A park management system and method based on digital twin platform

Abstract

The present invention is applicable to the field of park management technology, and provides a park management system and method based on a digital twin platform, including: a data acquisition module; an edge computing node module; a digital twin model generation module; a real-time data update module; an intelligent simulation and prediction module; and an intelligent decision-making module. The present invention realizes comprehensive intelligent and refined control of park management, and can acquire and process multi-source data in the park in real time. It greatly reduces the delay in data processing through edge computing, and ensures the timeliness and accuracy of the data. At the same time, the digital twin model generation module creates a virtual model corresponding to the actual park based on the processed data, and ensures that the virtual model is highly consistent with the actual environment through an adaptive adjustment mechanism and real-time data update, thereby providing a reliable basis for subsequent simulation and prediction. Finally, through the intelligent decision-making module, the system can generate and implement the optimal management strategy based on the simulation results.

Description

Technical Field

[0001] The present invention belongs to the field of park management technology, and in particular relates to a park management system and method based on a digital twin platform. Background Art

[0002] With the continuous advancement of science and technology, digital twin technology has gradually become an important means of modern management. This technology can provide managers with a platform to simulate, analyze and predict actual conditions in a virtual environment by building a virtual model that is highly similar to the real world, thereby assisting them in making more informed decisions. However, in the application scenario of park management, the current digital twin technology still has some areas that need improvement.

[0003] In the process of building a digital park twin model, it is necessary to conduct in-depth feature extraction and processing of various types of collected data in order to generate a model that can accurately reflect the real-time dynamics of the park. However, due to the limitations of current data processing capabilities and the maturity of feature extraction methods, the constructed digital twin model may sometimes be difficult to fully capture all the key features of the park, which may affect the accuracy and credibility of the model to a certain extent. Summary of the invention

[0004] The purpose of the present invention is to provide a campus management system and method based on a digital twin platform, aiming to solve the problems mentioned in the background technology.

[0005] The present invention is implemented in this way: a park management system based on a digital twin platform, the system comprising:

[0006] Data collection module, used to obtain data from each subsystem of IoT devices in the park in real time;

[0007] A data processing module is used to pre-process and fuse the data to obtain fused data;

[0008] A feature extraction module is used to extract spatial features and temporal features from fused data;

[0009] An intermediate state variable generation module is used to process spatial features and temporal features through a multi-layer perceptron to generate multiple intermediate state variables;

[0010] The initial model building module is used to process the intermediate state variables through LSTM to obtain the initial model;

[0011] A digital twin model determination module is used to iterate the initial model multiple times to obtain a digital twin model;

[0012] Intelligent simulation and prediction module, used to simulate and analyze the digital twin model and predict the future state to generate management solutions;

[0013] The intelligent decision-making module is used to generate management instructions based on the management plan and implement the management instructions through IoT devices.

[0014] Preferably, the digital twin model determination module includes:

[0015] The error analysis module is used to perform error analysis between the prediction results of the initial model and the actual data in each iteration, and generate a multi-dimensional error matrix based on the analysis results;

[0016] An adaptive optimization module is used to adjust the parameters in the initial model according to the multi-dimensional error matrix to obtain dynamic parameters;

[0017] The iterative convergence judgment module is used to calculate the convergence between the current model and the target model after each iteration to obtain the error convergence speed. When the preset error convergence speed is met, the iteration is stopped to obtain the digital twin model.

[0018] Preferably, the adaptive optimization module further includes:

[0019] The parameter sensitivity analysis unit is used to perform error sensitivity analysis on each parameter in the initial model according to the errors of each dimension in the multidimensional error matrix, and obtain the sensitive value of each parameter; the calculation formula of the sensitive value is:

[0020] ,

[0021] in, Representation parameters The sensitive value of is the i-th parameter of the initial model, is the error function in the y-th dimension, is the symbol of partial derivative, is the parameter of the y-th dimension, n is the number of dimensions of the multi-error matrix, is the gradient of the error function in the y-th dimension to the i-th parameter of the initial model; where, , is the actual value, is the initial model prediction value, c is the number of samples;

[0022] The gradient adjustment unit is used to perform multi-dimensional weighted processing on the gradient of the parameters according to the sensitivity value of each parameter in each iteration to obtain dynamic parameters; the calculation formula of the dynamic parameters is:

[0023] ,

[0024] in, is the dynamic parameter of the initial model after the current x+1th iteration, is the learning rate, which is used to control the step size of parameter adjustment. is the weighting factor of the sensitive value, n is the number of dimensions of the multi-error matrix;

[0025] A step length adjustment unit, used to dynamically adjust the learning rate η according to the error convergence speed during iteration, and to increase the step length when the deviation between the error convergence speed and the preset error convergence speed increases; and to reduce the step length when the deviation between the error convergence speed and the preset error convergence speed decreases;

[0026] The constraint condition verification unit is used to verify whether the adjusted initial model parameters meet the preset constraint conditions after each parameter adjustment, and to correct the parameters if the constraint conditions are not met.

[0027] Preferably, the intelligent simulation and prediction module includes:

[0028] The environmental analysis module is used to perform multi-dimensional analysis on different data sources according to environmental parameters and obtain a multi-dimensional analysis report;

[0029] The data source environment performance scoring module is used to score different data sources based on the multi-dimensional analysis report to obtain the performance score of the data source;

[0030] The data source weight module is used to calculate and obtain the comprehensive score of the data source according to the performance score of the data source under different environmental parameters, and generate the first weight of the data source according to the comprehensive score of the data source; the comprehensive score of the data source is calculated by the following calculation formula:

[0031] ,

[0032] in, For Moment The comprehensive score of the data sources, For Moment The weight of the environmental parameters, For Time data source In the Performance ratings under environmental parameters, For Moment The correction factor of the environmental parameter, For Moment The actual value of the environmental parameter, For Moment The baseline values ​​of environmental parameters, is the total number of environmental parameters.

[0033] Preferably, the intelligent simulation and prediction module further includes a data source weight adjustment module, which is used to adjust the first weight of the data source according to the historical performance score of the data source to obtain a second weight; the calculation formula of the second weight is:

[0034] ,

[0035] in, For The second weight of the moment, For The first weight set at the moment, is the calibration factor, For Historical performance scores of moment data sources, For Benchmark score at the moment.

[0036] Preferably, the intelligent simulation and prediction module includes a parameter automatic adjustment unit, which is used to obtain simulation parameters and adjust the digital twin model according to real-time fusion data during the simulation process; the calculation formula of the simulation parameters is:

[0037] ,

[0038] in: For The simulation parameters adjusted at all times, and Respectively in The simulated energy consumption and actual energy consumption of the zth subsystem at the moment, is the feedback coefficient of the z-th subsystem, is the environmental adjustment coefficient of the zth subsystem, For The change in environmental parameters of the zth subsystem at time, v is the total number of subsystems.

[0039] Preferably, the intelligent simulation and prediction module further includes:

[0040] A feedback loop module is used to calculate the deviation amount based on the real-time data and the simulation data;

[0041] The deviation modification module is used to adjust the simulation parameters according to the deviation to obtain the simulation parameter correction amount; the calculation formula of the simulation parameter correction amount is:

[0042] ,

[0043] in, For The simulation parameter correction amount at the time; is the correction factor; For Deviation parameter at the moment; For Prediction simulation parameters at time; is the adjustment coefficient; For Additional simulation parameters at time; Append simulation parameters for the benchmark.

[0044] Preferably, the intelligent decision-making module includes:

[0045] A management strategy generating unit, used for generating a plurality of management strategies according to the management scheme;

[0046] The management instruction determination unit is used to comprehensively score the management strategy according to the management cost and decision factors to obtain the management instruction; the calculation formula of the comprehensive score is:

[0047] ,

[0048] in, is the comprehensive score, is the weight of the qth strategy, is the score of the qth strategy, For time The cost of the qth strategy at time, is the baseline cost, and w is the number of alternative strategies.

[0049] Preferably, the score of the qth strategy is The calculation formula is as follows:

[0050] ,

[0051] in, For The score of the qth strategy at time, For The weight of the e-th decision factor at time, For The score of the qth strategy at the moment under the eth decision factor, For The reference value of the e-th decision factor at time, is the benchmark reference value, and r is the number of decision factors.

[0052] The present invention also provides a park management method based on a digital twin platform, the method comprising the following steps:

[0053] Obtain data from each subsystem of IoT devices in the park in real time;

[0054] Preprocess and fuse the data to obtain fused data, and extract spatial and temporal features from it;

[0055] The spatial features and temporal features are processed by a multi-layer perceptron to generate multiple intermediate state variables;

[0056] The intermediate state variables are processed through LSTM to obtain the initial model;

[0057] Iterate the initial model multiple times to obtain the digital twin model;

[0058] Conduct simulation analysis and future state prediction on the digital twin model to generate management solutions;

[0059] According to the management plan, management instructions are generated and implemented through IoT devices.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] Through the park management system of the present invention, real-time monitoring and data management of various Internet of Things devices in the park can be achieved, thereby effectively improving the intelligent level of park management; the data acquisition module can obtain data from each subsystem of the Internet of Things device in real time to ensure the timeliness and integrity of the data; the data processing module pre-processes and fuses the collected data to form reliable fused data, thereby providing an accurate data basis for subsequent feature extraction and modeling.

[0062] The feature extraction module can effectively extract the spatial and temporal features in the fused data, which can better reflect the operating status of the equipment and environmental changes in the park; through the intermediate state variable generation module, the spatial and temporal features are further processed using a multilayer perceptron to generate multiple intermediate state variables, which provide rich input information for subsequent model construction.

[0063] The initial model construction module processes the intermediate state variables through the long short-term memory network to establish the initial model, which can capture the temporal dependencies of the data and improve the predictive ability and accuracy of the model; the digital twin model determination module iterates the initial model multiple times to ultimately obtain a more accurate and stable digital twin model, thereby providing a high-fidelity virtual mapping for park management.

[0064] The intelligent simulation and prediction module uses the digital twin model to simulate and predict future states, generating reliable management plans that help foresee possible abnormal situations and take measures in advance; the intelligent decision-making module automatically generates corresponding management instructions based on the generated management plan and implements them through IoT devices, thereby realizing automated and intelligent management of the park, reducing manual intervention, and improving management efficiency and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0066] Figure 1 A schematic diagram of the architecture of a campus management system based on a digital twin platform provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0068] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0069] like Figure 1 FIG. 1 is a schematic diagram of the architecture of a park management system based on a digital twin platform provided by an embodiment of the present invention, wherein the system includes:

[0070] Data collection module, used to obtain data from each subsystem of IoT devices in the park in real time;

[0071] A data processing module is used to pre-process and fuse the data to obtain fused data;

[0072] A feature extraction module is used to extract spatial features and temporal features from fused data;

[0073] An intermediate state variable generation module is used to process spatial features and temporal features through a multi-layer perceptron to generate multiple intermediate state variables;

[0074] The initial model building module is used to process the intermediate state variables through LSTM to obtain the initial model;

[0075] A digital twin model determination module is used to iterate the initial model multiple times to obtain a digital twin model;

[0076] Intelligent simulation and prediction module, used to simulate and analyze the digital twin model and predict the future state to generate management solutions;

[0077] The intelligent decision-making module is used to generate management instructions based on the management plan and implement the management instructions through IoT devices.

[0078] In the embodiment of the present invention, the park management system of the present invention can realize real-time monitoring and data management of various Internet of Things devices in the park, thereby effectively improving the intelligent level of park management; the data acquisition module can obtain the data of each subsystem of the Internet of Things device in real time to ensure the timeliness and integrity of the data; the data processing module pre-processes and fuses the collected data to form reliable fused data, thereby providing an accurate data basis for subsequent feature extraction and modeling; the feature extraction module can effectively extract the spatial features and time features in the fused data, which can better reflect the operating status and environmental changes of the equipment in the park; through the intermediate state variable generation module, the spatial features and time features are further processed using a multi-layer perceptron to generate multiple intermediate state variables, which provide rich data for subsequent model construction. The initial model building module processes the intermediate state variables through the long short-term memory network (LSTM) to establish the initial model, which can capture the temporal dependencies of the data and improve the prediction ability and accuracy of the model. The digital twin model determination module iterates the initial model multiple times to finally obtain a more accurate and stable digital twin model, thereby providing a high-fidelity virtual mapping for park management. The intelligent simulation and prediction module uses the digital twin model to simulate and predict future states to generate reliable management plans, which helps to foresee possible abnormal situations and take measures in advance. The intelligent decision-making module automatically generates corresponding management instructions based on the generated management plan and implements them through IoT devices, thereby realizing automated and intelligent management of the park, reducing manual intervention and improving management efficiency and safety.

[0079] Among them, the data processing module is used to preprocess and fuse the real-time data of the Internet of Things devices and subsystems in the park to improve the quality and consistency of the data. The module includes a data cleaning unit, a data standardization unit and a fusion unit. The data cleaning unit is responsible for removing abnormal data and noise data to ensure the accuracy and completeness of the data. The data standardization unit converts data from different sources into a unified format and unit, so that the data is comparable and operational in subsequent processing. The fusion unit comprehensively processes the cleaned and standardized data to form fused data that can represent the current state of the park, thereby providing high-quality input for subsequent feature extraction and model building.

[0080] The feature extraction module is used to extract spatial and temporal features from the fused data to identify the operating mode and status of the equipment and systems in the park. The module includes a spatial feature extraction unit and a temporal feature extraction unit. The spatial feature extraction unit uses methods such as convolution operations to extract spatial information such as location, distribution, and adjacent relationships in the data to describe the layout and association of IoT devices and subsystems in the park. The temporal feature extraction unit analyzes the time series of the data to identify dynamic features such as periodicity, trend, and mutation points in the operation of equipment and systems, thereby helping to establish a comprehensive understanding of the park status and provide high-value feature input for subsequent model construction.

[0081] The intermediate state variable generation module is used to process the spatial features and temporal features output from the feature extraction module through a multi-layer perceptron to generate multiple intermediate state variables. These intermediate state variables, as abstract representations of features, can capture the potential operating modes and dynamic changes of equipment and systems. This module performs multi-layer nonlinear transformations on the input features to extract complex relationships and high-dimensional information between features to improve the expressiveness and generalization capabilities of the data, thereby laying the foundation for the construction of the initial model and enabling the model to better adapt to diverse and complex campus scenarios.

[0082] The initial model construction module is used to generate the initial state model of the park based on the intermediate state variables. Specifically, the intermediate state variables are processed through the long short-term memory network (LSTM) to capture the time dependency and long-term trend between the variables. This module inputs the intermediate state variables into the LSTM network and uses its ability to process time series data to generate an initial model that can describe the operating status of the park's equipment and systems. This initial model can reflect the current operating characteristics and potential problems of the park, and provide a basis for the subsequent construction and iterative optimization of digital twin models, thereby supporting the intelligent management and decision-making optimization of the park.

[0083] As a preferred embodiment of the present invention, the digital twin model determination module includes:

[0084] The error analysis module is used to perform error analysis between the prediction results of the initial model and the actual data in each iteration, and generate a multi-dimensional error matrix based on the analysis results;

[0085] An adaptive optimization module is used to adjust the parameters in the initial model according to the multi-dimensional error matrix to obtain dynamic parameters;

[0086] The iterative convergence judgment module is used to calculate the convergence between the current model and the target model after each iteration to obtain the error convergence speed. When the preset error convergence speed is met, the iteration is stopped to obtain the digital twin model.

[0087] In an embodiment of the present invention, by introducing a digital twin model determination module, it is possible to continuously improve the accuracy and stability of the model by combining error analysis and adaptive optimization on the basis of the initial model; the error analysis module compares the predicted results of the initial model with the actual data in each iteration to generate a multi-dimensional error matrix, which can promptly discover the deviation of the model, accurately identify the source of the error, and provide a direction for subsequent optimization; the adaptive optimization module is based on the multi-dimensional error matrix, and intelligently adjusts the parameters in the model, dynamically corrects the parameters, and ensures rapid convergence and improved accuracy of the model; the iterative convergence judgment module judges the convergence between the current model and the target model, and automatically stops the iteration when the preset convergence speed is met, which not only saves computing resources, but also ensures that the final digital twin model has sufficient accuracy; through this technical solution, it is possible to effectively cope with the changing data environment in park management, ensuring that the digital twin model can always reflect the real situation and adapt to actual needs.

[0088] Among them, the error analysis module is responsible for comparing and analyzing the prediction results of the initial model with the actual data after each iteration, and calculating the error value of the model; the specific implementation method is to first compare the predicted value of the model with the actual observed value point by point, and quantify the error by calculating the difference; then, according to the errors in different dimensions (such as time error, space error, etc.), generate a multidimensional error matrix, and each dimension of the matrix represents a specific error type; the error analysis module uses the matrix to conduct a detailed analysis of each part of the model, identify the weaknesses of the model in different dimensions, and provide a reference basis for the adaptive optimization module.

[0089] Among them, the iterative convergence judgment module is used to judge whether the current model has reached the preset convergence standard after each iteration; the convergence judgment standard can be whether the absolute value of the error is lower than a certain threshold, or whether the error change amplitude in several iterations is lower than a certain set value; when the judgment result shows that the convergence standard has been met, the module will stop the iteration process, thereby reducing unnecessary computing resource consumption and ensuring that the model achieves the expected accuracy; the result of the convergence judgment directly affects the construction accuracy and convergence speed of the digital twin model.

[0090] As another preferred embodiment of the present invention, the adaptive optimization module further includes:

[0091] The parameter sensitivity analysis unit is used to perform error sensitivity analysis on each parameter in the initial model according to the errors of each dimension in the multidimensional error matrix, and obtain the sensitive value of each parameter; the calculation formula of the sensitive value is:

[0092] ,

[0093] in, Representation parameters The sensitive value of is the i-th parameter of the initial model, is the error function in the y-th dimension, is the symbol of partial derivative, is the parameter of the y-th dimension, n is the number of dimensions of the multi-error matrix, is the gradient of the error function in the y-th dimension to the i-th parameter of the initial model; where, , is the actual value, is the initial model prediction value, c is the number of samples;

[0094] The gradient adjustment unit is used to perform multi-dimensional weighted processing on the gradient of the parameters according to the sensitivity value of each parameter in each iteration to obtain dynamic parameters; the calculation formula of the dynamic parameters is:

[0095] ,

[0096] in, is the dynamic parameter of the initial model after the current x+1th iteration, is the learning rate, which is used to control the step size of parameter adjustment. is the weighting factor of the sensitive value, n is the number of dimensions of the multi-error matrix;

[0097] A step length adjustment unit, used to dynamically adjust the learning rate η according to the error convergence speed during iteration, and to increase the step length when the deviation between the error convergence speed and the preset error convergence speed increases; and to reduce the step length when the deviation between the error convergence speed and the preset error convergence speed decreases;

[0098] The constraint condition verification unit is used to verify whether the adjusted initial model parameters meet the preset constraint conditions after each parameter adjustment, and to correct the parameters if the constraint conditions are not met.

[0099] In an embodiment of the present invention, the accuracy and efficiency of model parameter optimization are greatly improved through a multidimensional error matrix and a multidimensional weighted gradient adjustment mechanism; the multidimensional error matrix generation unit subdivides the model prediction error into multiple dimensions, including time dimension, space dimension, and feature dimension, etc., so that the source of the error can be analyzed more comprehensively. Through this detailed error decomposition method, the system can identify the parameters that have the greatest impact on the total error in each dimension, avoiding the situation in which the overall error masks problems in individual dimensions in traditional methods.

[0100] In the parameter sensitivity analysis unit, the system calculates the sensitivity value of each parameter in different dimensions. , can accurately identify parameters that are more sensitive to errors. This sensitivity value is calculated by considering the error weights of each dimension. , so that during the optimization process, the adjustment of parameters can not only reduce the overall error, but also effectively improve the error performance of specific dimensions, ensuring that the optimization process can make targeted adjustments for different error dimensions, thereby improving the overall performance of the model.

[0101] Multidimensional weighted gradient adjustment unit combines dimension-sensitive values With learning rate The dynamic adjustment mechanism of the optimization process not only depends on the error gradient, but also is affected by the errors and sensitivities of each dimension. This multi-dimensional weighted gradient adjustment method enables highly sensitive parameters to be adjusted to a greater extent during the optimization process, thereby converging to the optimal solution faster. Compared with the traditional single-dimensional optimization method, the present invention achieves a more balanced and efficient parameter adjustment in multiple dimensions, significantly improving the accuracy and robustness of the digital twin model.

[0102] Through the step size adjustment unit, the present invention can dynamically adjust the learning rate after each iteration , ensuring that the step size is increased and the optimization time is shortened when the error convergence speed is too slow; when the convergence speed is too fast or the error reduction is not significant, the step size is reduced to avoid skipping the optimal solution. Such a dynamic adjustment mechanism enhances the flexibility and adaptability of the optimization process and ensures that the model can maintain the optimal convergence speed at different stages.

[0103] The constraint verification unit further ensures that after each parameter adjustment, the model parameters still meet the preset physical or business constraints, avoiding unreasonable prediction results of the model due to excessive parameter optimization. Through this series of steps, the stability, accuracy and adaptability of the digital twin model of the park management system are significantly improved, enabling the system to manage and make decisions more effectively in complex environments.

[0104] As another preferred embodiment of the present invention, the intelligent simulation and prediction module includes:

[0105] The environmental analysis module is used to perform multi-dimensional analysis on different data sources according to environmental parameters and obtain a multi-dimensional analysis report;

[0106] The data source environment performance scoring module is used to score different data sources based on the multi-dimensional analysis report to obtain the performance score of the data source;

[0107] The data source weight module is used to calculate and obtain the comprehensive score of the data source according to the performance score of the data source under different environmental parameters, and generate the first weight of the data source according to the comprehensive score of the data source; the comprehensive score of the data source is calculated by the following calculation formula:

[0108] ,

[0109] in, is the comprehensive score of the jth data source at time t, is the weight of the kth environmental parameter at time t, is the performance score of data source j under the kth environmental parameter at time t, is the correction coefficient of data source j under the kth environmental parameter at time t, is the actual value of the kth environmental parameter at time t, is the reference value of the kth environmental parameter at time t, and m is the total number of environmental parameters.

[0110] In an embodiment of the present invention, through the multi-dimensional analysis function of the intelligent simulation and prediction module, the present invention can perform environmental performance scoring and weighting processing on different data sources, thereby making more scientific use of various data sources and improving the accuracy and reliability of simulation predictions; the environmental analysis module performs multi-dimensional analysis of the data source according to different environmental parameters, and generates a detailed analysis report, which provides a solid foundation for subsequent scoring and weighting; the data source environmental performance scoring module scores the data source through the analysis report, which can quantify the performance of different data sources in various environments, and provides an objective basis for judging the quality of the data source; the data source weight module uses the scoring results to weight the data source, ensuring that in the subsequent digital twin model construction and simulation prediction, the data source with excellent performance can play a greater role, thereby significantly improving the prediction accuracy of the system.

[0111] in, For Time data source In the The performance score under each environmental parameter is calculated by the following formula:

[0112] ,

[0113] in: Score the performance of data source j under the kth environmental parameter at time t; is the actual value of the kth environmental parameter at time t; is the reference value of the kth environmental parameter at time t, that is, the expected value or the reference value within the normal range; is the correction coefficient of the kth environmental parameter at time t, which is used to adjust the sensitivity of the score to prevent the score from over- or under-reflecting the actual fluctuation of the environmental parameter;

[0114] It reflects the degree of deviation of the data source under the environmental parameters.

[0115] when equal , the performance score is zero, indicating that the data source performs normally under the environment parameters.

[0116] when Greater than , the performance score is positive, indicating that the data source performs better than the benchmark under this environment parameter.

[0117] when Less than , the performance score is negative, indicating that the data source performs poorly under this environment parameter.

[0118] This scoring method can help the system identify the best or worst data source under different environmental conditions by quantitatively evaluating the performance of the data source, providing a scientific basis for subsequent intelligent decision-making.

[0119] Among them, the environmental analysis module uses multi-dimensional data analysis technology to comprehensively analyze various environmental parameters that affect the data source of the park management system; these environmental parameters include but are not limited to temperature, humidity, equipment load, network delay, etc.; the environmental analysis module uses multi-dimensional analysis tools (such as principal component analysis, cluster analysis, etc.) to model and analyze the relationship between environmental parameters and data source performance, and generate a multi-dimensional analysis report; this report can clearly show the degree of influence of various environmental parameters on data source performance, and provide a basis for subsequent data source environmental performance scoring.

[0120] As another preferred embodiment of the present invention, the intelligent simulation and prediction module further includes a data source weight adjustment module, which is used to adjust the first weight of the data source according to the historical performance score of the data source to obtain a second weight; the calculation formula of the second weight is:

[0121] ,

[0122] in, For The second weight of the moment, For The first weight set at the moment, is the calibration factor, For Historical performance scores of moment data sources, For The benchmark score of the moment.

[0123] In an embodiment of the present invention, the intelligent simulation and prediction module can not only perform initial weighting on the data source, but also adjust the weight according to the historical performance of the data source, thereby further improving the adaptability and prediction accuracy of the system; the data source weight adjustment module dynamically calibrates the data source weight based on the historical performance score, ensuring that the system can adapt to changes in the quality of the data source in real time and always maintain the optimal weight configuration; by introducing calibration coefficients and historical performance scores, the system can adjust the weight in time when the data source performance fluctuates, avoiding the sudden inaccuracy of the data source and causing a decrease in the overall prediction accuracy; this technical solution greatly improves the stability and reliability of the digital twin model, especially in scenarios with long-term operation or frequent changes in data sources, the advantages are particularly obvious.

[0124] As another preferred embodiment of the present invention, the intelligent simulation and prediction module includes a parameter automatic adjustment unit, which is used to obtain simulation parameters and adjust the digital twin model according to real-time fusion data during the simulation process; the calculation formula of the simulation parameters is:

[0125] ,

[0126] in: For The simulation parameters adjusted at all times, and Respectively in The simulated energy consumption and actual energy consumption of the zth subsystem at the moment, is the feedback coefficient of the z-th subsystem, is the environmental adjustment coefficient of the zth subsystem, For The change in environmental parameters of the zth subsystem at time, v is the total number of subsystems.

[0127] In an embodiment of the present invention, the intelligent simulation and prediction module further improves the dynamic adaptability of the digital twin model during the simulation process by introducing an automatic parameter adjustment unit; the unit automatically adjusts the simulation parameters based on real-time fusion data to ensure that the model can reflect the actual operating conditions in real time; the automatic parameter adjustment unit uses feedback coefficients and environmental adjustment coefficients to dynamically correct the simulation parameters, so that the simulation results are closer to the actual situation, thereby improving the reliability of the management solution; by introducing real-time data into the simulation parameter adjustment, the system can cope with complex and changeable scenarios and ensure the accuracy and timeliness of the simulation prediction results.

[0128] As another preferred embodiment of the present invention, the intelligent simulation and prediction module further includes:

[0129] A feedback loop module is used to calculate the deviation amount based on the real-time data and the simulation data;

[0130] The deviation modification module is used to adjust the simulation parameters according to the deviation to obtain the simulation parameter correction amount; the calculation formula of the simulation parameter correction amount is:

[0131] ,

[0132] in, For The simulation parameter correction amount at the time; is the correction factor; For Deviation parameter at the moment; For Prediction simulation parameters at time; is the adjustment coefficient; For Additional simulation parameters at time; Append simulation parameters for the benchmark.

[0133] In the embodiment of the present invention, by introducing a feedback loop module and a deviation modification module, the accuracy and reliability of the simulation parameters can be further enhanced; the feedback loop module can quickly discover inaccuracies in the simulation results by calculating the deviation between the simulation data and the actual data in real time; the deviation modification module corrects the simulation parameters according to the deviation, and makes the adjusted simulation parameters more consistent with the actual situation through adjustment coefficients and correction coefficients, thereby significantly reducing the simulation error; the module design ensures that the system can still maintain high-precision simulation capabilities in complex environments, providing more reliable data support for subsequent decision-making.

[0134] As another preferred embodiment of the present invention, the intelligent decision-making module includes:

[0135] A management strategy generating unit, used for generating a plurality of management strategies according to the management scheme;

[0136] The management instruction determination unit is used to comprehensively score the management strategy according to the management cost and decision factors to obtain the management instruction; the calculation formula of the comprehensive score is:

[0137] ,

[0138] in, is the comprehensive score, is the weight of the qth strategy, is the score of the qth strategy, For time The cost of the qth strategy at time, is the baseline cost, and w is the number of alternative strategies.

[0139] Among them, the score of the qth strategy The calculation formula is as follows:

[0140] ,

[0141] in, For The score of the qth strategy at time, For The weight of the e-th decision factor at time, For The score of the qth strategy at the moment under the eth decision factor, For The reference value of the e-th decision factor at time, is the benchmark reference value, and r is the number of decision factors.

[0142] In the embodiment of the present invention, the intelligent decision-making module ensures that the decisions made in the park management process can take into account both cost-effectiveness and actual effects through management strategy generation and instruction screening; the management strategy generation unit generates multiple alternative management strategies based on simulation and prediction results, providing managers with multiple choices; the management instruction determination unit screens out the optimal management instructions by comprehensively scoring the alternative strategies, considering the management cost and the weight of the decision-making factors, and ensures the scientificity and economy of the decision; the comprehensive scoring formula takes into account the influencing factors of multiple dimensions, ensuring that the final management instructions not only have cost advantages, but also can achieve the best management effect; further by refining the strategy scoring, the accuracy and rationality of the final decision are improved; the comprehensive scoring formula adds consideration of different decision factors, and through the combination of the decision factor weights and the scoring results, the feasibility and effect of each management strategy can be more comprehensively evaluated; by comparing the benchmark reference value with the actual reference value, the scoring mechanism ensures the applicability and superiority of the final selected strategy in reality; this design greatly improves the scientificity of the decision, enabling the park management system to make the best decision in a complex environment.

[0143] The management strategy generation unit can be implemented by the following technologies:

[0144] Optimization algorithms: such as genetic algorithms and multi-objective optimization algorithms, used to find the optimal combination among multiple possible management strategies.

[0145] Machine learning model: By learning from historical data, it predicts the execution effects of different strategies and generates optimized management strategies.

[0146] For example, in a smart park, the management strategy generation unit generates multiple management strategies based on the management plan, including: reducing air-conditioning power to save energy, increasing lighting brightness to improve working environment comfort, optimizing power distribution to balance loads, etc. These strategies will be passed to the intelligent decision-making module for further analysis and selection.

[0147] Among them, the decision factors refer to various parameters and indicators that need to be considered in the process of generating management instructions, including but not limited to cost, efficiency, security, resource utilization, user satisfaction, etc.; in the process of screening management instructions, the system first scores each candidate management strategy, and these scores are calculated based on different decision factors. The higher the score, the better the performance under the decision factor; then, the system weights and summarizes the scores of each strategy according to the preset decision factor weights, and finally selects the management strategy with the best overall performance as the final management instruction; this multi-factor decision-making mechanism can ensure that the park management system is both economical and efficient, and safe and reliable in actual operation.

[0148] The present invention also provides a park management method based on a digital twin platform, which is applied to the above system and includes the following steps:

[0149] Obtain data from each subsystem of IoT devices in the park in real time;

[0150] Preprocess and fuse the data to obtain fused data, and extract spatial and temporal features from it;

[0151] The spatial features and temporal features are processed by a multi-layer perceptron to generate multiple intermediate state variables;

[0152] The intermediate state variables are processed through LSTM to obtain the initial model;

[0153] Iterate the initial model multiple times to obtain the digital twin model;

[0154] Conduct simulation analysis and future state prediction on the digital twin model to generate management solutions;

[0155] According to the management plan, management instructions are generated and implemented through IoT devices.

[0156] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A park management system based on a digital twin platform, characterized in that: The system comprises: Data collection module, used to obtain data from each subsystem of IoT devices in the park in real time; A data processing module is used to pre-process and fuse the data to obtain fused data; A feature extraction module is used to extract spatial features and temporal features from fused data; An intermediate state variable generation module is used to process spatial features and temporal features through a multi-layer perceptron to generate multiple intermediate state variables; The initial model building module is used to process the intermediate state variables through LSTM to obtain the initial model; A digital twin model determination module is used to iterate the initial model multiple times to obtain a digital twin model; Intelligent simulation and prediction module, used to simulate and analyze the digital twin model and predict the future state to generate management solutions; Intelligent decision-making module, used to generate management instructions according to the management plan and implement the management instructions through IoT devices; The digital twin model determination module includes: The error analysis module is used to perform error analysis between the prediction results of the initial model and the actual data in each iteration, and generate a multi-dimensional error matrix based on the analysis results; An adaptive optimization module is used to adjust the parameters in the initial model according to the multi-dimensional error matrix to obtain dynamic parameters; Iterative convergence judgment module, which is used to calculate the convergence between the current model and the target model after each iteration to obtain the error convergence speed. When the preset error convergence speed is met, the iteration is stopped to obtain the digital twin model; The adaptive optimization module also includes: The parameter sensitivity analysis unit is used to perform error sensitivity analysis on each parameter in the initial model according to the errors of each dimension in the multidimensional error matrix, and obtain the sensitive value of each parameter; the calculation formula of the sensitive value is: ， in, Representation parameters The sensitive value of is the i-th parameter of the initial model, is the error function in the y-th dimension, is the symbol of partial derivative, is the parameter of the y-th dimension, n is the number of dimensions of the multi-error matrix, is the gradient of the error function in the y-th dimension to the i-th parameter of the initial model; where, , is the actual value, is the initial model prediction value, c is the number of samples; The gradient adjustment unit is used to perform multi-dimensional weighted processing on the gradient of the parameter according to the sensitivity value of each parameter in each iteration to obtain a dynamic parameter; the calculation formula of the dynamic parameter is: ， in, is the dynamic parameter of the initial model after the current x+1th iteration, is the learning rate, which is used to control the step size of parameter adjustment. is the weighting factor of the sensitive value, n is the number of dimensions of the multi-error matrix; A step length adjustment unit, used to dynamically adjust the learning rate η according to the error convergence speed during iteration, and to increase the step length when the deviation between the error convergence speed and the preset error convergence speed increases; and to reduce the step length when the deviation between the error convergence speed and the preset error convergence speed decreases; The constraint checking unit is used to check whether the dynamic parameters of the initial model meet the preset constraint conditions after each parameter adjustment, and to correct the dynamic parameters if the constraint conditions are not met; The intelligent simulation and prediction module includes: The environmental analysis module is used to perform multi-dimensional analysis on different data sources according to environmental parameters and obtain a multi-dimensional analysis report; The data source environment performance scoring module is used to score different data sources based on the multi-dimensional analysis report and obtain the performance scores of the data sources under different environmental parameters; The data source weight module is used to calculate and obtain the comprehensive score of the data source according to the performance score of the data source under different environmental parameters, and generate the first weight of the data source according to the comprehensive score of the data source; the comprehensive score of the data source is calculated by the following calculation formula: , in, For Moment The comprehensive score of the data sources, For Moment The weight of the environmental parameters, For Time data source In the Performance ratings under environmental parameters, For Moment The correction factor of the environmental parameter, For Moment The actual value of the environmental parameter, For Moment The baseline values ​​of environmental parameters, is the total number of environmental parameters; The intelligent decision-making module includes: A management strategy generating unit, used for generating a plurality of management strategies according to the management scheme; The management instruction determination unit is used to comprehensively score the management strategy according to the management cost and decision factors to obtain the management instruction; the calculation formula of the comprehensive score is: ， in, is the comprehensive score, is the weight of the qth strategy, is the score of the qth strategy, For time The cost of the qth strategy at time, is the baseline cost, w is the number of alternative strategies; The score of the qth strategy The calculation formula is as follows: ， in, For The score of the qth strategy at time, For The weight of the e-th decision factor at time, For The score of the qth strategy at the moment under the eth decision factor, For The reference value of the e-th decision factor at time, is the benchmark reference value, and r is the number of decision factors.

2. According to the digital twin platform-based park management system of claim 1, it is characterized in that: The intelligent simulation and prediction module also includes a data source weight adjustment module, which is used to adjust the first weight of the data source according to the historical performance score of the data source to obtain a second weight; the calculation formula of the second weight is: ， in, For The second weight of the moment, For The first weight set at the moment, is the calibration factor, For Historical performance scores of moment data sources, For Benchmark score at the moment.

3. According to claim 2, a park management system based on a digital twin platform is characterized in that: The intelligent simulation and prediction module includes a parameter automatic adjustment unit, which is used to obtain simulation parameters and adjust the digital twin model according to real-time fusion data during the simulation process; the calculation formula of the simulation parameters is: ， in: For The simulation parameters adjusted at all times, and Respectively in The simulated energy consumption and actual energy consumption of the zth subsystem at the moment, is the feedback coefficient of the z-th subsystem, is the environmental adjustment coefficient of the zth subsystem, For The change in environmental parameters of the zth subsystem at time, v is the total number of subsystems.

4. According to claim 3, a park management system based on a digital twin platform is characterized in that: The intelligent simulation and prediction module also includes: A feedback loop module is used to calculate the deviation amount based on the real-time data and the simulation data; The deviation modification module is used to adjust the simulation parameters according to the deviation to obtain the simulation parameter correction amount; the calculation formula of the simulation parameter correction amount is: ， in, For The simulation parameter correction amount at the moment; is the correction factor; For Deviation parameter at the moment; For Prediction simulation parameters at time; is the adjustment coefficient; For Additional simulation parameters at time; Append simulation parameters for the benchmark.

5. A park management method based on a digital twin platform, characterized in that: Applied to the system according to any one of claims 1 to 4, the method comprises the following steps: Obtain data from each subsystem of IoT devices in the park in real time; Preprocess and fuse the data to obtain fused data, and extract spatial and temporal features from it; The spatial features and temporal features are processed by a multi-layer perceptron to generate multiple intermediate state variables; The intermediate state variables are processed through LSTM to obtain the initial model; Iterate the initial model multiple times to obtain the digital twin model; Conduct simulation analysis and future state prediction on the digital twin model to generate management solutions; According to the management plan, management instructions are generated and implemented through IoT devices.

Images