Smart park-oriented cooperative control and dynamic resource scheduling method and system

Abstract

The invention discloses a cooperative control and dynamic resource scheduling method and system for a smart park, and relates to the technical field of program control. According to the method, the heterogeneous sensor network is deployed, and multi-source data is subjected to denoising, feature extraction and classified transmission; constructing a digital twinborn model, embedding equipment physical law constraints, optimizing the learning rate of an RCN prediction network in combination with historical data and a bee colony algorithm, and realizing high-precision state prediction; when the real-time data or the predicted data exceeds a threshold value, triggering global scheduling, generating an initial allocation scheme by using an improved contract network protocol, and generating a Pareto optimal scheme set of comprehensive cost, equipment utilization rate and service delay through a pea population algorithm; verifying the feasibility of the scheme based on digital twinning simulation, and dynamically distributing conflict resources by adopting virtual auction; the network load is reduced, the prediction progress is improved, the equipment load peak-valley difference is compressed, and the park resource scheduling efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of program control, and in particular, relates to a collaborative control and dynamic resource scheduling method and system for a smart park. BACKGROUND

[0002] In recent years, with the rapid development of Internet of Things (IoT), artificial intelligence (AI) and digital twin technology, smart parks, as an important part of urban intelligence, have become a core scenario for improving industrial efficiency and reducing operating costs. However, the complexity and heterogeneity of subsystems (such as energy, transportation, and security) within the park are increasing, leading to multiple challenges for traditional management methods.

[0003] Chinese Patent CN114637269B discloses an online management and scheduling system for water resources in a smart park and a method thereof; the present application integrates a remote scheduling platform, a water resource allocation system, and a water resource detection terminal, wherein the remote scheduling platform establishes data connection with the latter two through a communication network, and the water resource allocation system is connected with the detection terminal through a water supply pipe and communicated with at least one external water source supply pipeline through a flow guide pipe; the system uses a method covering three key steps of system configuration, water supply operation, and water supply scheduling.

[0004] The prior art lacks a unified data fusion and classification transmission mechanism, resulting in a coexistence of emergency response lag and conventional data bandwidth waste; traditional prediction models are disconnected from physical system laws, making it difficult to accurately reflect dynamic coupling relationships such as device energy consumption and load, and fixed learning rates result in low utilization efficiency of historical data; resource scheduling relies on a single threshold triggering mechanism, lacks multi-objective collaborative optimization capability, and is prone to cause device overload and resource allocation conflicts; existing conflict resolution solutions use static priority strategies, which cannot dynamically adapt to sudden demands, causing insufficient supply of resources for key subsystems. SUMMARY

[0005] (I) Technical problems solved

[0006] To solve the problems in the related art, the present application provides a collaborative control and dynamic resource scheduling method for a smart park. Through digital twinning and deep learning, combined with improved optimization algorithm technology, the present application solves the problems of data silos and insufficient real-time performance, disconnection between prediction models and physical laws, rigid resource scheduling and frequent conflicts, and weak dynamic adaptation capability.

[0007] (II) Technical solutions

[0008] To solve the above technical problems, the present application is implemented by the following technical solutions:

[0009] S1, deploy a heterogeneous sensor network, collect raw data through the heterogeneous sensor network, process the raw data to obtain real-time data, and set a classification transmission rule;

[0010] S2, construct a digital twin model of a smart park, transmit the real-time data to the digital twin model of the smart park according to the classification transmission rule, obtain a resource matrix and generate a physical law constraint, construct an RCN prediction network according to the physical law constraint, use historical smart park data in combination with an optimization algorithm to improve and optimize the RCN prediction network, and obtain an optimized RCN smart park data prediction network; input the real-time data into the optimized RCN smart park data prediction network to obtain prediction data;

[0011] S3, determine whether the real-time data or the prediction data exceeds a threshold value, and when there is a threshold value that exceeds the threshold value, trigger global scheduling and generate a Pareto solution set in combination with a seeking optimization algorithm;

[0012] Determine whether there is a solution in the Pareto solution set that will not cause resource conflict through the digital twin model of the smart park, if there is, select the optimal solution as the final scheduling solution, otherwise, select the solution with the smallest conflict degree, combine the auction algorithm to obtain the final scheduling solution, and use the final scheduling solution for resource scheduling;

[0013] Preferably, the S1 comprises the following steps:

[0014] S11, according to various subsystems of a smart park, set a subsystem set, install energy consumption monitoring devices, people flow detection devices, and device state sensors in the smart park to cover various subsystems in the subsystem set, and obtain a heterogeneous sensor network;

[0015] S12, collect raw data through the heterogeneous sensor network, denoise and normalize the raw data by each regional edge server, extract key features, and stamp time, to obtain real-time data;

[0016] S13, set an emergency event type set and a regular monitoring data type set;

[0017] When transmitting the real-time data, according to the emergency event type set and the regular monitoring data type set, the data in the real-time data is transmitted according to the classification transmission rule; the classification transmission rule is to set a transmission priority according to the data type: the emergency event is the highest level and is transmitted in real time; the regular monitoring data is periodically uploaded after batch compression;

[0018] The above steps break the data limitation of traditional single sensor type by deploying special sensors for subsystem characteristics to form a heterogeneous sensor network covering the whole domain; the original data is denoised, normalized (and feature extracted) to generate standardized real-time data with timestamps, reducing the cloud computing pressure and ensuring data timeliness; network resources are dynamically allocated according to data types, with emergency events marked as the highest priority and transmitted to the decision center in real time through a dedicated channel; regular data is compressed and packaged for batch uploading during off-peak hours; this mechanism reduces the transmission delay of emergency events and reduces the bandwidth occupancy rate of regular data, achieving optimal allocation of network resources.

[0019] Preferably, S2 comprises the following steps:

[0020] S21, set the device physical parameter set of the smart park; construct a three-dimensional model of the smart park based on BIM and GIS, import the device physical parameters into the three-dimensional model of the smart park according to the device physical parameter set, and establish an energy consumption-load correlation equation constraint condition to obtain a smart park digital twin model;

[0021] S22, transmit the real-time data to the smart park digital twin model, update the real-time dynamic parameters in the model, and generate a resource state matrix; according to the resource state matrix, obtain the physical law constraint of the smart park digital twin model;

[0022] S23, according to the physical law constraint of the smart park digital twin model, construct an RCN prediction network, collect historical smart park data, and improve and optimize the RCN network by using an optimization algorithm to obtain an optimized RCN smart park data prediction network;

[0023] S24, input the real-time data into the optimized RCN smart park data prediction network to obtain predicted data;

[0024] The above steps construct a high-precision three-dimensional model, import device physical parameters, define energy consumption-load correlation equations and other physical constraints, so that the digital twin model strictly follows the real physical law, avoids the "black box" deviation of the pure data-driven model, synchronizes real-time data to the twin model, generates a multi-dimensional resource state matrix, and extracts physical law constraints based on matrix analysis to provide prior knowledge for the prediction model, embeds physical constraints in the RCN, dynamically searches for the optimal learning rate using an optimization algorithm, and trains the model to learn data features and physical laws, thereby improving the prediction accuracy, inputs real-time data into the optimized RCN network, and outputs the state prediction of the future period to provide advanced decision-making basis for dynamic scheduling and avoid the lag of the traditional method "after response"; through the deep integration of physical law constraints and data-driven, the interpretability of the prediction result is guaranteed, and the model convergence speed is improved through dynamic learning rate optimization, realizing the leap from "perceiving the present" to "predicting the future".

[0025] Preferably, the S23 comprises the following steps:

[0026] S231, constructing an RCN prediction network according to the physical law constraints of the digital twin model of the smart park, and setting the learning rate of the RCN prediction network; setting the training accuracy threshold and training accuracy of the RCN prediction network;

[0027] S232, collecting historical smart park data, training the RCN prediction network using the historical smart park data, finding the optimal learning rate of the RCN prediction network in the training process combined with an optimization algorithm, obtaining the optimal solution, taking the optimal solution as the learning rate of the RCN prediction network, and obtaining an optimized RCN smart park data prediction network;

[0028] The above steps are specific steps for constructing an initial RNC network to an optimized RCN smart park data prediction network, which avoids the prediction result of the model falling into "mathematically feasible but physically infeasible" through physical law constraints, and overcomes the defect that the traditional gradient descent method is easy to fall into local optimum by dynamically generating a learning rate optimization combined with an optimization algorithm, thereby realizing the dual improvement of prediction accuracy and training efficiency.

[0029] Preferably, the step of finding the learning rate of the RCN prediction network in the training process combined with an optimization algorithm in S232 to obtain the optimal solution comprises the following steps:

[0030] S2321, constructing a bee population, setting the size of the bee population, and setting the maximum number of optimization iterations;

[0031] According to the learning rate of the RCN prediction network, randomly setting the initial position of the bee population, obtaining a bee population initial position set, and

[0032] S2322, define a fitness function according to the training accuracy threshold and the training accuracy;

[0033] S2323, perform an iteration operation on the initial position set of the bee colony, the higher the fitness value, the better the position; in each iteration process, the fitness value of each position in the initial position set of the bee colony is calculated according to the fitness function, the positions of each bee in the initial position set of the bee colony are updated in descending order of the fitness value, and the best bee individual position in the bee colony and the global best bee position are obtained in each iteration process;

[0034] S2324, repeat S2323, when the maximum optimization iteration number is reached, stop iteration, and take the global best bee position as the optimal solution;

[0035] The above steps set the bee colony size and the maximum iteration number through population initialization, randomly generate an initial position set within a reasonable learning rate range, each bee represents a candidate learning rate, and ensure full coverage of the search space; define a fitness function to convert the deviation of the training accuracy from the threshold into a fitness value, the higher the value, the better the learning rate; in each iteration, the fitness of all bees is calculated and sorted, guiding low fitness bees to move to high fitness areas, while retaining some bees to randomly explore new areas to prevent local optimization; after multiple iterations, the global best position corresponds to the optimal learning rate, which improves search efficiency and reduces prediction error; the bee colony algorithm balances exploration and development through a mechanism that not only utilizes swarm intelligence for rapid convergence, but also maintains diversity through random disturbance, avoiding the over-reliance on initial values of traditional gradient descent methods, ensuring that the learning rate optimization has both efficiency and robustness.

[0036] Preferably, the S3 comprises the following steps:

[0037] S31, set a set of dynamic scheduling trigger thresholds; when real-time data or predicted data breaks a scheduling trigger threshold in the set of dynamic scheduling trigger thresholds, trigger global scheduling;

[0038] After triggering global scheduling, each subsystem in the subsystem set publishes the current resource availability and demand list to the decision center;

[0039] The decision center publishes a task tender, each subsystem in the subsystem set bids according to its own state, and the decision center obtains an initial allocation scheme set by using an improved contract net protocol and combining with the Nash equilibrium strategy of game theory;

[0040] S32, generate a Pareto scheme set according to the initial allocation scheme set and a multi-objective optimization algorithm;

[0041] S33, set the priority of resource allocation of each subsystem according to the importance of each subsystem in the subsystem set to obtain a priority rule;

[0042] inputting the schemes in the pareto scheme set into the digital twin model of the smart park respectively, simulating and verifying the schemes in the pareto scheme set, and obtaining verification results;

[0043] if the verification result is that no scheme in the pareto scheme set will cause resource conflict, selecting an optimal scheme from the schemes that will not cause resource conflict according to the priority rule, and using the optimal scheme as a final scheme for resource scheduling;

[0044] if the verification result is that all the schemes in the pareto scheme set will cause resource conflict, selecting a scheme with the minimum conflict degree, virtually auctioning the conflict resources in the scheme with the minimum conflict degree, each subsystem submitting a bid, the highest bidder obtaining the priority, obtaining a final scheme, and using the final scheme for resource scheduling;

[0045] The above steps generate an initial resource allocation scheme set by starting a global scheduling mechanism, combining the resources and demand conditions available to each subsystem, using an improved contract net protocol and Nash equilibrium strategy in game theory, generate an initial resource allocation scheme set; a pareto scheme set is generated from the initial allocation scheme set by combining a multi-objective optimization algorithm, and an optimal balanced solution is found in a set of conflicting objectives, thereby realizing efficient and reasonable allocation of resources, digital twin simulation and dynamic conflict resolution, ensuring the feasibility and fairness of the scheduling scheme; the pareto scheme set is input into the twin model for simulation, resource conflicts are detected by physical law constraints, and conflict-free schemes are selected; according to the importance of the subsystem, a priority rule is set to select the optimal scheme from the conflict-free schemes; if all the schemes are conflicting, the conflict degree is calculated, the scheme with the lowest conflict degree is selected, and a virtual auction is started for the conflict resources, each subsystem submits a bid containing demand urgency and budget, the decision center calculates the weighted score according to the "urgency x budget", and the energy submits "allocated resources, the highest bidder uses first; this mechanism improves the demand satisfaction rate of high-priority subsystems and avoids resource monopoly; through the combination of pre-verification and dynamic auction, the physical feasibility of the scheduling scheme is ensured, and through the urgency-budget two-dimensional bidding, the fair and efficient allocation of scarce resources is realized, and the dynamic demand of the emergency is adapted.

[0046] Preferably, the S32 comprises the following steps:

[0047] S321, constructing a target function combining comprehensive cost, equipment utilization rate and service response time;

[0048] S322, constructing a pea population, setting the size of the pea population, each pea in the pea population representing a treatment scheme in the initial allocation scheme set, and setting the maximum number of optimization iterations;

[0049] S323, iteratively operating on the pea population; selecting and crossing peas in the pea population according to the multi-objective function to obtain an operated pea population;

[0050] S324, repeating S323, and stopping iteration when a maximum number of iterations for optimization is reached to obtain a set of Pareto solutions;

[0051] The above steps realize intelligent scheduling decision of resources in the smart park through dynamic threshold triggering and multi-objective collaborative optimization; when real-time data or predicted data breaks through the threshold, global scheduling is triggered to avoid the problems of "early intervention" or "response lag" caused by traditional fixed threshold; an improved contract net protocol is used, each subsystem publishes a resource list to the decision center, and a Nash equilibrium strategy based on game theory is used to balance local interests and global optimization to generate an initial allocation solution set; a multi-objective function of comprehensive cost, equipment utilization rate and service delay is constructed, and a pea population algorithm is used for selection and crossing operation to iteratively generate a set of Pareto solutions; the peak-valley difference of equipment load is reduced, and the service response time is compressed; through the coupling of dynamic threshold and multi-objective optimization, the "patching east to patch west" problem caused by single-objective optimization is avoided, and diversified scheduling strategies are provided through the set of Pareto solutions, supporting decision makers to flexibly select the optimal solution according to the real-time scene.

[0052] The collaborative control and dynamic resource scheduling system for the smart park is used to implement the above-mentioned collaborative control and dynamic resource scheduling method for the smart park, and comprises a heterogeneous sensor network and a hierarchical transmission module, a digital twin and optimization prediction module, a dynamic scheduling and Pareto optimization module, and a conflict resolution and resource scheduling module.

[0053] The heterogeneous sensor network and hierarchical transmission module is used to realize multi-source data acquisition by deploying a heterogeneous sensor network; the original data is denoised and normalized, key features are extracted and time stamped to generate real-time data; based on the set of emergency event types and the set of regular monitoring data, classification transmission rules are set;

[0054] The digital twin and optimization prediction module is used to construct a three-dimensional model of the smart park, import physical parameters of equipment, and establish an energy consumption-load correlation equation to form a digital twin model; the model dynamic parameters are updated through real-time data to generate a resource state matrix and physical law constraints; an RCN prediction network is constructed in combination with the constraints, a bee colony optimization algorithm and historical data are used to train the network, and finally an optimized RCN prediction model is obtained;

[0055] The dynamic scheduling and Pareto optimization module is used for triggering global scheduling when real-time or predicted data exceeds a threshold, each subsystem publishes resource demand, and the decision center generates an initial allocation scheme based on an improved contract net protocol and Nash equilibrium strategy; a multi-objective function of comprehensive cost, equipment utilization and service delay is constructed, and a pea population is used to generate a Pareto solution set;

[0056] The conflict resolution and resource scheduling module is used for inputting the Pareto solution into a digital twin model for simulation verification, screening conflict-free solutions and selecting the optimal solution according to the subsystem priority; if all solutions are in conflict, the solution with the smallest conflict degree is selected, and the conflicting resources are virtually auctioned; the subsystems bid based on demand urgency and budget, and the higher bidder obtains the use right.

[0057] (Three) beneficial effects

[0058] The present application has the following beneficial effects:

[0059] The present application has the following beneficial effects:

[0060] The present application embeds physical law constraints into the RCN prediction network architecture, establishes energy-consumption-load correlation equations and other physical constraint conditions, and dynamically optimizes the learning rate using an improved bee colony algorithm, so that the prediction network can maintain the advantages of time series modeling while reducing prediction error and improving prediction accuracy.

[0061] The present application has the following beneficial effects:

[0062] The present application has the following beneficial effects:

[0063] Of course, any product implementing the present application does not necessarily need to achieve all the advantages described above at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the description of the embodiments. Obviously, the drawings described in the following description are only some embodiments of the application, and all other drawings obtained by those of ordinary skill in the art without creative effort based on these drawings belong to the protection scope of the application.

[0065] Figure 1 The flowchart of the method for collaborative control and dynamic resource scheduling of the smart park of the application;

[0066] Figure 2 The flowchart of the optimized RCN smart park data prediction network obtained in the method for collaborative control and dynamic resource scheduling of the smart park of the application;

[0067] Figure 3 The flowchart of obtaining the initial allocation scheme set in the method for collaborative control and dynamic resource scheduling of the smart park of the application;

[0068] Figure 4 The flowchart of obtaining the Pareto scheme set in the method for collaborative control and dynamic resource scheduling of the smart park of the application;

[0069] Figure 5 The flowchart of obtaining the final scheme in the method for collaborative control and dynamic resource scheduling of the smart park of the application;

[0070] Figure 6 The flowchart of the method for collaborative control and dynamic resource scheduling of the smart park of the application. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some embodiments of the application, but not all the embodiments. Based on the embodiments of the application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the protection scope of the application.

[0072] In the description of the application, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inner" and the like indicate the orientation or positional relationship, which are only used to facilitate the description of the application and simplify the description, and do not indicate or imply that the indicated component or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the application.

[0073] Embodiment one:

[0074] Please refer to Figure 1 , Figure 2 , Figure 3 ,Figure 4 、 Figure 5 The application discloses a smart park-oriented collaborative control and dynamic resource scheduling method, comprising the following steps:

[0075] S1, deploying a heterogeneous sensor network, collecting raw data through the heterogeneous sensor network and processing to obtain real-time data; setting a classification transmission rule;

[0076] The S1 comprises the following steps:

[0077] S11, according to various subsystems of the smart park, setting a subsystem set e = {e1, e2,..., e i ,...e r}, wherein e i represents the i-th subsystem, and r represents the total number of subsystems; installing energy consumption monitoring devices (such as smart meters, temperature and humidity sensors), people flow detection devices (such as infrared counters, cameras), device state sensors (such as vibration, current monitoring) in the smart park to cover various subsystems in the subsystem set, obtaining a heterogeneous sensor network; the subsystems are, for example, energy, transportation and security subsystems;

[0078] S12, collecting raw data through the heterogeneous sensor network, denoising and normalizing the raw data by each regional edge server, extracting key features (such as people flow density peak value, device abnormal vibration frequency), and stamping a time stamp to obtain real-time data;

[0079] S13, setting an emergency event type set a = {a1, a2,..., a i ,...a n}, wherein a i represents the i-th emergency event, and n represents the total number of emergency event types; setting a regular monitoring data type set b = {b1, b2,..., b i ,...b m}, wherein b i represents the i-th regular monitoring data, and m represents the total number of regular monitoring data types;

[0080] When transmitting the real-time data, according to the emergency event type set and the regular monitoring data type set, the data in the real-time data is transmitted according to the classification transmission rule; the classification transmission rule is to set a transmission priority according to the data type: the emergency event (such as fire alarm) is the highest level and is transmitted in real time; the regular monitoring data (such as room temperature) is periodically uploaded after being batch compressed;

[0081] S2, constructing a digital twin model of the smart park; transmitting real-time data to the digital twin model of the smart park according to a classification transmission rule to obtain a resource matrix and generate a physical law constraint; constructing an RCN prediction network according to the physical law constraint, improving and optimizing the RCN prediction network using historical smart park data in combination with an optimization algorithm to obtain an optimized RCN smart park data prediction network; inputting real-time data into the optimized RCN smart park data prediction network to obtain prediction data;

[0082] The S2 includes the following steps:

[0083] S21, setting a device physical parameter set d = {d1, d2,..., d i ,...d q} of the smart park, wherein d i represents a physical parameter of the i-th device, and q represents the total number of devices; constructing a three-dimensional model of the smart park based on BIM and GIS, importing device physical parameters (such as air conditioner energy efficiency ratio and elevator load upper limit) into the three-dimensional model of the smart park according to the device physical parameter set, and establishing an energy consumption-load correlation equation constraint condition (such as "air conditioner power = energy efficiency ratio x refrigerating capacity") to obtain a digital twin model of the smart park;

[0084] S22, transmitting the real-time data to the digital twin model of the smart park to update real-time dynamic parameters (such as current power consumption and parking occupancy) in the model and generate a resource state matrix (such as a row representing a device and a list representing time / energy consumption / load parameters); obtaining a physical law constraint (such as "elevator load ≤ 80% rated value") of the digital twin model of the smart park according to the resource state matrix.

[0085] S23, constructing an RCN prediction network according to the physical law constraint of the digital twin model of the smart park, collecting historical smart park data, and improving and optimizing the RCN network in combination with an optimization algorithm to obtain an optimized RCN smart park data prediction network;

[0086] The S23 includes the following steps:

[0087] S231, constructing an RCN prediction network according to the physical law constraint of the digital twin model of the smart park, setting a learning rate of the RCN prediction network; setting a training accuracy threshold value of the RCN prediction network as g1 and a training accuracy as g2;

[0088] S232, collecting historical smart park data, training the RCN prediction network using the historical smart park data, finding an optimal learning rate of the RCN prediction network in combination with an optimization algorithm during the training process, obtaining an optimal solution, taking the optimal solution as the learning rate of the RCN prediction network, and obtaining an optimized RCN smart park data prediction network;

[0089] The learning rate of the RCN prediction network in the training process in S232 is combined with the optimization algorithm to find the optimal solution, including the following steps:

[0090] S2321, construct a bee population, set the size of the bee population as j, and the bee population is represented as k = {k1, k2,..., k i ,...,k j}, where k i represents the i-th bee in the bee population; set the maximum number of optimization iterations;

[0091] According to the learning rate of the RCN prediction network, the initial position of the bee population is randomly set, and the initial position set of the bee population is obtained f = {f1, f2,..., f i ,...,f j}, where f i represents the position of the i-th bee in the bee population;

[0092] S2322, define the fitness function according to the training accuracy threshold g1 and the training accuracy g2, and the fitness function formula is as follows,

[0093]

[0094] S2323, perform iteration operation on the initial position set of the bee population, the higher the fitness value, the better the position; in each iteration process, according to the fitness function, the fitness value of each position in the initial position set of the bee population is calculated, and the position of each bee in the initial position set of the bee population is updated according to the fitness value from high to low, and the best bee individual position in the bee population and the global best bee position are obtained in each iteration process;

[0095] S2324, repeat S2323, when the maximum number of optimization iterations is reached, stop iteration, and take the global best bee position as the optimal solution;

[0096] S24, input the real-time data into the optimized RCN smart park data prediction network to obtain predicted data;

[0097] S3, determine whether the real-time data or the predicted data exceeds the threshold value, when there is a threshold value, trigger global scheduling and generate a set of Pareto solutions combined with the optimization algorithm;

[0098] Determine whether there is a solution in the set of Pareto solutions that will not cause resource conflict through the smart park digital twin model, if there is, select the optimal solution as the final scheduling scheme, otherwise, select the solution with the smallest conflict degree, combine the auction algorithm to obtain the final scheduling scheme, and use the final scheduling scheme for resource scheduling;

[0099] The S3 includes the following steps:

[0100] S31, set a dynamic scheduling trigger threshold set c = {c1, c2,..., c i ,...c p}, wherein c i represents the i-th scheduling trigger threshold, and p represents the total number of scheduling trigger thresholds; (such as the flow of people in a certain area exceeding 80% of the designed capacity, the energy consumption of a single air conditioner exceeding the limit for 5 minutes in a row), when the real-time data or the predicted data breaks through the scheduling trigger threshold in the dynamic scheduling trigger threshold set, global scheduling is triggered;

[0101] After triggering global scheduling, each subsystem in the subsystem set publishes the current resource available amount and demand list to the decision center;

[0102] The decision center publishes a task tender, and each subsystem in the subsystem set bids according to its own state, and the decision center obtains an initial allocation scheme set by using an improved contract net protocol and combining with the Nash equilibrium strategy of game theory;

[0103] S32, generating a Pareto scheme set according to the initial allocation scheme set and combining with a multi-objective optimization algorithm;

[0104] The S33 includes the following steps:

[0105] S321, constructing a target function combining comprehensive cost, device utilization rate, and service response time; the multi-objective function formula is as follows,

[0106]

[0107] wherein, represents minimizing the comprehensive cost, E i represents the energy consumption value of the i-th device, C i represents the operation and maintenance cost of the i-th resource, and a represents the total number of devices; represents maximizing the device utilization rate, U t represents the utilization rate of the t-th device, minmax(T z ) represents minimizing service delay, T k represents the response delay of the k-th service;

[0108] S322, constructing a pea population, setting the size of the pea population as v, and then the pea population is represented as o = {o1, o2,..., o i ,...,o v}, each pea in the pea population represents a treatment scheme in the initial allocation scheme set, and the maximum optimization iteration number is set;

[0109] S323, iteratively operating on the pea population; selecting and crossing peas in the pea population according to the multi-objective function to obtain an operated pea population;

[0110] S324, repeating S323, and stopping iteration when a maximum number of iterations for optimization is reached to obtain a set of Pareto solutions;

[0111] S33, setting a priority of resource allocation of a subsystem according to an importance of the subsystem in the set of subsystems to obtain a priority rule;

[0112] inputting each solution in the set of Pareto solutions into a digital twin model of the smart park to simulate and verify the solution in the set of Pareto solutions to obtain a verification result;

[0113] if the verification result is that there is a solution in the set of Pareto solutions that does not cause resource conflict, selecting an optimal solution from the solutions that do not cause resource conflict according to the priority rule, and using the optimal solution as a final solution for resource scheduling;

[0114] if the verification result is that all solutions in the set of Pareto solutions cause resource conflict, selecting a solution with the smallest conflict degree; virtually auctioning conflict resources (such as standby generator capacity) in the solution with the smallest conflict degree, each subsystem in the set of subsystems submitting a bid (including demand urgency and budget), and the highest bidder obtaining priority use right to obtain a final solution for resource scheduling.

[0115] Embodiment two:

[0116] Please refer to Figure 6 , the collaborative control and dynamic resource scheduling system for smart parks, which is used to implement the above-mentioned collaborative control and dynamic resource scheduling method for smart parks, and includes a heterogeneous sensor network and hierarchical transmission module, a digital twin and optimization prediction module, a dynamic scheduling and Pareto optimization module, and a conflict resolution and resource scheduling module;

[0117] The heterogeneous sensor network and hierarchical transmission module is used to realize multi-source data acquisition by deploying a heterogeneous sensor network; denoising and normalizing the original data, extracting key features and timestamping to generate real-time data; setting classification transmission rules based on the set of emergency event types and the set of routine monitoring data;

[0118] The digital twin and optimization prediction module is used to construct a three-dimensional model of the smart park, import device physical parameters, and establish an energy consumption-load correlation equation to form a digital twin model; update the dynamic parameters of the model through real-time data to generate a resource state matrix and physical law constraints; construct an RCN prediction network combined with the constraints, train the network using a bee colony optimization algorithm and historical data, and finally obtain an optimized RCN prediction model;

[0119] The dynamic scheduling and Pareto optimization module is used to trigger global scheduling when real-time or predicted data exceeds a threshold, each subsystem publishes resource requirements, and the decision center generates an initial allocation scheme based on the improved contract net protocol and Nash equilibrium strategy; a multi-objective function of comprehensive cost, equipment utilization and service delay is constructed, and a pea population is used to generate a Pareto solution set;

[0120] The conflict resolution and resource scheduling module is used to input the Pareto solution into the digital twin model for simulation verification, screen conflict-free solutions and select the optimal solution according to the subsystem priority; if all solutions are in conflict, the solution with the smallest conflict degree is selected, and the conflicting resources are virtually auctioned; the subsystems bid based on demand urgency and budget, and the highest bidder obtains the use right.

[0121] In the description of the present specification, the description of the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are contained in at least one embodiment or example of the invention. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0122] The preferred embodiments of the above disclosed invention are only used to help explain the invention. The preferred embodiments do not describe all the details and limit the invention to the specific embodiments described. Obviously, many modifications and changes can be made according to the content of the present specification. The present specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can well understand and utilize the invention.

Claims

1. A collaborative control and dynamic resource scheduling method for smart parks, characterized by: The following steps are involved: S1. Collect and process smart park data to obtain real-time data; set classification and transmission rules; S2. Transmit real-time data to the digital twin model of the smart park according to the classification transmission rules, generate physical law constraints after obtaining the resource matrix; and construct the RCN prediction network based on the physical law constraints; Input the real-time data into the optimized RCN smart park data prediction network obtained by combining historical smart park data with optimization algorithms to obtain the predicted data; S3. Determine whether the real-time data or predicted data exceeds the threshold. If so, trigger global scheduling and generate a Pareto solution set in conjunction with the optimization algorithm. The final scheduling plan is obtained through the digital twin model of the smart park, the Pareto solution set and the auction algorithm, and the final scheduling plan is used for resource scheduling.

2. The collaborative control and dynamic resource scheduling method for smart parks according to claim 1 is characterized in that: The S2 comprises the following steps: S21. Import the physical parameters of the equipment into the three-dimensional model of the smart park and establish constraints to obtain the digital twin model of the smart park; S22. Transmitting the real-time data to the smart park digital twin model to generate a resource state matrix; and obtaining physical law constraints of the smart park digital twin model based on the resource state matrix; S23. Based on the physical constraints of the smart park digital twin model, an RCN prediction network is constructed. Historical smart park data is collected and optimized using an optimization algorithm to improve the RCN network, resulting in an optimized RCN smart park data prediction network. S24. Input the real-time data into the optimized RCN smart park data prediction network to obtain predicted data.

3. The collaborative control and dynamic resource scheduling method for smart parks according to claim 2 is characterized in that: The S23 includes the following steps: S231, constructing an RCN prediction network, setting a learning rate, a training accuracy threshold, and a training accuracy of the RCN prediction network; S232. Use historical smart park data to train the RCN prediction network, and combine the optimization algorithm to find the learning rate of the RCN prediction network to obtain the optimal solution. The optimal solution is used as the learning rate of the RCN prediction network to obtain an optimized RCN smart park data prediction network.

4. The collaborative control and dynamic resource scheduling method for smart parks according to claim 3 is characterized in that: In S232, the optimization algorithm is combined to find the learning rate of the RCN prediction network to obtain the optimal solution, which includes the following steps: S2321. Randomly set the initial position of the bee population according to the learning rate of the RCN prediction network to obtain a bee population initial position set; set a maximum number of optimization iterations; S2322. Define a fitness function according to the training accuracy threshold and the training accuracy; S2323, performing an iterative operation on the initial position set of the bee population, and obtaining the best individual bee position in the bee population and the global best bee position according to the fitness function; S2324. Repeat S2323. When the maximum number of optimization iterations is reached, stop the iteration and take the global best bee position as the optimal solution.

5. The collaborative control and dynamic resource scheduling method for smart parks according to claim 1 is characterized in that: The S3 includes the following steps: S31. When real-time data or predicted data exceeds the dynamic scheduling trigger threshold, global scheduling is triggered and an initial allocation plan set is generated; S32, generating a Pareto solution set based on the initial allocation solution set and a multi-objective optimization algorithm; S33, setting priority rules; Input the solutions in the Pareto solution set into the smart park digital twin model for simulation verification to obtain verification results; If the verification result shows that there is a solution in the Pareto solution set that does not conflict with resources, the optimal solution is selected as the final solution based on the priority rule for resource scheduling; If the verification result shows that all the solutions in the Pareto solution set will cause resource conflicts, the solution with the smallest conflict degree is selected; a virtual auction is conducted on the conflicting resources in the solution with the smallest conflict degree to obtain the final solution, and the final solution is used for resource scheduling.

6. The collaborative control and dynamic resource scheduling method for smart parks according to claim 5 is characterized in that: The S32 includes the following steps: S321, constructing a multi-objective function; S322. Construct a pea population according to the initial allocation plan set and set a maximum number of optimization iterations; S323, performing iterative operations on the pea population; performing selection and hybridization operations on the peas in the pea population according to the multi-objective function to obtain an operated pea population; S324. Repeat S323. When the maximum number of optimization iterations is reached, stop the iteration and obtain the Pareto solution set.

7. The collaborative control and dynamic resource scheduling system for smart parks is characterized by: Used to implement the collaborative control and dynamic resource scheduling method for smart parks as described in any one of claims 1-6.

8. A storage medium, characterized in that: A program is stored thereon, and when the program is executed by the processor, it implements the collaborative control and dynamic resource scheduling method for smart parks as described in any one of claims 1-6.

Images