep reinforcement learning-based method for coordinated zoning of mineral development and ecological protection, the optimization of long and short time scales includes long-term optimization and short-term optimization. The long-term optimization uses a deep recursive Q-network to extract historical time-series data to obtain historical trends, and adjusts the boundaries and total development volume of the coordinated development and ecological protection zone according to the historical trends. The short-term optimization adjusts the boundaries and total development volume of the coordinated development and ecological protection zone according to real-time monitoring data and historical time-series data.

[0043] Another aspect of the present invention provides a zoning device for coordinated mineral development and ecological protection, comprising:

[0044] The first module is used to determine the ecological sensitivity index and resource development potential index of the target area based on the ecological index and resource development index of the target area.

[0045] The second module is used to divide the area into initial zones based on the ecological sensitivity index and the resource development potential index. The initial zones include zoning constraints and mining area zoning types.

[0046] The third module is used to determine the state transition probability matrix of the target area based on the dynamic environmental characteristics and control factors of the target area using a Markov decision process.

[0047] The fourth module is used to determine the reward function and the global collaborative reward function based on the mining area zoning type, zoning constraints and the state transition probability equation of the target area. Based on the target function, reward function and global collaborative reward function, the module uses intelligent agents and deep reinforcement learning to make predictions and obtain the development intensity prediction results.

[0048] The fifth module is used to process the development intensity prediction results using long and short time scale optimization to obtain the zoning results of coordinated development and ecological protection.

[0049] Another aspect of the present invention provides an electronic device, including a processor and a memory;

[0050] The memory is used to store programs;

[0051] The processor executes the program to implement the method as described above.

[0052] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the methods described above.

[0053] The beneficial effects of this invention are as follows: by leveraging the distributed collaborative capabilities of multi-agent deep reinforcement learning and the temporal memory capabilities of DRQN ​​(Deep Recursive Q-Network), and by combining the initial partitioning of the upper layer to set differentiated agents and decision parameters for different types of mining areas, by mining the temporal patterns of historical development and pollution data, and by combining the initial partitioning results based on ecological sensitivity and resource development potential, differentiated management of different types of mining areas can be achieved, thereby realizing dynamic collaborative zoning of resource development and ecological environment in multiple mining areas. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the collaborative zoning system for mineral development and ecological protection according to an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of the process of the collaborative zoning method for mineral development and ecological protection based on deep reinforcement learning in an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of the Markov decision-making process according to an embodiment of the present invention;

[0057] Figure 4 This is a schematic diagram of the development intensity prediction process according to an embodiment of the present invention;

[0058] Figure 5 This is a schematic diagram of a collaborative zoning device for mineral development and ecological protection based on deep reinforcement learning, according to an embodiment of the present invention. Detailed Implementation

[0059] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings. Throughout the description, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" can be used interchangeably. Terms such as "first," "second," etc., are used only to distinguish technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features. In the following description, the consecutive reference numerals for method steps are for ease of review and understanding. Adjusting the implementation order of steps, in conjunction with the overall technical solution of the present invention and the logical relationship between the various steps, will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0060] refer to Figure 1 It includes a target area, a data acquisition device, and an analysis device. The target area is a mineral development area that includes different types of mines. After the data acquisition device collects data from the target area, it sends the collected data to the analysis device. After analyzing the data acquisition results, the analysis device determines the zoning results of the coordinated development of mineral resources and ecological protection in the target area.

[0061] In some embodiments, the data acquisition device is mainly used to acquire mineral resource data, ecological environment monitoring data, historical development and pollution data, and economic statistics data. Therefore, the data acquisition device can be a satellite remote sensing monitoring device, a mineral mining sensor, an environmental quality monitoring station, and a historical database interface.

[0062] In some embodiments, the analysis device may be a personal computer, tablet computer, mobile terminal, or server.

[0063] refer to Figure 2 The flowchart shown is a collaborative zoning method for mineral development and ecological protection based on deep reinforcement learning, which includes, but is not limited to, steps S100~S500:

[0064] S100 determines the ecological sensitivity index and resource development potential index of the target area based on the ecological index and resource development index of the target area.

[0065] In some embodiments, an ecological sensitivity index is determined based on an ecological index and the weight of each ecological index. for:

[0066] ;

[0067] Among them, the ecological index For vegetation coverage, Soil erosion modulus, pH value of the water body Air Quality Index. For species richness index, As an ecological vulnerability index;

[0068] The resource development potential index is determined based on the resource development index and the weight of each resource development index. for:

[0069] ;

[0070] Among them, the resource development index The remaining recoverable reserves, For ore grade, The depth of the ore deposit. To improve the convenience of transportation in the mining area.

[0071] It is understood that the parameters of the ecological index, resource development index, weight of each ecological index and weight of each resource development index in the embodiments of the present invention have been standardized to between 0 and 1.

[0072] S200 is divided into initial zones based on the ecological sensitivity index and the resource development potential index. The initial zones include zoning constraints and mining area zoning types.

[0073] In some embodiments, the target area is divided into initial zones using discriminant analysis based on the ecological sensitivity index and the resource development potential index. The initial zones include core protection zones, restricted development zones, moderate development zones, and key development zones. The zoning constraints include development intensity and protection strength.

[0074] In some embodiments, discriminant analysis is used to partition the initial partitions, and the initial partitioning rule is: when It was designated as a core protected area at that time; and It was a restricted development zone at the time; and The area designated as a key development zone is currently designated as such; the rest are designated as moderately developed zones.

[0075] S300 uses a Markov decision process to determine the state transition probability matrix of the target area based on the dynamic environmental characteristics and control factors of the target area.

[0076] In some embodiments, reference Figure 3 The flowchart shown below illustrates the Markov decision-making process, which includes, but is not limited to, steps S310 to S340:

[0077] S310, acquire the dynamic environmental characteristics of the target area, including the current status, historical time series data, surrounding mining area correlation data and mining area zoning type.

[0078] In some embodiments, the current state is represented by a state space, which uses a multidimensional tensor to characterize dynamic environmental features, including the current state (real-time development intensity, pollution value), historical time series data (development-pollution sequence of the past 12 months, processed by LSTM), surrounding mining area correlation data (such as historical pollution of upstream mining areas), and mining area type identification (core protection zone, restricted development zone, moderate development zone, key development zone).

[0079] ;

[0080] Among them, LSTM stands for Long Short-Term Memory Network, which is a special type of recurrent neural network (RNN) used to process and predict time series data.

[0081] in, This is the current state vector (including real-time development intensity, real-time pollution value, soil pH value, etc.). The historical time series vector (including the development intensity sequence, pollution index sequence, rainfall, etc. of the past 12 months) is processed through the LSTM layer of a deep recurrent Q network (i.e., DRQN ​​network). The correlation vector for surrounding mining areas (including historical pollution peaks and current development intensity of upstream / adjacent mining areas); Identify mining area types (core protection zone, restricted development zone, moderate development zone, key development zone).

[0082] S320, obtain the regulatory factors of the target area, including the development intensity adjustment amount, ecological input protection and zoning boundary correction parameters.

[0083] In some embodiments, the action space includes adjustments to mineral development intensity, ecological protection investment, and zoning boundary correction coefficients, wherein the action space is:

[0084] ;

[0085] This refers to the adjustment range of mineral development intensity (i.e., the increase or decrease in mining volume, a continuous value). The adjustment range of development intensity in the core protection zone is [0, 0], i.e., development is prohibited; the adjustment range of development intensity in the restricted development zone is [-15%, 5%]; the adjustment range of development intensity in the moderate development zone is [-10%, 10%]; and the adjustment range of development intensity in the key development zone is [-5%, 15%].

[0086] For investments in ecological protection measures (including pollution control costs and ecological restoration area), the minimum investment in ecological protection measures in core protected areas is 30% of the mining area's output value; in restricted development zones, the minimum investment is 20% of the mining area's output value; in moderately developed zones, the minimum investment is 15% of the mining area's output value; and in key development zones, the minimum investment is 10% of the mining area's output value.

[0087] The dynamic correction parameter for the zoning boundary is a quantitative indicator that dynamically changes the zoning type of the mining area based on actual development and ecological protection effects.

[0088] ;

[0089] When the ecological environment of a certain area suffers significant pollution, and the pollution level rises for six consecutive months and exceeds a preset threshold, a restricted development zone can be converted into a core protection zone, a moderately developed zone into a restricted development zone, and a key development zone into a moderately developed zone. If resources in a key development zone are exhausted (e.g., remaining recoverable reserves are less than 10% of initial reserves) and ecological restoration is difficult (e.g., the cost of historically accumulated pollution has increased for three consecutive quarters), it will be converted into a restricted development zone. If the remaining reserves in a moderately developed zone are sufficient (e.g., remaining reserves are more than 50% of initial reserves) and the pollution control level is good (e.g., pollution levels are less than 0.1 for 12 consecutive months and there is no upward trend), it will be converted into a key development zone. Simultaneously, the conversion must meet the constraint that the ecological protection red line is not encroached upon.

[0090] S330 uses dynamic environmental characteristics as the state space and control factors as the action space, and uses a Gaussian distribution to determine the state transition probability.

[0091] In some embodiments, the state transition probability is modeled using a Gaussian distribution, as shown in the following formula:

[0092] ;

[0093] The mean Obtained by training with historical data, variance This indicates the impact of environmental noise on state transitions. Let be the transition probability function. This indicates that after taking action 'a', the state transitions to a new state. The probability distribution; It follows a Gaussian distribution; This indicates that after taking action a in state s, the user transitions to a new state. The mean, Obtained through training with historical data; This is a preset value, representing the variance of the new state obtained after taking action a in state s.

[0094] It is understandable that the above actions refer to the implementation of mineral development or environmental protection measures in different mining areas.

[0095] S340 adjusts the action based on the input development intensity, determines the state transition probability matrix based on the state transition probability, and determines the state transition probability equation for the contamination value.

[0096] Adjust actions based on input development intensity Determine the state transition probability equation based on the state transition probability. for:

[0097] ;

[0098] in, Remaining recoverable reserves Given the current development intensity, For time step.

[0099] Among them, time step Take 1 for the long-term scale (annual) and 1 / 12 for the short-term scale (monthly);

[0100] The state transition probability equation for pollution values for:

[0101] ;

[0102] in The cumulative pollution value, To determine the pollution coefficient, a value of 0.15 is used. The governance efficiency coefficient is set to 0.1. For noise, Investment in ecological protection measures;

[0103] S350, determine the state transition probability equation of the target area based on the state transition probability matrix of the remaining mineral reserves and the state transition probability equation of the pollution value.

[0104] S400: Based on the mining area zoning type, zoning constraints, and the state transition probability equation of the target area, the reward function and the global collaborative reward function are determined. Based on the target function, reward function, and global collaborative reward function, an intelligent agent and deep reinforcement learning are used for prediction to obtain the development intensity prediction result. In some embodiments, refer to... Figure 4 The schematic diagram of the development intensity prediction process shown includes, but is not limited to, steps S410 to S430:

[0105] S410, determine the global reward function based on the state transition probability equation of the target region.

[0106] Wherein, the global reward function at time t for:

[0107] ;

[0108] Where n is the number of mining areas of the same type. This represents the mineral resource development revenue at time t for the i-th mining area of ​​the same type. Let be the development cost of the i-th similar mining area at time t. Let be the ecological environment restoration cost of the i-th mining area of ​​the same type at time t. The historical cumulative pollution impact cost of the i-th type of mining area at time t. For the i-th mining area of ​​the same type, the historical influence weighting coefficient is determined according to the mining area zoning type;

[0109] S420 determines the reward function for each mining area type based on the global reward function, mineral resource development revenue, ecological environment restoration costs, and historical cumulative pollution impact costs.

[0110] In some embodiments, different types of agents are set up for different types of mining areas obtained from the initial partitioning at the upper layer. The type of mining area determines the range of resource development intensity and the minimum limit of ecological environmental protection in that area, and corresponding reward functions and parameters are designed based on the connotation of different types of mining areas. Core protection areas focus on ecological protection rewards, while key development areas appropriately focus on development revenue rewards while ensuring the ecology. The specific formula is as follows:

[0111] Core protected area: ;

[0112] Restricted Development Zones: ;

[0113] Moderately developed zones: ;

[0114] Key development zones: ;

[0115] in, Let t be the revenue from mineral resource development. The current cost of ecological environment restoration (only for pollution damage caused immediately during the development process, including dust pollution control costs, noise pollution control costs, and immediate vegetation restoration costs). Historical cumulative pollution impact costs extracted for DRQN ​​(for the types of pollution that accumulate over time after development, including costs for soil heavy metal remediation, groundwater pollution control, and geological hazard prevention).

[0116] In some embodiments, the upper layer refers to the initial partitioning, which determines four functional zones—core protection zone, restricted development zone, moderate development zone, and key development zone—based on ecological sensitivity and resource development potential, and sets initial thresholds for development intensity and protection strength. The lower layer is a dynamic collaborative optimization model combining multi-agent deep reinforcement learning (MADRL) and deep recursive Q-network (DRQN). Different types of DRQN ​​agents with LSTM layers are configured for different types of mining areas partitioned by the upper layer. Multi-mining area collaboration is achieved through a global coordination layer, with the goal of maximizing the comprehensive economic and ecological benefits. The decision-making scope and network parameters of the agents are constrained by the initial partitioning results of the upper layer.

[0117] S430 assigns agents to different mining area zoning types based on the objective function, reward function, and global collaborative reward function. The agents process historical time series data using a deep recursive Q-network (DRQN network), and then make predictions through a global coordination layer to obtain the development intensity prediction results for each mining area zoning type.

[0118] In some embodiments, the objective function is:

[0119] ;

[0120] Where T represents the decision cycle.

[0121] In some embodiments, a distributed intelligent agent is employed. The distributed intelligent layer consists of different types of DRQN ​​intelligent agents configured for core protected areas, restricted development areas, moderately developed areas, and key development areas, respectively. Each type of intelligent agent uses a DRQN ​​network to process historical time-series data. The convolutional neural network (CNN) extracts current state features, and the long short-term memory neural network (LSTM) mines historical time-series correlations to output local development strategies. The network parameters and processing focuses of each type of intelligent agent are different, as detailed below:

[0122] Core protected area agent: In the network structure, ecological data input accounts for 70% of the weight, and development data input accounts for 30%. Its main function is to manage the ecological red line area, strictly prohibit development activities, and provide negative rewards for any development behavior. The primary goal is to minimize the cost of ecological restoration, and its action space is limited to... , Ten thousand yuan.

[0123] The restricted development zone agent has an ecological data input weight of 55% and a development data input weight of 45% in the network structure. Its main function is to control the development scale, prevent over-development from damaging the ecology, and adjust the development intensity cautiously. The annual mineral production limit is 5% of the regional reserves per year.

[0124] Moderately developed intelligent agents: In the network structure, the weight of ecological data input accounts for 50%, and the weight of development data input accounts for 50%, balancing development and protection, and ensuring ecological stability while rationally developing resources.

[0125] Intelligent agents in key development zones: In the network structure, ecological data input accounts for 40% of the weight and development data input accounts for 60%. Under the premise of meeting the ecological baseline requirements, optimize the development strategy to improve the returns, with efficient resource development as the core, and the recovery rate should be maintained above 85%.

[0126] Meanwhile, to ensure collaborative cooperation among multiple agents, a global coordination layer is introduced, achieving global collaborative management through a centralized evaluation mechanism. The global coordination layer adopts a multi-agent centralized evaluation-distributed execution architecture, collecting current status and historical data from all different types of mining areas, calculating global rewards and feeding them back to each agent, coordinating cross-regional decision-making conflicts, and ensuring optimal overall benefits.

[0127] The global coordination layer achieves coordination by calculating the global value correction coefficient, the formula of which is as follows: ,in This is the sensitivity parameter (value 1.2). This represents the historical global average collaborative reward. The local policies of each agent will be adjusted according to... Make corrections, and the probability of the corrected action is: , The synergistic tendency factor is set at 0.8 for core protected areas, 0.6 for restricted development areas, 0.5 for moderately developed areas, and 0.3 for key development areas.

[0128] In some embodiments, wherein Let t be the revenue from mineral development. For development costs, Given the current cost of ecological environment restoration, Cost of historical cumulative pollution impact (calculated based on time-series features extracted by DRQN). For the decision-making cycle, This represents the historical influence weighting coefficient.

[0129] Historical influence weighting coefficient Core protected area The value is 0.7, indicating a restricted development zone. The value is 0.6, indicating a moderately developed development zone. The value is 0.4, and it is a key development zone. The value is 0.3.

[0130] Current ecological environment restoration costs Calculation method: ,in Costs for dust pollution control For noise pollution control costs, Costs for immediate vegetation restoration.

[0131] Costs of historical cumulative pollution Calculation method: ,in Costs for soil heavy metal remediation Costs for groundwater pollution remediation Costs for geological disaster prevention and control.

[0132] In some embodiments, agents assigned to each mining area are trained separately based on their own historical data, according to the mining area zoning type. After the separate training is completed, the agents are trained collaboratively by the global coordination layer based on global rewards and optimized collaborative strategies.

[0133] In some embodiments, the training of the agent includes individual training and collaborative training.

[0134] Individual training is conducted, with each agent learning the time-series patterns of "development-pollution" based on its own historical data and constraints of the corresponding type of mining area (e.g., an agent in a key development area uses LSTM to capture the "correlation between development intensity and revenue"). During training, an independent training sample set is constructed for each agent, which includes monthly development intensity, pollution index, and revenue data for that type of mining area over the past 5 years.

[0135] Collaborative training involves connecting 50 different types of agents to the global coordination layer, where they interact in a shared environment. The collaborative strategy is optimized through global rewards, and the process is iterated 100,000 times, with each round simulating one calendar year, until the global reward fluctuation is less than 3%.

[0136] Based on the development intensity prediction results, S500 is processed using long-term and short-term time scale optimization to obtain the zoning results of coordinated development and ecological protection.

[0137] In some embodiments, long-term timescale optimization includes short-term timescale optimization and long-short-term timescale optimization.

[0138] On a long-term (annual) scale, the zoning boundaries and total development volume are adjusted based on historical trends extracted by DRQN ​​(such as annual pollution accumulation patterns). The adjustment frequency and magnitude vary for different types of mining areas. The core protection zone boundary has the lowest adjustment frequency and the smallest magnitude, while the key development zone boundary has a relatively higher adjustment frequency and a relatively larger magnitude.

[0139] On a short-term (monthly) scale, the mining intensity is fine-tuned by combining real-time monitoring data with recent historical data (such as the development-pollution correlation of the past 3 months). The basis and sensitivity of fine-tuning vary for different types of mining areas.

[0140] Figure 5 This is a schematic diagram of a collaborative zoning device for mineral development and ecological protection based on deep reinforcement learning, according to an embodiment of the present invention. The device includes a first module 510, a second module 520, a third module 530, a fourth module 540, and a fifth module 550.

[0141] The system comprises five modules: First, determining the ecological sensitivity index and resource development potential index of the target area based on its ecological and resource development indices; second, dividing the area into initial zones based on these indices, including zoning constraints and mining area zoning types; third, determining the state transition probability matrix of the target area using a Markov decision process based on its dynamic environmental characteristics and regulatory factors; fourth, determining the reward function and global collaborative reward function based on the mining area zoning type, zoning constraints, and the target area's state transition probability matrix, and using an agent and deep reinforcement learning to predict the development intensity; and fifth, processing the development intensity prediction results using long-term and short-term timescale optimization to obtain the zoning results for coordinated development and ecological protection.

[0142] For example, with the cooperation of the first, second, third, fourth, and fifth modules in the device, the embodiment device can implement any of the aforementioned methods for coordinated zoning of mineral development and ecological protection. Specifically, it determines the ecological sensitivity index and resource development potential index of the target area based on the ecological index and resource development index of the target area; divides the area according to the ecological sensitivity index and resource development potential index to obtain initial partitions, where the initial partitions include partition constraints and mining area zoning types; determines the state transition probability matrix of the target area using a Markov decision process based on the dynamic environmental characteristics and control factors of the target area; determines the reward function and global collaborative reward function based on the mining area zoning type, partition constraints, and the state transition probability matrix of the target area; uses an intelligent agent and deep reinforcement learning to predict the development intensity based on the objective function, reward function, and global collaborative reward function, obtaining the development intensity prediction result; and processes the development intensity prediction result using long-term and short-term time scale optimization to obtain the coordinated zoning result for development and ecological protection. The beneficial effects of this invention are as follows: by leveraging the distributed collaborative capabilities of multi-agent deep reinforcement learning and the temporal memory capabilities of DRQN, and by combining the initial partitioning at the upper layer to set differentiated agents and decision parameters for different types of mining areas, by mining the temporal patterns of historical development and pollution data, and by combining the initial partitioning results based on ecological sensitivity and resource development potential, differentiated management of different types of mining areas can be achieved, thereby realizing dynamic collaborative zoning of resource development and ecological environment in multiple mining areas.

[0143] This invention also provides an electronic device, which includes a processor and a memory;

[0144] The memory stores the program;

[0145] The processor executes the program to perform the aforementioned deep reinforcement learning-based method for collaborative zoning of mineral development and ecological protection; the electronic device has the function of carrying and running the software system for collaborative zoning of mineral development and ecological protection provided in the embodiments of the present invention, such as a personal computer, minicomputer, mainframe, workstation, network or distributed computing environment, standalone or integrated computer platform, or communicating with charged particle tools or other imaging devices, etc.

[0146] This invention also provides a computer-readable storage medium storing a program that is executed by a processor to implement the deep reinforcement learning-based collaborative zoning method for mineral development and ecological protection as described above.

[0147] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented in the embodiments of this invention. Alternative embodiments are contemplated, in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0148] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned deep reinforcement learning-based method for collaborative zoning of mineral development and ecological protection.

[0149] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, considering the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed in the embodiments of the invention, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0150] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0151] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can include, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0152] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0153] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0154] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0155] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0156] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for collaborative zoning of mineral development and ecological protection based on deep reinforcement learning, characterized in that, include: Based on the ecological index and resource development index of the target area, determine the ecological sensitivity index and resource development potential index of the target area; The initial zoning is obtained by dividing the area according to the ecological sensitivity index and the resource development potential index. The initial zoning includes zoning constraints and mining area zoning types. Based on the dynamic environmental characteristics and control factors of the target area, a Markov decision process is used to determine the state transition probability matrix of the target area. The reward function and global collaborative reward function are determined based on the mining area zoning type, zoning constraints, and state transition probability equation of the target area. Based on the target function, reward function, and global collaborative reward function, an intelligent agent and deep reinforcement learning are used to make predictions to obtain the development intensity prediction results. Based on the development intensity prediction results, the long and short time scales were optimized to obtain the zoning results of coordinated development and ecological protection. The step of determining the state transition probability matrix of the target area using a Markov decision process based on the dynamic environmental characteristics and control factors of the target area includes: Acquire the dynamic environmental characteristics of the target area, including the current status, historical time series data, surrounding mining area correlation data, and mining area zoning type; Obtain the regulatory factors for the target area, including the adjustment amount of development intensity, ecological input for protection, and parameters for correction of zoning boundaries; Using dynamic environmental characteristics as the state space and control factors as the action space, the state transition probabilities are determined using a Gaussian distribution: ； in, Let be the transition probability function. This indicates that after taking action 'a', the state transitions to a new state. The probability distribution; It follows a Gaussian distribution; This indicates that after taking action a in state s, the user transitions to a new state. The mean, Obtained through training with historical data; This is a preset value, representing the variance of the new state obtained after taking action a in state s; Adjust actions based on input development intensity The state transition probability matrix for determining the remaining mineral reserves is based on the state transition probabilities. for: ； in, Remaining recoverable reserves Given the current development intensity, For time step; And, determine the state transition probability equation for the pollution value, wherein the state transition probability equation for the pollution value is... for: ； in The cumulative pollution value, To develop pollution coefficients, The governance efficiency coefficient. It is noise, and ; The state transition probability equation of the target area is determined based on the state transition probability matrix of the remaining mineral reserves and the state transition probability equation of the pollution value. The process involves determining the reward function and global collaborative reward function based on the mining area's zoning type, zoning constraints, and the state transition probability equation of the target area. Then, based on the target function, reward function, and global collaborative reward function, an agent and deep reinforcement learning are used for prediction to obtain the development intensity prediction result, including: The global cooperative reward function at time t is determined based on the state transition probability equation of the target region. for: ； Where n is the number of mining areas of the same type. This represents the mineral resource development revenue at time t for the i-th mining area of ​​the same type. Let be the development cost of the i-th similar mining area at time t. Let be the ecological environment restoration cost of the i-th mining area of ​​the same type at time t. The historical cumulative pollution impact cost of the i-th type of mining area at time t. For the i-th mining area of ​​the same type, the historical influence weighting coefficient is determined according to the mining area zoning type; The reward function for each mining area type is determined based on the global collaborative reward function, the revenue from mineral resource development, the cost of ecological environment restoration, and the cost of historical cumulative pollution impact. Based on the objective function, reward function, and global collaborative reward function, agents are assigned to different mining area zoning types. These agents process historical time-series data using a deep recursive Q-network, and then a global coordination layer performs predictions to obtain the development intensity prediction results for each mining area zoning type. The objective function is: ； Where T is the decision-making period.

2. The method for collaborative zoning of mineral development and ecological protection based on deep reinforcement learning according to claim 1, characterized in that, The initial partitioning of the target area based on the ecological sensitivity index and resource development potential index includes: The ecological sensitivity index is determined based on the ecological indices and the weights of each ecological index. for: ； Among them, the ecological index For vegetation coverage, For soil erosion modulus, pH value of the water body Air Quality Index. For species richness index, As an ecological vulnerability index; The resource development potential index is determined based on the resource development index and the weight of each resource development index. for: ； Among them, the resource development index The remaining recoverable reserves, For ore grade, The depth of the ore deposit. To improve the convenience of transportation in the mining area.

3. The method for collaborative zoning of mineral development and ecological protection based on deep reinforcement learning according to claim 1, characterized in that, The ecological sensitivity index and resource development potential index used in this study were used to determine the initial zoning using discriminant analysis, including: Based on the resource development potential index and ecological sensitivity index, the target area is divided using discriminant analysis to obtain initial zoning. The initial zoning includes core protection zones, restricted development zones, moderate development zones, and key development zones. Zoning constraints include development intensity and protection strength.

4. The method for collaborative zoning of mineral development and ecological protection based on deep reinforcement learning according to claim 1, characterized in that, The training of the agent includes: Based on the mining area zoning type, the agents assigned to each mining area zoning type are trained separately using their own historical data. After completing the individual training, they are trained collaboratively through the global coordination layer using global rewards and optimized collaborative strategies.

5. The method for collaborative zoning of mineral development and ecological protection based on deep reinforcement learning according to claim 4, characterized in that, The optimization of long and short time scales includes long-term optimization and short-term optimization. The long-term optimization uses a deep recursive Q-network to extract historical time-series data to obtain historical trends, and adjusts the boundaries and total development volume of the development and ecological protection collaborative zone according to the historical trends. The short-term optimization adjusts the boundaries and total development volume of the development and ecological protection collaborative zone according to real-time monitoring data and historical time-series data.

6. A mineral development and ecological protection collaborative zoning device according to any one of claims 1-5, characterized in that, include: The first module is used to determine the ecological sensitivity index and resource development potential index of the target area based on the ecological index and resource development index of the target area. The second module is used to divide the area into initial zones based on the ecological sensitivity index and the resource development potential index. The initial zones include zoning constraints and mining area zoning types. The third module is used to determine the state transition probability matrix of the target area based on the dynamic environmental characteristics and control factors of the target area using a Markov decision process. The fourth module is used to determine the reward function and the global collaborative reward function based on the mining area zoning type, zoning constraints and the state transition probability equation of the target area. Based on the target function, reward function and global collaborative reward function, the module uses intelligent agents and deep reinforcement learning to make predictions and obtain the development intensity prediction results. The fifth module is used to process the development intensity prediction results using long and short time scale optimization to obtain the zoning results of coordinated development and ecological protection.

7. An electronic device, characterized in that, Including the processor and memory; The memory is used to store programs; The processor executes the program to implement the method for coordinated zoning of mineral development and ecological protection as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a program, which is executed by a processor to implement the method for coordinated zoning of mineral development and ecological protection as described in any one of claims 1-5.

Images