CN-121766734-B - Multi-scale urban garden carbon sink lifting and ecological cooperative regulation and control system

Abstract

The invention relates to a multi-scale urban garden carbon sink lifting and ecological cooperative regulation and control system, in particular to the urban garden field, which comprises the steps of establishing a computable and evolutionable knowledge graph, deeply fusing multi-source data with field rules, dynamically driving a space optimization decision, and upgrading static planning into a closed loop process of 'prediction-simulation-learning'; performing high-fidelity and uncertainty-containing long-term dynamic deduction and risk assessment on the digital twin body scheme; the simulation result is fed back to calibrate knowledge rules through meta reinforcement learning, and a successful mode back feeding knowledge base is extracted from the knowledge rules, and the process realizes collaborative self-evolution of planning knowledge, an optimization model and a simulation environment, so that the accuracy, the robustness and the foresight of decision making are continuously improved, and finally, the multi-scale collaborative optimization of carbon sink, ecology, social service and economic cost is achieved.

Inventors

• LIU SHILIANG
• XIE YUHENG
• WANG XINYU
• WU MENGXI
• CHEN QIBING
• WANG XURUI
• YANG WENLIN

Assignees

• 四川农业大学
• 四川昱泽景观规划设计有限公司

Dates

Publication Date

20260512

Application Date

20260305

Claims (8)

1. 1. The system is characterized by comprising a decision factor library construction module, an optimization scheme generation module, a scheme dynamic deduction module and a co-evolution module which are connected in sequence, wherein the decision factor library construction module is used for generating a decision factor library; the decision factor library construction module is used for responding to the triggering of planning tasks, integrating multisource heterogeneous data including soil, weather, remote sensing, community population, cost history and real-time Internet of things monitoring data, constructing a knowledge graph taking site conditions, plant species, ecological services and resource consumption as entities, dynamically expressing quantitative relations among the entities through computable association rules, and forming and outputting a structured decision factor library knowledge packet containing static attributes and dynamic association logic; The optimization scheme generating module is used for calling association rules in a decision factor library knowledge packet to dynamically assemble a multi-objective optimization function, discretizing a planning space into grid units, defining plant configuration and maintenance grades as decision variables, solving a function taking carbon sink, ecological diversity, social service and economic cost as target vectors by adopting an improved NSGA-III algorithm based on a knowledge graph pre-screening solution, and outputting a pareto optimal scheme set and a resource allocation blueprint and a multidimensional benefit pre-estimation value corresponding to the pareto optimal scheme set; The scheme dynamic deduction module is used for establishing an independent digital twin body for each pareto optimal scheme in the pareto optimal scheme set, wherein the digital twin body is a mixed simulation system integrating a plant growth model, a carbon circulation model and a people stream simulation model, and under the condition of accessing real-time Internet of things monitoring data and introducing random disturbance factors to simulate uncertain events, the mixed simulation system is operated to deduct the dynamic change and risk conditions of carbon sink accumulation, ecological indexes, people stream distribution and maintenance cost of each pareto optimal scheme in a future preset period, and a scheme dynamic deduction result comprising each deduction result is generated; Generating a scheme dynamic deduction result comprising various deduction results, which is specifically completed by the following three substeps: Step E1, data acquisition and aggregation, namely continuously acquiring and recording four types of time series data in the mixed simulation running process, wherein the first type is carbon sink accumulated time series data, the second type is time series data of one or more ecological indexes representing ecological conditions, the third type is accumulated maintenance cost time series data, the fourth type is real-time temperature and precipitation time series data serving as a simulation driving source, and meanwhile, continuously acquiring and recording people flow distribution data expressed in a space thermodynamic diagram form and used for evaluating social service effects, and recording information of all random disturbance events triggered in the simulation process to form a disturbance event log; e2, quantifying risk conditions, namely, aiming at the carbon sink objective function value, the ecological diversity objective function value, the social service objective function value and the economic cost objective function value in the multi-dimensional benefit pre-evaluation value corresponding to the current scheme, which are output from the optimization scheme generation module, executing the following quantification operation, namely, counting the probability distribution of the carbon sink accumulation quantity, the ecological index, the social service quantification index and the accumulated maintenance cost which are acquired in the sub-step E1 and correspond to the objective function value in the sub-step E1 and are calculated by people stream distribution data, wherein the value of the accumulated carbon sink accumulation quantity, the ecological index and the accumulated maintenance cost is at the end of a simulation preset period, and calculating the risk probability value that the value of the simulation preset period is lower than the value of the item by a preset percentage in the corresponding multi-dimensional benefit pre-evaluation value based on the probability distribution; e3, encapsulating the results, namely encapsulating the four types of time series data, the people stream distribution data and the disturbance event log which are acquired and recorded in the substep E1 and each risk probability value calculated in the substep E2 into a structured scheme dynamic deduction result for each pareto optimal scheme; The co-evolution module is used for comparing the scheme dynamic deduction result of each pareto optimal scheme with the corresponding multidimensional benefit pre-estimated value in the optimal scheme generation module to generate a deviation matrix, updating the confidence coefficient and the weight of the corresponding association rule in the decision factor library knowledge packet through the meta reinforcement learning framework based on the deviation matrix, abstracting the scheme characteristic with excellent performance in the scheme dynamic deduction result into a new association rule for back feeding to the knowledge graph, and realizing the closed loop co-evolution of the decision factor library knowledge packet, the multi-objective optimization function and the improved NSGA-III algorithm in the optimal scheme generation module, the digital twin in the scheme dynamic deduction module and the mixed simulation system; The specific process for comparing the scheme dynamic deduction result of each pareto optimal scheme with the corresponding multidimensional benefit predicted value in the optimal scheme generation module to generate a deviation matrix is as follows: Extracting the carbon sink accumulation amount obtained by final statistics of a simulation preset period, the final value of the time series data of the second type ecological index recorded in the substep E1, the social service quantization index obtained by calculation of people stream distribution data and the actual result value of accumulated maintenance cost according to the scheme dynamic deduction result of each pareto optimal scheme output by the scheme dynamic deduction module; meanwhile, extracting a carbon sink objective function value, an ecological diversity objective function value, a social service objective function value and an economic cost objective function value from the multidimensional benefit predicted value corresponding to the scheme output by the optimizing scheme generating module as corresponding expected values; calculating the relative deviation between the actual result value and the expected value in each benefit dimension of the carbon sink, the ecological, the social service and the economic cost one by one, wherein for the carbon sink, the ecological and the social service indexes, the relative deviation is obtained by subtracting the expected value from the actual result value and dividing the expected value by the expected value, and for the economic cost, the relative deviation is obtained by subtracting the actual result value from the expected value and dividing the expected value by the expected value; Then, performing bias tracing operation, namely determining the association rule of each of the decision factor library knowledge packages participating in forming the expected value of the benefit dimension of each of the pareto optimal schemes according to the calculation process of the carbon sink objective function, the ecological diversity objective function, the social service objective function and the economic cost objective function in the optimization scheme generation module, and multiplying the relative bias value by the ratio of the contribution of the association rule to the expected value of the benefit dimension to obtain a bias signal of the association rule generated by the scheme in the benefit dimension; organizing all association rules and the comprehensive deviation values of the association rules in each benefit dimension into a two-dimensional table, namely forming a deviation matrix; updating the confidence coefficient and the weight of the corresponding association rule in the decision factor library knowledge packet through the element reinforcement learning framework based on the deviation matrix, and abstracting the scheme characteristic with excellent performance in the scheme dynamic deduction result into a new association rule for feeding back to the knowledge graph, wherein the specific process comprises the following steps: Firstly, constructing a loss function of each association rule according to the comprehensive deviation amount of the rule in each dimension of carbon sink, ecological diversity, social service and economic cost based on a deviation matrix, wherein the value of the loss function is the result of summing the product of the preset preference weight of each dimension and the square of the corresponding comprehensive deviation amount; Secondly, introducing a meta-strategy network, wherein the network takes the historical performance data feature vector of each association rule as input, and dynamically outputs an adaptive learning rate for parameter updating aiming at the rule, and the adaptive learning rate comprises a first learning rate for updating confidence coefficient parameters and a second learning rate for updating weight parameters; Then, performing a parameter updating operation, wherein the operation comprises the following steps of calculating a partial derivative of a loss function on an association rule confidence coefficient parameter as an updating gradient of the confidence coefficient parameter, calculating a partial derivative of the loss function on an association rule weight parameter as an updating gradient of the weight parameter, subtracting a product of a first learning rate and the confidence coefficient parameter updating gradient from a current confidence coefficient parameter of the association rule to obtain a confidence coefficient parameter intermediate value, subtracting a product of a second learning rate and the weight parameter updating gradient from a current weight parameter of the association rule to obtain a weight parameter intermediate value, and limiting the confidence coefficient parameter intermediate value and the weight parameter intermediate value to be within an effective numerical range from zero to one respectively so as to obtain an updated confidence coefficient parameter and a weight parameter; At the same time, a feature abstraction and rule-back operation of the excellent scheme is performed, which comprises four sub-steps performed in sequence: Step F1, screening excellent schemes, namely setting a carbon sink, ecological diversity, a achievement rate threshold of social service indexes and a control rate threshold of economic cost indexes according to various risk probability values recorded in a scheme dynamic deduction result and actual result values obtained by statistics at the end of a simulation preset period, screening all pareto optimal schemes meeting threshold conditions, and forming a scheme set with excellent performance; F2, mining a characteristic mode, namely performing data mining on the resource allocation blueprints of all schemes in the screened excellent scheme set, and identifying repeated plant species combinations which occur together, soil and climate condition ranges corresponding to the combinations, and stable ecological service potential, social service effect and resource consumption demand characteristics of the combinations in simulation; A substep F3, generating and formalizing a new rule, namely formalizing the stable mode identified in the substep F2 into a new computable association rule, defining the condition part of the rule based on the identified soil and climate condition range and plant species combination, outputting the combination of the ecological service potential estimation value, the social service potential estimation value and the resource consumption demand estimation value, and endowing the new rule with an initial confidence coefficient parameter and an initial weight parameter; and F4, the knowledge base is fed back, namely all new computable association rules generated in the step F3 are added into a computable association rule list of the decision-making factor base knowledge packet, so that the expansion of the knowledge map is completed.
2. 2. The multi-scale urban garden carbon sink lifting and ecological cooperative regulation and control system according to claim 1, wherein the decision factor library construction module integrates the specific operations of multi-source heterogeneous data: The method comprises the steps of responding to a planning task trigger, acquiring and locking a geographic boundary of a target planning area, acquiring and integrating data of the pH value, the organic matter content and the volume weight of soil, historical precipitation, temperature sequence, sunshine hours, real-time precipitation and temperature monitoring data of weather, remote sensing vegetation indexes, surface temperature and land utilization classification data, community population density distribution, age structure and movement signaling thermodynamic diagram data, cost seedling purchase unit price, transportation cost and historical maintenance operation record data, and real-time monitoring data from an Internet of things sensor, wherein the real-time monitoring data comprises real-time precipitation and temperature data provided by a weather sensor, real-time soil volume water content, leaf surface humidity and stem flow rate data provided by the soil and plant sensor, and performing space-time alignment and normalization processing on all acquired multi-source heterogeneous data to form a standardized data set.
3. 3. The multi-scale urban garden carbon sink lifting and ecological cooperative control system according to claim 2, wherein the specific operation of constructing the quantitative relation of the entity which takes the site condition, plant species, ecological service and resource consumption as the entity and dynamically expresses the entity through the computable association rule is as follows: Defining a site condition entity, a plant species entity, an ecological service entity and a resource consumption entity in a knowledge graph based on the formed standardized data set, endowing each site condition entity with an attribute vector comprising the pH value, the organic matter content, the volume weight, the precipitation amount, the temperature, the sunshine hours, the vegetation index, the surface temperature, the soil volume moisture content, the leaf surface humidity and the trunk stem flow rate of the corresponding space grid, and endowing each plant species entity with an attribute vector comprising the photosynthesis rate, the transpiration coefficient and the negative resistance of the corresponding species; The method comprises the steps of establishing a relation suitable for, providing and consuming entities, converting domain knowledge and a data rule into a calculable production rule, forming a calculable association rule list containing all rules, wherein each calculable production rule is an independent data structure, the data structure comprises a condition part, an output part and a confidence part, wherein the condition part is a logic for judging based on attribute matching of a standing condition entity and a plant species entity, the output part is a quantitative estimated value for a certain ecological service supply amount or resource consumption demand amount, the quantitative estimated value is called an ecological service potential estimated value or resource consumption demand estimated value, and is associated with a specific algorithm for calculating the value, the confidence part records a confidence coefficient and a weight coefficient of the rule, the calculation process of the ecological service potential estimated value or the resource consumption demand estimated value is that the membership degree of the optimum range of plant photosynthesis is determined firstly, the rainfall and the soil moisture content is determined respectively, a plant membership degree is determined according to plant membership characteristics, a conversion factor of a leaf area and biomass allocation coefficient is finally determined according to plant membership degree, the conversion factor is calculated, the conversion of the plant membership degree is calculated based on the plant membership degree and the water content volume ratio is calculated, and the water content is multiplied by the calculation rule, and the relation of the plant membership factor is calculated, and the relation of the plant membership factor is calculated.
4. 4. The system for improving and controlling ecological cooperation of multi-scale urban garden carbon sink according to claim 3, wherein the concrete process for forming the decision factor library comprising static attribute and dynamic association logic is as follows: The defined site condition entity, plant species entity, ecological service entity and resource consumption entity are packaged into a structured decision factor library knowledge package, wherein attribute vectors are endowed for the entities, and the relationship between the entities is suitable for providing and consuming and the generated computable association rule set is established; The decision factor library knowledge packet provides an application programming interface, and the application programming interface supports two calling operations, namely, a first calling operation is to query and return corresponding estimated values of various ecological service potentials and estimated values of resource consumption requirements according to input identification information of the ground condition entities and identification information of plant species entities, and attribute vectors of the associated ground condition entities and plant species entities, and a second calling operation is to return a list of computable association rules, wherein each rule in the list exists in a data structure form and comprises condition judgment logic, a specific algorithm for calculating the estimated values of the ecological service potentials or the estimated values of the resource consumption requirements, confidence parameters and weight parameters.
5. 5. The system for improving and ecologically coordinating and regulating the carbon sink of the multi-scale urban garden according to claim 4, wherein the specific process of calling the association rules in the decision-making factor library knowledge package to dynamically assemble the multi-objective optimization function in the optimization scheme generating module is as follows: the method comprises the steps of calling an application programming interface provided by a decision factor library knowledge package, acquiring various ecological service potential estimated values and resource consumption demand estimated values packaged by the decision factor library knowledge package through a first calling operation, acquiring a complete computable association rule list through a second calling operation, discretizing a planning space into grid units with equal areas to form a grid unit set, defining decision variables of each grid unit in the set as a triplet comprising plant species numbers, planting densities and maintenance intensity levels, and dynamically constructing four objective functions, namely a carbon sink objective function, an ecological diversity objective function, a social service objective function and an economic cost objective function based on the decision variables, the acquired ecological service potential estimated values and the resource consumption demand estimated values; Screening all rules which take carbon as output and are applicable to plant species numbers selected by the standing conditions and decision variables of the grid cells from a computable association rule list to form a rule screening result set, calculating the product of a carbon sink potential estimated value of each rule in the rule screening result set and confidence coefficient parameters and weight parameters of the rule, and summing the product results of all rules in the rule screening result set to obtain a carbon sink contribution weighted sum aiming at the grid cells; Then multiplying the area of the grid unit, the planting density in the decision variable and the calculated carbon sink contribution weighted sum to obtain the contribution of the grid unit to the total carbon sink target; Finally, accumulating and summing the contribution of all grid cells in the grid cell set, wherein the result is the carbon sink objective function value; The method comprises the steps of selecting all grid cells in a grid cell set from a computable association rule list, filtering out rules which take relevant indexes of ecological diversity as output and are suitable for plant species numbers selected by standing conditions and decision variables of the grid cells to form an ecological diversity rule screening result set, calculating the product of an ecological diversity potential estimation value corresponding to each rule in the ecological diversity rule screening result set and confidence coefficient parameters and weight parameters of the rule, and summing up the product results of all rules in the result set to obtain an ecological diversity contribution weighted sum of the grid cells, wherein the ecological diversity potential estimation value is calculated by the corresponding association rules in a decision factor library, and multiplying the area of the grid cells, the planting density in decision variables of the grid cells and the calculated ecological diversity contribution weighted sum to obtain the contribution of the grid cells to the total ecological diversity target; The social service objective function is calculated by screening all rules which take the social service related index as output and the condition part of the rules is applicable to the plant species number selected by the grid unit from a computational association rule list to form a social service rule screening result set, calculating the product of a social service potential estimation value corresponding to each rule in the social service rule screening result set and the confidence coefficient parameter and weight parameter of the rule, summing the product results of all rules in the result set to obtain the social service contribution weighted sum of the grid unit, calculating the social service contribution weighted sum of the grid unit by the corresponding association rules in a decision factor library, multiplying the area of the grid unit, the planting density in the decision variable of the grid unit and the calculated social service contribution weighted sum by a service demand weight coefficient determined based on population distribution data to obtain the contribution of the grid unit to the total social service objective, and finally summing the contribution of all the grid units to obtain the social service objective function value, wherein the contribution of the grid unit is the service objective function value; The economic cost objective function is calculated by screening all rules which take resource consumption requirements as output and are suitable for plant species numbers selected by a decision variable from a computable association rule list to form a resource consumption rule screening result set, calculating the product of a resource consumption requirement estimated value corresponding to each rule in the resource consumption rule screening result set and a confidence coefficient parameter and a weight parameter of the rule, summing the product results of all rules in the result set to obtain the comprehensive resource consumption weighted requirement of the grid unit, multiplying the area of the grid unit, the planting density in a decision variable thereof, the coefficient converted by a maintenance intensity level and the calculated comprehensive resource consumption weighted requirement to obtain the resource consumption cost of the grid unit, and finally, accumulating and summing the resource consumption cost of all grid units and combining the initial cost of nursery stock and transportation to obtain the economic cost objective function value; The dynamically constructed carbon sink objective function, the ecological diversity objective function, the social service objective function and the economic cost objective function jointly form a multi-objective optimization function.
6. 6. The system for improving and ecologically coordinating and regulating carbon sink in multi-scale urban gardens according to claim 5, wherein the specific process for solving the function taking carbon sink, ecological diversity, social service and economic cost as target vectors by adopting an improved NSGA-III algorithm based on knowledge-graph pre-screening solution is as follows: Extracting mandatory constraint conditions about plant stereoscopic tolerance from a computable association rule list, wherein the mandatory constraint conditions comprise soil pH value and salt content required by plant species to be within tolerance range, and taking the mandatory constraint conditions as hard constraint of optimization problem; And after the algorithm is terminated, outputting a group of pareto optimal scheme set, wherein each pareto optimal scheme in the set comprises two parts of contents, one part is a plant species number, planting density and maintenance intensity level determined for each grid unit, and is called a resource allocation blueprint of the scheme, and the other part is a carbon sink objective function value, an ecological diversity objective function value, a social service objective function value and an economic cost objective function value corresponding to the scheme, which are commonly called as a multidimensional benefit pre-estimation value of the scheme.
7. 7. The system for improving and ecologically collaborative regulation and control of multi-scale urban garden carbon sink of claim 6, wherein the scheme dynamic deduction module establishes independent digital twin bodies for each pareto optimal scheme in the pareto optimal scheme set by the following specific process: extracting a resource allocation blueprint corresponding to each pareto optimal scheme in the pareto optimal scheme set output by the optimal scheme generating module; based on the resource allocation blueprint, an independent digital twin body is instantiated for the pareto optimal scheme, wherein the digital twin body comprises a hybrid simulation system which is composed of three mutually coupled models, namely a plant growth model, a carbon circulation model and a people stream simulation model; The initial state of the plant growth model is set according to plant species numbers and planting densities of each grid unit in a resource allocation blueprint, the plant growth model is a hybrid model combining mechanism and data driving, the mechanism part of the plant growth model is based on photosynthesis, respiration and biomass allocation theory, and the data driving part of the plant growth model acquires photosynthesis rate, transpiration coefficient physiological and ecological parameters of configured plant species and association rules affecting plant growth by calling an application programming interface of a decision factor library knowledge packet; the initial parameter part of the carbon circulation model is derived from an attribute vector of a site condition entity of a corresponding grid in a decision factor library knowledge packet, and the carbon circulation model is connected with the output of the plant growth model so as to receive biomass change data generated by plant growth simulation; And the artificial abortion model constructs a basic environment according to plant space configuration described by the resource allocation blueprint and the geographic information system data.
8. 8. The system for improving and ecologically coordinating and regulating carbon sink in multi-scale urban gardens of claim 7, wherein the method for generating the scheme dynamic deduction result comprising various deduction results comprises the following specific processes: In the simulation deduction process, continuously accessing real-time monitoring data from an Internet of things sensor for each digital twin body, and inputting the real-time monitoring data into each model of the hybrid simulation system in real time as an environment driving variable; At the same time, a random disturbance factor is systematically introduced on a simulation time axis, wherein the random disturbance factor is a random event sequence generator for simulating extreme drought, flood and plant disease and insect outbreak uncertainty events, an event sequence is generated according to a preset time-space distribution rule, and a disturbance influence calculation process is introduced in a plant growth model, wherein the calculation process comprises the steps of firstly calculating theoretical total primary productivity based on model states, then calculating an environmental stress function value jointly determined by real-time temperature, precipitation and soil volume water content; Accumulating the potential productivity loss proportion caused by all triggered random disturbance events, and acting on the theoretical total primary productivity together with the environmental stress function value, so as to calculate the net primary productivity after disturbance; the hybrid simulation system is used for coupling and running the plant growth model, the carbon circulation model and the people stream simulation model in a time stepping mode in a preset future time period, the process is called hybrid simulation running, and the social service effect and the time-dependent change curve of accumulated carbon sink accumulation, ecological indexes and people stream distribution of each scheme are dynamically simulated.

Description

Multi-scale urban garden carbon sink lifting and ecological cooperative regulation and control system Technical Field The invention relates to the field of urban gardens, in particular to a multi-scale urban garden carbon sink lifting and ecological cooperative regulation and control system. Background In the current urban updating process taking green low carbon as a guide, integrating green space to cooperatively promote carbon sink capacity, ecological service function and social welfare becomes a core issue, and in the old urban reconstruction project, a composite community park is often constructed by utilizing limited plots, and planning and construction face multiple and interrelated targets, namely, not only is a considerable carbon sink needed to be quickly formed to offset carbon emission generated by regional development, so as to meet the requirement of advanced carbon assessment, but also urgent requirements of residents in a high-density living area on leisure space are met, the ecological function of constructing local biodiversity gallery nodes is borne, and the low-cost sustainable maintenance of the whole life cycle is realized under the external constraints of annual financial budget, nursery stock market fluctuation and the like, and the multiple targets are jointly acted on the same limited land, fund, nursery stock and water resource, so that the problem of how to accurately allocate various resource units to different functional areas in the initial planning and subsequent management becomes a challenge related to environment, society and economic complex trade-off is solved. However, the existing landscaping management and planning technical system is difficult to effectively support the complex decisions of multi-objective collaborative optimization, the mainstream landscaping information management system is mainly used for realizing visualization and simple query of space information and lack of embedded intelligent decision support modules even if combined with a geographic information system, although relatively independent carbon sink metering models or ecosystem service assessment tools are developed in academia, the technical models are usually separated from actual resource planning, distribution and scheduling service flows, the fundamental technical problem is that the existing system cannot quantitatively fuse heterogeneous indexes such as carbon sink increment, biological diversity index, human average leisure area, full life cycle cost and the like and solve the heterogeneous indexes optimally under a unified frame, so that the special decisions such as 'how many plants are combined and distributed in a specific space can realize comprehensive benefit optimization' cannot provide scheme simulation and comparison based on data and models, and simultaneously, cross-domain data such as high-precision carbon sink prediction data, population dynamic data, nursery stock and adaptive dispersion data and the like which are needed by support decisions are usually separated from each other, the scheme cannot be easily analyzed by the dynamic feedback from the prior art, and the dynamic state is difficult to realize the real-time, and the dynamic state data is difficult to be adjusted to realize the real-time, and the real-time feedback of the dynamic state information is difficult to realize the real-time, and the real-time performance is difficult to realize the real-time analysis of the dynamic state, and the dynamic state is difficult to realize the real-time, and the quality control is based on the dynamic information, and the dynamic state has a feedback has been difficult to be distributed, and the quality, A series of risks such as insufficient ecological benefit or long-term maintenance cost. Disclosure of Invention Aiming at the technical problems in the prior art, the invention provides a multi-scale urban garden carbon sink lifting and ecological cooperative regulation system, which solves the problems in the background art through a decision factor library construction module, an optimization scheme generation module, a scheme dynamic deduction module and a cooperative evolution module. The technical scheme for solving the technical problems comprises a decision factor library construction module, an optimization scheme generation module, a scheme dynamic deduction module and a co-evolution module which are connected in sequence, wherein the decision factor library construction module is used for generating a decision factor library; the decision factor library construction module is used for responding to the triggering of planning tasks, integrating multisource heterogeneous data including soil, weather, remote sensing, community population, cost history and real-time Internet of things monitoring data, constructing a knowledge graph taking site conditions, plant species, ecological services and resource consumption as entities, dynamically expressing quantitative r