Urban resource optimal configuration method and system based on multi-target cooperation

Abstract

The invention relates to the technical field of resource optimal configuration, in particular to an urban resource optimal configuration method and system based on multi-objective collaboration. The method comprises the following steps: preprocessing acquired urban resource related data; constructing a dynamic space-time knowledge graph based on the preprocessed urban resource related data; constructing a reinforcement learning-neural network hybrid optimization model, and taking the dynamic space-time knowledge graph as embedded representation of the model; performing multi-target training optimization on the model by adopting quantum calculation and a classical optimization algorithm; and generating a resource configuration decision based on the optimized model. The urban resource information can be more comprehensively and accurately acquired by constructing a multi-source data acquisition network, integrating space-time, structuralization, text, image and other multi-modal heterogeneous data, and performing feature extraction and fusion by using a space-time diagram convolutional network, a BERT model, an improved ResNet model and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of resource optimization configuration, and in particular to a method and system for optimizing urban resource configuration based on multi-objective collaboration. Background Art

[0002] In recent years, the rapid development of emerging technologies such as big data, artificial intelligence, and the Internet of Things has provided new technical means for optimizing urban resource allocation. IoT devices can collect real-time data on the usage and environmental status of urban resources, providing a wealth of real-time data for resource allocation. Big data technologies can integrate heterogeneous data from multiple sources and uncover underlying patterns. Artificial intelligence algorithms have demonstrated powerful capabilities in solving complex problems and optimizing decision-making.

[0003] However, the application of these technologies in the field of optimizing urban resource allocation still faces numerous challenges. Existing technologies often focus on a single area or single goal, lacking systematic integration. Data fusion and processing techniques are still imperfect, making it difficult to effectively handle multimodal and heterogeneous data. Algorithmic models lack adaptability and robustness in the complex and dynamic environments of cities, and their interpretability is poor, making them difficult to meet the actual needs of urban resource allocation decision-making.

[0004] There is an urgent need for a new solution to break through the limitations of traditional technologies, make full use of the advantages of emerging technologies, solve key problems in the optimal allocation of urban resources, achieve efficient, intelligent and coordinated allocation of urban resources, and promote sustainable urban development. Summary of the Invention

[0005] In order to solve the above-mentioned problems, the present invention provides a method and system for optimizing urban resource allocation based on multi-objective collaboration.

[0006] In a first aspect, the present invention provides a method for optimizing urban resource allocation based on multi-objective collaboration, which adopts the following technical solutions:

[0007] A multi-objective collaborative urban resource optimization method includes:

[0008] Obtain data related to urban resources;

[0009] Preprocess the acquired urban resource-related data;

[0010] Construct a dynamic spatiotemporal knowledge graph based on pre-processed urban resource-related data;

[0011] Build a reinforcement learning-neural network hybrid optimization model and use the dynamic spatiotemporal knowledge graph as the model's embedding representation;

[0012] Use quantum computing and classical optimization algorithms to perform multi-objective training and optimization on the model;

[0013] Generate resource allocation decisions based on the optimized model.

[0014] Furthermore, the acquisition of urban resource-related data includes building a multi-source data acquisition network, adding urban micro-environment sensors and social media data in addition to traditional data sources, and dividing the data into four categories: spatiotemporal data, structured data, text data, and image data.

[0015] Furthermore, the obtained urban resource-related data is preprocessed, including using a spatiotemporal graph convolutional network to extract a node feature matrix from the urban resource spatiotemporal graph composed of nodes and edges, and performing convolution operations through convolution kernels; using a Transformer-based BERT model to extract features from text data, converting the text input into a word vector sequence, and calculating the context representation of each word through a multi-head attention mechanism; for image data, using an improved ResNet model to extract features, introducing a self-attention module, and adding attention weight calculation to the residual block to enhance the ability to extract key image features.

[0016] Furthermore, the method constructs a dynamic spatiotemporal knowledge graph based on the preprocessed urban resource-related data, including using a BiLSTM-CRF model based on an attention mechanism to perform entity recognition on text data, and adopting a joint learning framework to extract complex relationships; designs a multi-dimensional graph architecture, adopts an R-tree index structure in the spatial dimension, stores the graph state of the time interval corresponding to the time slice in the time dimension, and uses a deep learning model to extract image feature vectors and entity feature vectors for the association of graph entities in the image data domain, and calculates the association degree through cosine similarity.

[0017] Furthermore, the construction of a dynamic spatiotemporal knowledge graph based on pre-processed urban resource-related data also includes setting a trigger mechanism driven by multi-source data, triggering graph updates by priority weights and setting trigger thresholds; and after the graph is updated, using a graph edit distance algorithm to calculate the difference before and after the graph update, using a graph neural network GNN to update node features based on graph differences, and through multi-layer iterative learning, predicting potential entities and relationships, updating the knowledge graph to achieve evolutionary reasoning.

[0018] Furthermore, the construction of the reinforcement learning-neural network hybrid optimization model takes the dynamic spatiotemporal knowledge graph as the embedded representation of the model, including introducing a dual network structure and a competitive network structure based on DQN, wherein the dual network structure is used to reduce the Q value over-estimation problem by separating the target network and the evaluation network; the Q network is divided into an advantage function and a value function through the network structure, thereby calculating the Q values of the target network and the evaluation network, which are used to efficiently learn the value of the state and the advantage of the action.

[0019] Furthermore, the construction of the reinforcement learning-neural network hybrid optimization model takes the dynamic spatiotemporal knowledge graph as the embedding representation of the model, and also includes optimizing the Transformer encoder for the long sequence dependencies and complex semantic relationships of urban resource configuration data. In the multi-head attention mechanism, a dynamic weight adjustment strategy is introduced to assign dynamic weights to the attention calculations of different heads according to the importance and association strength of entities in the knowledge graph; finally, the embedding representation of the dynamic spatiotemporal knowledge graph is deeply integrated with other data features, wherein the graph attention network GAT is used to perform embedding learning on the knowledge graph to obtain the node embedding vector, and the knowledge graph embedding vector is integrated with the data features through a gating mechanism.

[0020] Furthermore, the method uses quantum computing and classical optimization algorithms to perform multi-objective training and optimization of the model, including introducing a hybrid framework of quantum computing and classical optimization algorithms, and using the parallelism of quantum computing to accelerate the search process; using a quantum genetic algorithm to perform a global search of model parameters, using quantum bit encoding to represent model parameters, and representing the state of quantum individuals with probability amplitudes; combining dynamic spatiotemporal knowledge graphs to perceive the spatiotemporal changes of urban resources in real time; and introducing a dynamic weight adjustment strategy for multiple objectives in urban resource allocation, by adjusting model parameters to ensure that the configuration plan is always close to the optimal Pareto frontier, thereby achieving dynamic balance optimization of multiple objectives.

[0021] Furthermore, the resource allocation decision is generated based on the optimized model, including inputting preprocessed real-time data into the optimized reinforcement learning-neural network hybrid model, and the model generates multiple candidate resource allocation plans based on the learned strategy and the current city resource status; constructing a multi-dimensional evaluation index system to comprehensively evaluate the candidate plans; using the hierarchical analysis method (AHP) to determine the weight of each evaluation index, calculate the comprehensive score of each candidate plan, sort the candidate plans according to the comprehensive score, and screen out the plan with the highest score as the final resource allocation decision plan.

[0022] The second aspect is a multi-objective collaborative urban resource optimization configuration system, comprising:

[0023] The data acquisition module is configured to acquire data related to urban resources;

[0024] The preprocessing module is configured to preprocess the acquired urban resource related data;

[0025] The graph module is configured to construct a dynamic spatiotemporal knowledge graph based on the preprocessed urban resource-related data;

[0026] The embedding module is configured to build a reinforcement learning-neural network hybrid optimization model and use the dynamic spatiotemporal knowledge graph as the embedding representation of the model;

[0027] The optimization module is configured to use quantum computing and classical optimization algorithms to perform multi-objective training optimization on the model;

[0028] The decision module is configured to generate resource allocation decisions based on the optimized model.

[0029] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, a method for optimizing and configuring urban resources based on multi-objective collaboration.

[0030] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; and the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded and executed by the processor to implement the method for optimizing and configuring urban resources based on multi-objective collaboration.

[0031] In summary, the present invention has the following beneficial technical effects:

[0032] This technical solution focuses on optimizing the allocation of urban resources, integrating multi-source data, building knowledge graphs and hybrid models, and achieving multi-objective optimization and dynamic decision-making. Its technical effects are significant, as follows:

[0033] 1. By building a multi-source data acquisition network, integrating multimodal heterogeneous data such as spatiotemporal, structured, textual, and image data, and applying spatiotemporal graph convolutional networks, BERT models, and improved ResNet models for feature extraction and fusion, we can more comprehensively and accurately acquire urban resource information. Furthermore, the construction of a dynamic spatiotemporal knowledge graph enables the refined extraction of multi-granular entities and relationships. Leveraging intelligent dynamic update and maintenance mechanisms, this graph reflects the spatiotemporal changes of urban resources in real time, providing rich and accurate knowledge support for subsequent decision-making.

[0034] 2. A hybrid reinforcement learning-neural network optimization model, combining a dual-network architecture, a competitive network structure, and optimized Transformer encoders, enhances the model's ability to learn urban resource allocation strategies, focusing more closely on key resource information and improving the accuracy of policy generation. Prioritized experience replay, curriculum learning strategies, and adversarial training are employed during training to accelerate training convergence and enhance model generalization. Hierarchical policy generation and a multi-agent collaborative decision-making mechanism make resource allocation decisions more systematic and rational. A knowledge graph-based decision interpretation mechanism improves the interpretability of decisions and enhances the trust of decision makers.

[0035] 3. A hybrid framework combining quantum computing and classical optimization algorithms, combined with meta-learning-guided transfer optimization, rapidly finds the optimal solution for model parameters and improves model adaptability. Parameter adjustment based on spatiotemporal dynamics and multi-objective dynamic balance optimization enable the model to promptly adapt to spatiotemporal changes in urban resources and multi-objective demands. Multi-dimensional evaluation and feedback optimization integrate cross-domain knowledge and combine human-machine collaborative feedback to continuously optimize resource allocation plans. Through a comprehensive process of resource allocation plan implementation, monitoring, effect evaluation, and feedback optimization, a closed-loop system is formed to continuously improve the efficiency and effectiveness of urban resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a schematic diagram of a method for optimizing urban resource configuration based on multi-objective collaboration according to Example 1 of the present invention. DETAILED DESCRIPTION

[0037] The present invention will be further described in detail below with reference to the accompanying drawings.

[0038] Example 1

[0039] Reference Figure 1 , a multi-objective collaborative urban resource optimization configuration method of this embodiment includes:

[0040] Obtain data related to urban resources;

[0041] Preprocess the acquired urban resource-related data;

[0042] Construct a dynamic spatiotemporal knowledge graph based on pre-processed urban resource-related data;

[0043] Build a reinforcement learning-neural network hybrid optimization model and use the dynamic spatiotemporal knowledge graph as the model's embedding representation;

[0044] Use quantum computing and classical optimization algorithms to perform multi-objective training and optimization on the model;

[0045] Generate resource allocation decisions based on the optimized model.

[0046] Specifically:

[0047] Complete technical solution for optimizing urban resource allocation

[0048] 1. Multimodal Heterogeneous Data Fusion and Preprocessing

[0049] (1) Data collection and classification

[0050] A multi-source data collection network was constructed. In addition to traditional data sources, new data sources included urban microenvironment sensors (such as air quality and noise monitoring equipment) and social media data (residents' discussions on resource needs). Data was divided into four categories: spatiotemporal data (land coordinates, spatiotemporal information on infrastructure distribution), structured data (project approval data, real estate registration information), text data (policy documents, public opinion feedback), and image data (satellite remote sensing images, building exterior images).

[0051] (2) Feature extraction and fusion

[0052] (1) Spatiotemporal data: Using the spatiotemporal graph convolutional network (ST-GCN), for the urban resource spatiotemporal graph G = (V, E) consisting of nodes V and edges E, the node feature matrix X∈R N×C (N is the number of nodes, C is the feature dimension), through the convolution kernel W∈R K×C×F (K is the convolution kernel size, F is the output feature dimension) perform convolution operation:

[0053]

[0054] Where Yi,f is the output of node i in feature dimension f, and N(i) is the set of neighboring nodes of node i.

[0055] (2) Text data: Use the Transformer-based BERT model for feature extraction, convert the text input into a word vector sequence T = [t1, t2, ..., tn], and calculate the contextual representation of each word through a multi-head attention mechanism:

[0056]

[0057] Where Q, K, and V are query, key, and value vectors respectively, and dk is the key vector dimension.

[0058] (3) Image data: Using the improved ResNet model, we introduce a self-attention module, add attention weight calculation to the residual block, and enhance the ability to extract key image features.

[0059] The above extracted features are fused through the gated fusion mechanism:

[0060] F = σ(G1X1+G2X2+G3X3+G4X4), where X1, X2, X3, X4 are spatiotemporal, structured, text, and image data features respectively, G1, G2, G3, G4 are learnable gating matrices, and σ is the activation function.

[0061] 2. Construction of Dynamic Spatiotemporal Knowledge Graph

[0062] (1) Refined extraction of multi-granularity entities and relationships

[0063] Hierarchical entity extraction: Use the BiLSTM-CRF model based on the attention mechanism for entity recognition. Suppose the input text sequence is x = [x1, x2, ..., xn], and the BiLSTM forward propagation calculates the hidden state Backward propagation calculates the hidden state Get bidirectional hidden state The multi-head attention mechanism calculates the attention weight αt,i:

[0064]

[0065] e t,i =score(h t ,h i )

[0066] Where score is the similarity calculation function, and the attention output is obtained by weighted summation Input zt into the CRF layer and calculate the probability of the label sequence y = [y1,y2,...,yn]:

[0067]

[0068] Where A is the transfer matrix, P is the emission matrix, Y x is the set of all possible label sequences.

[0069] (2) Complex relation extraction: Using a joint learning framework, we set the entity extraction loss function as Lentity, the relation classification loss function as Lrelation, and the joint loss function as Lrelation.

[0070] L=λ1L entity +λ2L relation , where λ1 and λ2 are weight coefficients.

[0071] For the derivation of implicit relationships, we use the probabilistic graphical model Bayesian network. Let the variable set be X = X1, X2, ..., Xn, and calculate the conditional probability P(Xi|Xi) according to the Bayesian formula:

[0072]

[0073] where X -i Indicates division by X i Other variable sets outside.

[0074] (2) High-dimensional heterogeneous dynamic spatiotemporal knowledge graph structure design

[0075] (1) Multi-dimensional graph architecture: In the spatial dimension, an R-tree index structure is used. For a spatial object o, its minimum bounding rectangle (MBR) is MBR(o) = (x min ,ymin ,x max ,y max ), R-tree node n contains the child node pointer set P and the MBR (n) covering the child node. In the time dimension, the time slice t corresponds to the time interval [start t ,end t ], store the spectrum state St in this interval.

[0076] (2) Heterogeneous data fusion structure: For the association between image data and graph entities, the deep learning model is used to extract the image feature vector fimage and the entity feature vector fentity, and the association degree is calculated by cosine similarity. When sim exceeds the threshold, an association is established.

[0077] (3) Intelligent dynamic update and maintenance mechanism

[0078] (1) Multi-source data-driven trigger mechanism: Assume that the data type set is D = d1, d2, ..., dm, the corresponding priority weights are w = w1, w2, ..., wm, and the priority of data di is pi = wi * si, where si is the data importance score. When data di arrives and pi exceeds the trigger threshold T, the graph is triggered to update.

[0079] (2) Incremental update and conflict resolution strategy: The graph edit distance algorithm is used to calculate the difference between the graph before and after the update. Assume that the graphs G1 = (V1, E1) and G2 = (V2, E2). The graph edit distance d(G1, G2) is obtained by minimizing the cost of inserting, deleting, and modifying nodes and edges:

[0080]

[0081] Where ops is the operation sequence and cost(op) is the cost of the operation op.

[0082] (3) Automatic evolution and reasoning of knowledge graph: Using graph neural network (GNN) for reasoning, let the node feature matrix H l For the l-th layer node features, the node features are updated through graph convolution operations:

[0083]

[0084] Where N(u) is the set of neighboring nodes of node u, Cu,v is the normalization constant, and W l is the weight matrix, b l is the bias vector. Through multi-layer iterative learning, potential entities and relationships are predicted and the knowledge graph is updated.

[0085] 3. Reinforcement Learning-Neural Network Hybrid Optimization Model

[0086] (1) Model architecture in-depth design

[0087] (1) Deep Q Network (DQN) Enhanced Structure

[0088] Based on the traditional DQN, a dual network structure (DoubleDQN) and a competitive network structure (DuelingDQN) are introduced. The dual network structure reduces the problem of Q value overestimation by separating the target network and the evaluation network. Let the evaluation network parameter be θ and the target network parameter be θ - , the evaluation network parameters are copied to the target network every certain number of steps. When calculating the target Q value, the evaluation network is used to select the action, and the target network calculates the Q value:

[0089]

[0090] The competitive network structure divides the Q network into two parts: advantage function A(s,a;θ) and value function V(s;θ).

[0091]

[0092] Calculate the Q value, where |A| is the size of the action space. This structure can more efficiently learn the value of states and the advantages of actions, improving the model's ability to learn urban resource allocation strategies.

[0093] (2) Transformer encoder optimization

[0094] The Transformer encoder is optimized for the long sequence dependencies and complex semantic relationships of urban resource configuration data. A dynamic weight adjustment strategy is introduced in the multi-head attention mechanism. According to the importance and association strength of entities in the knowledge graph, dynamic weights are assigned to the attention calculations of different heads. Let the importance score of entity ei be Iei, and the formula is used to calculate the weight of each head. Calculate the weight βh of the h-th attention head, where αh,i is the weight of the h-th attention head when calculating the attention of entity ei. The final attention output is:

[0095] Make the model more focused on key resource information and improve the accuracy of strategy generation.

[0096] (3) Knowledge Graph Embedding Fusion

[0097] The embedded representation of the dynamic spatiotemporal knowledge graph is deeply integrated with other data features. The graph attention network (GAT) is used to embed the knowledge graph and obtain the node embedding vector hv. For each node v, the formula e ij =LeakyReLU(a T [Wh i ||Whj ]) Calculate the attention coefficient eij between nodes i and j, where a is the attention vector and W is the learnable weight matrix. The attention weight is obtained after normalization Then we get the node embedding Fusing the knowledge graph embedding vector with data features through a gating mechanism: fusion =σ(G kg h kg +G data F data ), where hkg is the knowledge graph embedding vector, Fdata is other data features, Gkg and Gdata are learnable gating matrices, which realize the organic combination of knowledge graph information and resource data and provide richer information for model decision-making.

[0098] (2) Innovation and Optimization of the Training Process

[0099] (1) Priority Experience Replay

[0100] Improve the experience replay mechanism and use Prioritized Experience Replay. Assign a priority to each experience sample based on the TD error of the sample δ = |yQ(s, a; θ)| Where β is a parameter that controls the degree of influence of priority (0<β≤1). When sampling, samples are drawn from the experience pool according to the priority ratio. The formula is Here, α is a parameter that adjusts the sampling bias (0 < α ≤ 1). This approach enables the model to learn important empirical samples more frequently, accelerating training convergence. It is particularly suitable for learning complex and critical decision-making scenarios in urban resource allocation.

[0101] Course Learning Strategies

[0102] (2) Introducing a curriculum learning strategy, designing training task sequences from simple to complex based on the difficulty of urban resource allocation problems. In the early stages of training, simple resource allocation scenarios are set, such as land use adjustments in a single region, to allow the model to quickly learn basic strategies. As training progresses, the difficulty of the tasks is gradually increased, introducing complex scenarios such as multi-region collaborative configuration and consideration of multiple resource types. In this way, the model is helped to gradually establish complex decision-making capabilities, avoiding learning difficulties or falling into local optimality due to excessive task difficulty in the early stages of training.

[0103] (3) Adversarial training enhances generalization

[0104] Adversarial training is used to improve the model's generalization ability. An adversarial network is introduced, which attempts to generate interference data that can cause the main model to make incorrect decisions. The main model then strives to identify the interference data and maintain correct decisions. During training, an adversarial loss term Ladv is added to the main model's loss function. The main model and the adversarial network are optimized through a minimax game. The adversarial loss function is Ladv = \mathbbEx\simpdata[\log(1-D(x))]+\mathbbEx'\simpadv[\log(D(x'))], where D is the adversarial network (discriminator), pdata is the real data distribution, and padv is the interference data distribution generated by the adversarial network. Through adversarial training, the model's robustness to uncertainties and anomalies in urban resource allocation is enhanced.

[0105] (3) Strategy Generation and Decision Optimization

[0106] (1) Hierarchical strategy generation

[0107] A hierarchical strategy generation framework is constructed, categorizing urban resource allocation decisions into three levels: macro, meso, and micro. The macro level formulates an overall resource allocation strategy, such as the resource allocation ratios for different functional areas of the city. The meso level, based on the macro strategy, determines resource allocation plans for specific areas, such as the project portfolio for a commercial district. The micro level develops detailed resource utilization plans for specific projects, such as the material procurement and construction schedule for a single building.

[0108] Strategy generation at each level is handled by a different sub-network, which collaborates through information transfer and feedback mechanisms. The macro-strategy layer provides guidance to the meso-strategy layer, which further refines and provides feedback to adjust macro-strategy. The micro-strategy layer then develops specific implementation plans based on the meso-strategy layer's requirements and provides feedback to both the meso- and macro-strategies. This creates a multi-layered, coordinated strategy generation system, improving the systematicity and rationality of resource allocation decisions.

[0109] (2) Multi-agent collaborative decision-making

[0110] The Multi-Agent Reinforcement Learning mechanism is introduced, which treats different types of resources (such as land, capital, and manpower) or different management departments as independent agents. Each agent makes decisions based on its own goals and constraints, and interacts with other agents through communication and collaboration mechanisms. The joint action space A = A1 × A2 × … × A n Represents the action combination of all agents and the joint reward function The rewards of each agent are comprehensively considered, where Ri is the reward of the i-th agent and wi is the weight coefficient. Through training, agents learn to cooperate and jointly optimize urban resource allocation. For example, the land resource agent collaborates with the capital agent to reasonably allocate funds while ensuring the quality of land development, thereby maximizing resource utilization efficiency.

[0111] (3) Explainable decision support

[0112] To improve the explainability of decisions, a decision explanation mechanism based on a knowledge graph is introduced. After generating a resource allocation strategy, the entity relationships and rules in the knowledge graph are used to provide an explanation for the decision. For example, when the model decides to increase the construction of public facilities in a certain area, relevant information such as the population density of the area, the coverage of existing facilities, and the requirements of policies and regulations for public facility configuration are extracted from the knowledge graph. Natural language generation technology is used to explain the decision reasons to decision makers in an easy-to-understand manner, thereby enhancing their trust and understanding of the model's decisions and facilitating the execution and adjustment of decisions in real-world applications.

[0113] 4. Model Optimization and Dynamic Adjustment

[0114] (1) Algorithm fusion innovation and optimization

[0115] (1) A hybrid framework of quantum computing and classical optimization algorithms is introduced to accelerate the search process by utilizing the parallelism of quantum computing. A quantum genetic algorithm (QGA) is used to perform a global search of model parameters. Quantum bits (qubits) are encoded to represent model parameters, and the states of quantum individuals are represented by probability amplitudes. In quantum gate operations, the probability amplitude of the qubit is adjusted by rotating the gate U(θ):

[0116]

[0117] Where α and β are the probability amplitudes of the quantum bits, and θ is the rotation angle. In each evolutionary generation, quantum measurements are used to collapse the quantum state into a classical solution. This solution is then locally optimized using classical local search algorithms (such as simulated annealing), forming a quantum-classical collaborative optimization mechanism that rapidly finds the optimal solution for the model parameters and improves the model's adaptability in complex urban resource configuration scenarios.

[0118] (2) Meta-learning-guided transfer optimization

[0119] A meta-learning framework is constructed to learn the optimization patterns in resource allocation tasks in different urban areas or different time periods. The gradient-based meta-learning algorithm MAML is used. During the meta-training phase, the adapted parameters θ' are obtained for multiple source tasks Tsource through a small number of gradient updates:

[0120]

[0121] Where α is the learning rate, is the loss function of the source task. When θ′ is obtained through meta-learning, it is used as the initial parameter and quickly optimized in combination with the target task data. Through transfer learning, the convergence speed of the model in the new scenario is accelerated, reducing the training time and data requirements.

[0122] (2) Dynamic Adaptive Adjustment Mechanism

[0123] (1) Parameter adjustment of spatiotemporal dynamic perception

[0124] Combined with the dynamic spatiotemporal knowledge graph, the spatiotemporal changes of urban resources can be perceived in real time. A spatiotemporal attention mechanism is designed to assign dynamic weights to model parameter adjustments based on the spatiotemporal association information of entities in the knowledge graph. For the time dimension, the attention weight ωt of different time slices is calculated:

[0125]

[0126] Where ε is the set of entities in the knowledge graph, and TimeScore(e,t) represents the time-related score of entity e at time t. For the spatial dimension, the influence weights of spatial neighborhood entities are calculated based on the R-tree index structure, ultimately resulting in a comprehensive spatiotemporal weight. This is used to adjust the step size and direction of model parameter updates, enabling the model to promptly adapt to new demands arising from spatiotemporal changes in urban resource allocation.

[0127] (2) Multi-objective dynamic balance optimization

[0128] A dynamic weight adjustment strategy is introduced to address the multiple objectives (economic, social, and ecological benefits) in urban resource allocation. A target importance assessment model is established. Based on real-time data and policy guidance, the weights of each target are dynamically calculated using the analytic hierarchy process (AHP) combined with a fuzzy comprehensive evaluation method. When the external environment changes (such as policy adjustments or emergencies), the target weights are reassessed and the model optimization direction is adjusted. Furthermore, a Pareto front dynamic tracking algorithm is employed to monitor changes in the Pareto front in real time within the target space. By adjusting model parameters, the allocation plan is kept close to the optimal Pareto front, achieving dynamic balanced optimization of multiple objectives.

[0129] (3) Multi-dimensional evaluation feedback optimization

[0130] (1) Cross-domain knowledge integration evaluation

[0131] Construct a cross-disciplinary knowledge-integrated evaluation system that, in addition to traditional resource allocation indicators, integrates knowledge from multiple fields, including economics, sociology, and ecology. Introducing complex network analysis methods, the urban resource allocation system is viewed as a complex network. Indicators such as node degree centrality and betweenness centrality are used to assess the importance and influence of resource nodes within the system. Simultaneously, ecological models (such as the ecological footprint model) are used to assess the impact of resource allocation on the ecological environment. Feeding these multi-disciplinary evaluation results into model optimization ensures that the model not only focuses on short-term benefits but also comprehensively considers long-term system stability and sustainability.

[0132] (2) Human-machine collaborative feedback optimization

[0133] A feedback mechanism for human-machine collaboration has been established. Decision-makers can intuitively view the model's resource allocation plans and evaluation results through a visual interface and provide feedback based on their practical experience and expertise. Natural language processing technology is used to convert decision-maker feedback into parameter adjustment signals that the model can understand. Combined with the model's own evaluation results, this forms a two-way feedback optimization loop. For example, if a decision-maker believes that the public facility allocation plan for a certain area does not conform to the living habits of local residents, the model will process this feedback, adjust the relevant parameters, and regenerate the allocation plan. This achieves a complementary advantage between human and machine, and improves the rationality and feasibility of the resource allocation plan.

[0134] 5. Resource Allocation Decision Generation Based on Optimization Model

[0135] (1) Configuration plan generation

[0136] Preprocessed real-time data is fed into an optimized reinforcement learning-neural network hybrid model. Based on the learned strategy and the current state of urban resources, the model generates multiple candidate resource allocation scenarios. Each scenario includes detailed information such as land use adjustment plans, project construction priorities, and resource allocation amounts. For example, for land resource allocation, the scenario will clearly define the development type (residential, commercial, public facilities, etc.) and development intensity of different plots; for project resource allocation, it will determine the launch time and funding allocation for each project.

[0137] (2) Program evaluation and screening

[0138] A multi-dimensional evaluation index system will be established to comprehensively assess candidate solutions. Evaluation indicators include, but are not limited to, economic benefit indicators (such as expected land value appreciation and project investment return), social benefit indicators (such as the number of new jobs and the degree of improvement in residents' living convenience), ecological benefit indicators (such as carbon emission reduction and the increase in green space coverage), and feasibility indicators (such as policy compliance, technical feasibility, and difficulty in raising funds).

[0139] The analytic hierarchy process (AHP) is used to determine the weight of each evaluation indicator and calculate the comprehensive score of each candidate solution:

[0140]

[0141] Where Scorej is the comprehensive score of the jth candidate solution, wi is the weight of the i-th evaluation indicator, and Scoreij is the score of the jth candidate solution on the i-th evaluation indicator. The candidate solutions are sorted according to the comprehensive scores, and the solution with the highest score is selected as the final resource allocation decision solution.

[0142] VI. Implementation and Monitoring of Resource Allocation Plans

[0143] (1) Implementation of the plan

[0144] Develop a detailed implementation plan for the resource allocation plan, clarifying the responsibilities, division of labor, and timelines for each department. For example, the land management department will be responsible for approving land use changes and land transfers; the construction department will be responsible for project construction supervision; and the finance department will be responsible for fund allocation and oversight. During the implementation process, establish a cross-departmental collaborative working mechanism to ensure close coordination among all links and ensure the smooth implementation of the resource allocation plan.

[0145] (2) Real-time monitoring

[0146] Leveraging IoT devices, big data platforms, and dynamic spatiotemporal knowledge graphs, we monitor the implementation of resource allocation plans in real time. We collect real-time information on land development progress, project construction status, and resource usage, and analyze it against the planned schedule. For example, we monitor construction site progress through IoT sensors, comparing actual progress with planned progress. We also leverage big data platforms to analyze fund usage to ensure proper allocation and use within budget. If deviations are detected, early warning signals are issued and emergency adjustment mechanisms are activated.

[0147] 7. Effect Evaluation and Feedback Optimization

[0148] (1) Effect evaluation

[0149] After the resource allocation plan is implemented, a comprehensive effectiveness evaluation will be conducted. Using a combination of quantitative and qualitative methods, the achievement of economic, social, and ecological benefits will be assessed. For example, economic benefits will be assessed by compiling data such as actual land value appreciation and project returns; social benefits will be assessed through questionnaires and field interviews to understand residents' satisfaction with the improved living environment; and ecological benefits will be assessed by monitoring changes in environmental indicators (such as air and water quality).

[0150] (2) Feedback Optimization

[0151] Based on the results of the effectiveness evaluation, analyze the problems and shortcomings of the resource allocation plan. Feedback the evaluation results into various links such as data collection, model training, and optimization and adjustment, and iteratively optimize the technical plan. For example, if it is found that the resource allocation of a certain area fails to achieve the expected economic benefits, the data characteristics of that area are re-analyzed, the model parameters or optimization algorithm are adjusted, and the resource allocation plan is regenerated, forming a closed-loop urban resource optimization allocation system to continuously improve the efficiency and effectiveness of urban resource allocation.

[0152] This solution fully demonstrates the entire process, from data processing to resource optimization. If you have any suggestions for adjustments to the steps or algorithms in this solution, or would like to add specific details, please feel free to submit them.

[0153] Example 2

[0154] This embodiment provides a system for optimizing urban resource allocation based on multi-objective collaboration, including:

[0155] The data acquisition module is configured as

[0156] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device for a method for optimizing and configuring urban resources based on multi-objective collaboration.

[0157] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to implement the urban resource optimization configuration method based on multi-objective collaboration.

[0158] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimizing urban resource allocation based on multi-objective collaboration, characterized in that: include: Obtain data related to urban resources; Preprocess the acquired urban resource-related data; Construct a dynamic spatiotemporal knowledge graph based on pre-processed urban resource-related data; Build a reinforcement learning-neural network hybrid optimization model and use the dynamic spatiotemporal knowledge graph as the model's embedding representation; Use quantum computing and classical optimization algorithms to perform multi-objective training and optimization on the model; Generate resource allocation decisions based on the optimized model.

2. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 1 is characterized in that: The acquisition of urban resource-related data includes building a multi-source data acquisition network, adding urban micro-environment sensors and social media data in addition to traditional data sources, and dividing the data into four categories: spatiotemporal data, structured data, text data, and image data.

3. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 2 is characterized in that: The obtained urban resource-related data is preprocessed, including using a spatiotemporal graph convolutional network to extract a node feature matrix from the urban resource spatiotemporal graph composed of nodes and edges, and performing a convolution operation through a convolution kernel; using a Transformer-based BERT model to extract features from text data, converting the text input into a word vector sequence, and calculating the contextual representation of each word through a multi-head attention mechanism; using an improved ResNet model to extract features from image data, introducing a self-attention module, and adding attention weight calculation to the residual block to enhance the ability to extract key image features.

4. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 3 is characterized in that: The method constructs a dynamic spatiotemporal knowledge graph based on preprocessed urban resource-related data, including using a BiLSTM-CRF model based on an attention mechanism to perform entity recognition on text data and adopting a joint learning framework to extract complex relationships; designs a multi-dimensional graph architecture, adopts an R-tree index structure in the spatial dimension, stores the graph state of time slices corresponding to time intervals in the time dimension, and uses a deep learning model to extract image feature vectors and entity feature vectors for the association of graph entities in the image data domain, and calculates the association degree through cosine similarity.

5. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 4 is characterized in that: The construction of a dynamic spatiotemporal knowledge graph based on pre-processed urban resource-related data also includes setting a trigger mechanism driven by multi-source data to trigger graph updates through priority weights and setting trigger thresholds; After the graph is updated, the graph edit distance algorithm is used to calculate the difference before and after the graph update. The graph neural network GNN is used to update the node features based on the graph differences. Through multi-layer iterative learning, potential entities and relationships are predicted, and the knowledge graph is updated to realize evolutionary reasoning.

6. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 5 is characterized in that: The reinforcement learning-neural network hybrid optimization model is constructed, and the dynamic spatiotemporal knowledge graph is used as the embedded representation of the model. This includes introducing a dual network structure and a competitive network structure based on DQN. The dual network structure is used to reduce the problem of Q value over-estimation by separating the target network and the evaluation network. The Q network is divided into an advantage function and a value function through the network structure, thereby calculating the Q values of the target network and the evaluation network, which are used to efficiently learn the value of the state and the advantage of the action.

7. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 6 is characterized in that: The proposed hybrid reinforcement learning-neural network optimization model uses a dynamic spatiotemporal knowledge graph as the model's embedded representation. It also optimizes the Transformer encoder for the long sequence dependencies and complex semantic relationships of urban resource allocation data. In the multi-head attention mechanism, a dynamic weight adjustment strategy is introduced to assign dynamic weights to the attention calculations of different heads based on the importance and association strength of entities in the knowledge graph. Finally, the embedded representation of the dynamic spatiotemporal knowledge graph is deeply integrated with other data features. The graph attention network GAT is used to embed the knowledge graph to obtain the node embedding vector, and the knowledge graph embedding vector is fused with the data features through a gating mechanism.

8. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 7 is characterized in that: The method uses quantum computing and classical optimization algorithms to perform multi-objective training optimization on the model, including introducing a hybrid framework of quantum computing and classical optimization algorithms, using the parallelism of quantum computing to accelerate the search process; using a quantum genetic algorithm to perform a global search on model parameters, using quantum bit encoding to represent model parameters, and representing the state of quantum individuals with probability amplitudes; combining dynamic spatiotemporal knowledge graphs to perceive the spatiotemporal changes of urban resources in real time; and introducing a dynamic weight adjustment strategy for multiple objectives in urban resource allocation, by adjusting model parameters to ensure that the configuration plan is always close to the optimal Pareto frontier, thereby achieving dynamic balance optimization of multiple objectives.

9. The method for optimizing urban resource allocation based on multi-objective collaboration according to claim 8, characterized in that: Generating resource allocation decisions based on the optimized model includes inputting pre-processed real-time data into the optimized reinforcement learning-neural network hybrid model. The model generates multiple candidate resource allocation plans based on the learned strategy and the current state of urban resources; and constructing a multi-dimensional evaluation index system to comprehensively evaluate the candidate plans. The analytic hierarchy process (AHP) is used to determine the weight of each evaluation indicator, calculate the comprehensive score of each candidate solution, sort the candidate solutions according to the comprehensive score, and select the solution with the highest score as the final resource allocation decision solution.

10. A city resource optimization configuration system based on multi-objective collaboration, characterized in that: include: The data acquisition module is configured to acquire city resource related data; The preprocessing module is configured to preprocess the acquired urban resource related data; The graph module is configured to construct a dynamic spatiotemporal knowledge graph based on the preprocessed urban resource-related data; The embedding module is configured to build a reinforcement learning-neural network hybrid optimization model and use the dynamic spatiotemporal knowledge graph as the embedding representation of the model; The optimization module is configured to use quantum computing and classical optimization algorithms to perform multi-objective training optimization on the model; The decision module is configured to generate resource allocation decisions based on the optimized model.

Images