Search

CN-122022190-A - Landscape garden ecological planning decision-making method based on industrial Internet knowledge graph

CN122022190ACN 122022190 ACN122022190 ACN 122022190ACN-122022190-A

Abstract

The invention provides a landscape garden ecological planning decision method based on an industrial Internet knowledge graph, which belongs to the cross technical field of landscape garden forestation, computer science and technology and industrial Internet of things engineering, and comprises the following steps of S1, acquiring standardized ecological-industrial fusion perception data; S2, generating a landscape garden ecological planning knowledge map base, S3, generating a dynamic association rule set of ecological factors-industrial perception data, S4, generating a preliminary ecological planning deduction scheme, and S5, generating a multi-dimensional ecological planning decision scheme set. The invention realizes the real-time data, the association quantification, the deduction and the dynamics of the landscape architecture ecological planning decision, the decision sceneries, the comprehensive verification and the iteration implementation, improves the suitability and the scientificity of the planning scheme and the ecological characteristics of the planning area and the industrial perception network characteristics, and provides a brand-new technical path and implementation method for the landscape architecture ecological planning.

Inventors

  • MENG YU

Assignees

  • 绵阳师范学院

Dates

Publication Date
20260512
Application Date
20260408

Claims (10)

  1. 1. The landscape garden ecological planning decision-making method based on the industrial Internet knowledge graph is characterized by comprising the following steps of: s1, constructing an heterogeneous adaptation system of an industrial Internet sensing network and landscape garden ecological monitoring nodes, and processing original data acquired in a planning area to obtain standardized ecological-industrial fusion sensing data; s2, integrating standardized ecological-industrial fusion perception data, landscape basic geographic data, landscape ecological planning industry specification data and industrial Internet equipment operation and maintenance data, constructing four fusion body layers of an industrial Internet knowledge graph, establishing a heterogeneous association mapping rule, and generating a landscape ecological planning knowledge graph basic library; S3, extracting core data from a landscape-garden ecological planning knowledge graph base library, performing time sequence alignment pretreatment, fusing an HIN2Vec heterogeneous network embedding algorithm and a TAMP time-aware multi-view time sequence association rule mining algorithm to realize bidirectional deep interaction, and performing mapping time sequence association mining on the pretreated core data to obtain a dynamic association rule set of ecological factor-industrial perception data; s4, building an industrial Internet edge cloud collaborative deduction framework, fusing GRAPHWAVE graph wave analysis algorithm and FedDyn dynamic federal learning algorithm to realize bidirectional deep interaction, taking standardized ecological-industrial fusion perception data as real-time input, a landscape garden ecological planning knowledge map base library as a carrier and a dynamic association rule set of ecological factor-industrial perception data as a deduction basis, and outputting a preliminary ecological planning deduction scheme; S5, dividing four types of core planning scenes, taking a dynamic association rule set of ecological factors-industrial perception data as core driving constraint, taking a preliminary ecological planning deduction scheme as an initial training sample, constructing and training a multi-scene landscape ecological planning decision model driven by a map, and generating a multi-dimensional ecological planning decision scheme set.
  2. 2. The landscape architectural ecological planning decision method based on the industrial internet knowledge graph according to claim 1, further comprising: s6, building an ecological-industrial two-dimensional verification system, and carrying out comprehensive verification and targeted correction on ecological dimensions and industrial dimensions on a multi-dimensional ecological planning decision scheme set by taking a landscape ecological planning knowledge base and a dynamic association rule set of ecological factors-industrial perception data as core basis to obtain a landscape ecological planning decision scheme which is qualified in verification; S7, implementing the landscape-garden ecological planning decision scheme which is qualified in verification in a planning area, continuously collecting real-time operation data of the planning area through an industrial Internet perception network and feeding back the real-time operation data to a landscape-garden ecological planning knowledge spectrum base, establishing a landscape-garden ecological planning knowledge spectrum dynamic iteration mechanism and decision scheme continuous optimization system, when a preset iteration trigger threshold is reached, re-executing the steps S3 to S6 by taking the landscape-garden ecological planning knowledge spectrum base after iteration update as a core, outputting the optimized landscape-garden ecological planning decision scheme, and progressively replacing and implementing the landscape-garden ecological planning decision scheme in the planning area to form a landscape-garden ecological planning closed-loop decision flow of perception-adaptation-modeling-excavation-deduction-decision-verification-iteration.
  3. 3. The landscape architecture ecological planning decision method based on the industrial Internet knowledge graph according to claim 1, wherein the HIN2Vec heterogeneous network embedding algorithm is characterized in that a heterogeneous entity-relation-entity triplet in a landscape architecture ecological planning knowledge graph base is used as input, a relation weight matrix, an entity initial embedding vector and a bias term are initialized, ecological factors in the knowledge graph and industrial perception equipment heterogeneous entities are converted into low-dimensional computable embedding vectors through model operation, model training is completed by taking minimum heterogeneous network embedding errors as targets, core data after time sequence alignment preprocessing is mapped into embedded time sequence data, cosine similarity of the ecological factors and the industrial perception equipment entity embedding vectors is calculated, and embedded time sequence data input and multi-view association weight are provided for TAMP time perception multi-view time sequence association rule mining algorithm.
  4. 4. The landscape architecture ecological planning decision method based on the industrial Internet knowledge graph is characterized in that the TAMP time-aware multi-view time sequence association rule mining algorithm takes embedded time sequence data output by an HIN2Vec heterogeneous network embedded algorithm as core input, combines multi-view association weights provided by the TAMP time-aware multi-view time sequence association rule mining algorithm, calculates time sequence association degrees of ecological factors and industrial perception data at different moments through setting time-aware sliding windows, quantitatively screens up standard association results through confidence and lifting degrees, finishes model training with minimum rule screening precision errors as targets, regularly refines the screened effective association results, and finally generates a dynamic association rule set of ecological factor-industrial perception data by defining preconditions, association objects, quantization thresholds and result characterization of each rule.
  5. 5. The landscape architecture ecological planning decision method based on the industrial Internet knowledge graph, which is disclosed in claim 4, is characterized in that the bidirectional deep interaction process of the HIN2Vec heterogeneous network embedding algorithm and the TAMP time-aware multi-view time sequence association rule mining algorithm is that the forward interaction provides embedded time sequence data and multi-view association weights for the HIN2Vec heterogeneous network embedding algorithm and the TAMP time-aware multi-view time sequence association rule mining algorithm, the algorithm is supported to complete time sequence association calculation and rule screening, the reverse interaction is TAMP time-aware multi-view time sequence association rule mining algorithm optimizes the multi-view association weights with the minimum rule screening precision error as a target, the relation weight matrix of the HIN2Vec heterogeneous network embedding algorithm is updated through gradient feedback, the representation capability of the algorithm on ecological-industrial association relation is improved, and bidirectional optimization and feature mutual feedback of the two types of algorithms are realized.
  6. 6. The landscape architecture ecological planning decision method based on the industrial Internet knowledge graph, as set forth in claim 1, is characterized in that the GRAPHWAVE graph wave analysis algorithm works by taking the topological structure of a landscape architecture ecological planning knowledge graph base as input, firstly calculating a degree matrix and an adjacent matrix of the graph and obtaining a Laplacian matrix, obtaining characteristic values and characteristic vectors of the Laplacian matrix through characteristic value decomposition, extracting graph wave characteristic vectors of all nodes in the graph based on the characteristic values and the characteristic vectors, generating node global fusion characteristics by fusing association degree values in dynamic association rule sets of ecological factors-industrial perception data, completing model training with minimized graph wave characteristic reconstruction errors, and providing global fusion characteristic constraint and dynamic adjustment coefficients for FedDyn dynamic federal learning algorithm.
  7. 7. The landscape architecture ecological planning decision-making method based on the industrial Internet knowledge graph as claimed in claim 6, wherein the FedDyn dynamic federal learning algorithm has the working process that the global fusion characteristic and the dynamic association rule set of ecological factor-industrial perception data output by the standardized ecological-industrial fusion perception data and the GRAPHWAVE graph wave analysis algorithm are taken as input, the local quick local deduction of a planning area is completed at the side end of the industrial Internet side cloud collaborative deduction architecture, all side end deduction results are integrated in the cloud and global fusion characteristic is combined to complete global optimization deduction, model training is completed by taking the joint loss of the minimum side cloud deduction deviation and GRAPHWAVE graph wave analysis algorithm training loss as a target, model parameters are updated through side cloud interaction, the deviation rate of the side cloud deduction result reaches a preset standard, and finally the structured preliminary ecological planning deduction scheme is output.
  8. 8. The landscape architecture ecological planning decision method based on the industrial Internet knowledge graph according to claim 7, wherein the bidirectional deep interaction process of the GRAPHWAVE graph wave analysis algorithm and the FedDyn dynamic federal learning algorithm is that forward interaction is that the GRAPHWAVE graph wave analysis algorithm provides global fusion characteristics and dynamic adjustment coefficients for the FedDyn dynamic federal learning algorithm, the global fusion characteristics serve as global constraints of edge side deduction to avoid local deduction from a global ecological rule, the dynamic adjustment coefficients support cloud to complete accurate global optimization deduction, reverse interaction is that the FedDyn dynamic federal learning algorithm takes minimum joint loss as a target optimization model parameter, the joint loss is updated through gradient feedback to the Laplace matrix characteristic value of the GRAPHWAVE graph wave analysis algorithm, the suitability of the global characteristics extracted by the algorithm to the cloud collaborative deduction is improved, and the global characteristic constraints and reverse loss optimization of the two algorithms are achieved.
  9. 9. The landscape garden ecological planning decision method based on the industrial internet knowledge graph according to claim 1, wherein the training logic of the graph-driven multi-scene landscape garden ecological planning decision model is characterized in that a dynamic association rule set of ecological factors-industrial perception data is used as a model hard constraint, the entity relation of a landscape garden ecological planning knowledge graph base is used as a model state space, the decision behavior of landscape garden ecological planning is used as a model action space, a dedicated reward function is designed for four types of core planning scenes, a preliminary ecological planning deduction scheme is used as an initial training sample, the model weight is dynamically adjusted by taking the latest knowledge data of the landscape garden ecological planning knowledge graph base in real time in the training process, training is stopped when the convergence errors of the reward functions in the four types of core planning scenes are less than or equal to 3%, and the model outputs a personalized planning scheme corresponding to the four types of core planning scenes and is integrated into a multi-dimensional ecological planning decision scheme set.
  10. 10. The landscape architecture ecological planning decision-making method based on the industrial internet knowledge graph according to claim 2, wherein the execution logic of the dynamic iteration mechanism of the landscape architecture ecological planning knowledge graph in S7 is that the real-time operation data comprises ecological operation real-time data, garden facility operation state data and industrial perception equipment operation and maintenance data, the mechanism sequentially executes entity completion, relation update and attribute correction operations on a landscape architecture ecological planning knowledge graph base according to the real-time operation data, the entity completion is the ecological, garden facility, industrial perception equipment type entities and corresponding attributes and association relations newly appeared in the implementation of the supplementation planning, the relation update is to adjust the association relation and quantization threshold between original entities, the attribute correction is to correct various attribute data of the original entities, all update operations keep log records, and the preset iteration trigger threshold is that the entity update amount of the landscape architecture ecological planning knowledge graph base is more than or equal to 10% or the relation correction amount is more than or equal to 15%.

Description

Landscape garden ecological planning decision-making method based on industrial Internet knowledge graph Technical Field The invention relates to the cross technical fields of landscape architecture, computer science and technology and industrial Internet of things engineering, in particular to a landscape architecture ecological planning decision-making method based on an industrial Internet knowledge graph. Background The landscape architecture ecological planning is a core technical means for realizing urban ecological protection, landscape construction and human living environment optimization, the traditional planning method in the current industry mainly relies on manual site investigation, static geographic information data and historical ecological monitoring data to develop planning design, and simple data analysis software is used for assisting in carrying out qualitative analysis of ecological factors, while the method has certain feasibility in basic planning, the technical limitations of the traditional method are increasingly prominent along with dynamic changes of urban ecological environment and popularization of industrial Internet technology in various fields, and the requirements of precision, dynamics and scientificalness of modern landscape architecture ecological planning cannot be met, and the specific technical pain points are represented in the following seven aspects: the traditional planning mainly relies on static geographic data and historical ecological data which are collected at one time, a global and real-time perception system is not built, dynamic changes of ecological factors and garden facilities in a planning area cannot be captured, and deviation exists between a planning scheme and an actual ecological state; the multi-source heterogeneous data cannot be fused and has no unified knowledge carrier, namely, the multi-source data such as ecological monitoring data, geographic data, industry standard data, facility operation and maintenance data and the like involved in the planning process have the isomerism of communication protocols, formats and dimensions, the traditional method lacks an effective fusion modeling means, the data is distributed in an 'island' manner, has no structured and associated knowledge carrier, and is difficult to realize the deep mining and associated analysis of the data; The ecological association analysis is qualitative, and no quantization rule basis is that the association analysis between ecological factors by the traditional planning is mostly artificial qualitative judgment, so that the association mining of the quantization time sequence of the ecological factors and planning support data can not be realized, and the lack of a landable quantization association rule leads to the lack of scientific data support for the planning scheme; the planned deduction has no real-time property, the local deduction and the global deduction are disjointed, the traditional scheme is designed in a static scheme, the real-time deduction capability is not realized, the partial progressive method introduces a simple deduction model, but only can realize the static deduction of a local area, the cooperative deduction of the local area and the global area of the planned area can not be realized, and the problems of 'local optimum and global suboptimal' are easy to occur; The traditional planning mostly adopts a 'one-cut' decision mode, and personalized schemes are not designed according to the actual requirements of different areas such as an ecological sensitive area, a landscape display area, a facility concentration area and the like of a planning area, so that the suitability and the practicability of the planning scheme are insufficient; verifying dimension uniqueness, namely performing compliance verification of ecological dimension only in traditional planning, and not considering suitability of a planning scheme, industrial sensing equipment and a sensing network after the industrial sensing network is introduced, so that if the sensing system is built subsequently, the planning scheme needs to be modified greatly, and the implementation cost is increased; the implementation process is not iterated, one-time decision static implementation is implemented, the traditional planning is a linear flow of 'design-implementation-ending', no effective dynamic iteration and optimization mechanism exists after the scheme is landed, and when the ecological characteristics and the facility state of a planning area change, the original scheme cannot be timely adjusted, so that the planning effect gradually deviates from expectations. Disclosure of Invention The invention provides a landscape garden ecological planning decision-making method based on an industrial Internet knowledge graph, which integrates heterogeneous perception, bian Yun synergy and dynamic iteration characteristics of an industrial Internet with the structuring association of the knowledge gra