Park intelligent comprehensive service system based on Internet of Things

Abstract

The invention relates to the technical field of smart park service resource scheduling and risk control, in particular to a park smart comprehensive service system based on the Internet of Things, which comprises a state acquisition unit used for identifying the identity of a user terminal to construct a user behavior mode data set; the demand prediction unit is used for generating a service demand vector; the risk assessment unit is used for generating system stability risk factors; the resource scheduling unit is used for generating an initial optimal resource allocation vector by adopting a preset resource scheduling robust optimization model based on the service demand vector; the self-adaptive control unit is used for comparing the system stability risk factor with a preset first risk threshold value and a preset second risk threshold value, and correcting the resource scheduling robust optimization model according to a comparison result so as to output a final optimal resource allocation vector; the resource utilization rate and the overall operation efficiency of the park are effectively improved, and the default risk of core services is greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of smart park service resource scheduling and risk control technology, specifically a smart integrated service system for parks based on the Internet of Things. Background Technology

[0002] In the current operation and management of smart parks, the Internet of Things-based integrated service system is the key to improving efficiency. Its core lies in the real-time scheduling of various service resources. Traditional resource scheduling strategies mostly adopt a passive response mode, allocating resources based on service requests that have occurred. This approach generally suffers from problems such as delayed response and rigid resource configuration.

[0003] In existing technologies, risk assessment dimensions are relatively singular, typically focusing only on hard indicators such as facility load. This fails to comprehensively assess the long-term reputational risks accumulated due to continuous fluctuations in service quality, as well as the short-term impact risks caused by abnormal user behavior patterns. Furthermore, existing scheduling methods lack dynamic adaptive control mechanisms, making it difficult to effectively balance the service experience of core users with overall operating costs when faced with demand fluctuations or sudden risks, resulting in limited system stability and service quality. Therefore, how to provide a smart park service system that can accurately predict service demand, conduct multi-dimensional comprehensive risk assessments, and achieve dynamic adaptive resource scheduling on this basis is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an Internet of Things-based smart integrated service system for industrial parks. Specifically, the technical solution of this invention includes:

[0005] The status acquisition unit is used to collect multi-dimensional service status data to build a historical service status dataset, and to perform user terminal identification to build a user behavior pattern dataset.

[0006] The demand forecasting unit is used to generate service demand vectors based on historical service status datasets and using a pre-defined generative AI model.

[0007] The risk assessment unit is used to calculate the service reputation index and the system interference index based on the historical service status dataset and the user behavior pattern dataset, respectively, and to generate a system stability risk factor by combining the service reputation index and the system interference index.

[0008] The resource scheduling unit is used to generate an initial optimal resource allocation vector based on the service demand vector and using a preset resource scheduling robust optimization model.

[0009] The adaptive control unit is used to compare the system stability risk factor with the preset first risk threshold and second risk threshold, and to modify the resource scheduling robust optimization model according to the comparison result, so as to output the final optimal resource allocation vector.

[0010] Furthermore, the multi-dimensional service status data includes: facility load data, user behavior data, and environmental parameter data; the user terminal identification process is as follows: user terminals accessing the service are divided into core users and edge users, and service request data of the two types of users are classified and stored.

[0011] Furthermore, the calculation process of the service reputation index is as follows: the minimum threshold of key service indicators is set according to the preset service level agreement and the service reputation index is initialized; based on the collected actual service quality performance data, and combined with the preset ideal service target value and the maximum tolerable deviation, the service reputation index is dynamically quantified using a preset iterative update model.

[0012] Furthermore, the calculation process of the system interference index is as follows: Based on the user behavior pattern dataset, the overall behavior pattern probability distribution of the current user group is estimated; a preset benchmark behavior pattern probability distribution model trained based on the historical data of core users is obtained; a preset Kullback-Leibler divergence operator is used to calculate the difference between the overall behavior pattern probability distribution of the current user group and the benchmark behavior pattern probability distribution model, and the difference is set as the system interference index.

[0013] Furthermore, the process of generating the system stability risk factor is as follows: obtain the reciprocal of the service reputation index and obtain the system interference index; perform a weighted summation of the reciprocal of the service reputation index and the system interference index, and set the resulting summation as the system stability risk factor.

[0014] Furthermore, the process by which the adaptive control unit corrects itself based on the comparison results includes:

[0015] If the system stability risk factor is greater than the second risk threshold, then a second-level correction is performed on the resource scheduling robust optimization model;

[0016] If the system stability risk factor is greater than the first risk threshold and less than or equal to the second risk threshold, then the resource scheduling robust optimization model is modified in the first stage.

[0017] If the system stability risk factor is less than or equal to the first risk threshold, the resource scheduling robust optimization model will not be modified, and the initial optimal resource allocation vector will be used as the final optimal resource allocation vector.

[0018] Furthermore, the first-level correction is as follows: in the objective function of the resource scheduling robust optimization model, an additional cost term for edge users related to the system stability risk factor is introduced.

[0019] Furthermore, the second-level correction is: to perform the first-level correction and to add a hard upper limit constraint on the service supply for edge users to the resource scheduling robust optimization model.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. This system introduces generative AI models to predict service demand in advance, realizing the transformation from passive response to proactive supply. This proactive prediction mechanism can prepare resources in advance, significantly shorten the service request response time of core users, and effectively improve resource utilization and overall park operation efficiency.

[0022] 2. This system constructs a two-dimensional risk assessment system based on service reputation and user behavior interference. It not only quantifies the long-term chronic risks accumulated due to continuous fluctuations in service quality, but also captures the short-term impact risks caused by abnormal user behavior in real time, thus forming a more comprehensive and profound insight into system stability.

[0023] 3. This system has established a closed-loop adaptive control mechanism linked to risk assessment. By setting graded risk thresholds, the system can dynamically adjust the resource scheduling strategy from flexible guidance to rigid restriction according to the real-time risk level, achieving a delicate balance between ensuring core services and controlling costs.

[0024] 4. This system can effectively guarantee the service experience of core users and the overall stability of the system. When faced with pressure scenarios such as a surge in concurrent requests from edge users, it can dynamically adjust the resource allocation strategy for edge users to prioritize the resource supply for core services, thereby greatly reducing the risk of default for core services. Attached Figure Description

[0025] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0026] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0028] Example 1:

[0029] Please see Figure 1 A smart integrated service system for industrial parks based on the Internet of Things (IoT) includes:

[0030] The status acquisition unit is used to collect multi-dimensional service status data to build a historical service status dataset, and to perform user terminal identification to build a user behavior pattern dataset.

[0031] The demand forecasting unit is used to generate service demand vectors based on historical service status datasets and using a pre-defined generative AI model.

[0032] The risk assessment unit is used to calculate the service reputation index and the system interference index based on the historical service status dataset and the user behavior pattern dataset, respectively, and to generate a system stability risk factor by combining the service reputation index and the system interference index.

[0033] The resource scheduling unit is used to generate an initial optimal resource allocation vector based on the service demand vector and using a preset resource scheduling robust optimization model.

[0034] The adaptive control unit is used to compare the system stability risk factor with the preset first risk threshold and second risk threshold, and to modify the resource scheduling robust optimization model according to the comparison result, so as to output the final optimal resource allocation vector.

[0035] This invention provides an IoT-based smart integrated service system for parks, which aims to solve the problems of delayed response, single risk assessment dimension, and inability to effectively balance core user service experience and operating costs in the scheduling of park service resources in the prior art.

[0036] The system includes:

[0037] The purpose of the status acquisition unit is to provide an accurate and comprehensive data foundation for subsequent demand forecasting and risk assessment. This unit collects multi-dimensional service status data in real time through an IoT sensor network deployed within the park, forming the original parameter set of the service environment, and constructs a historical service status dataset. Simultaneously, this unit identifies the user terminals accessing the service, classifying them into core users and peripheral users, and categorizes and stores the service request data of the two types of users to construct a user behavior pattern dataset. ;

[0038] The demand forecasting unit aims to generate accurate predictions of future service demand based on historical data, thereby shifting from passive response to proactive supply. This unit employs a pre-defined generative AI model. Its technical concept originates from sequence generation networks in deep learning; specifically, the model can adopt a sequence-to-sequence (Seq2Seq) architecture that includes a two-layer long short-term memory (LSTM) encoder and a single-layer LSTM decoder. A fully connected layer is connected after the decoder to map the decoded hidden states to the service demand vector. The output is dimension-matched; this model is based on a historical service status dataset. The service demand vector for a future scheduling cycle is generated using the following formula. :

[0039] in, : Refers to the service demand vector predicted at time t. Each dimension of the vector corresponds to the predicted demand for different services. It is a vector with clear physical dimensions. For example, the components can correspond to computing power demand TFLOPS, bandwidth demand Mbps, etc. : Refers to the historical service status dataset up to time t-1, which is multi-dimensional time series data collected and integrated by the status acquisition unit;

[0040] z: refers to a random noise vector, introduced to simulate the inherent randomness and uncertainty of group behavior. Its statistical distribution parameters are calibrated based on the fluctuation characteristics of historical demand data. Specifically, it is assumed that the random noise vector... It follows a multidimensional Gaussian distribution with a mean of zero. Its covariance matrix is ​​determined by statistical analysis of the error vector between historical forecast demand and actual demand, so that it can accurately reflect the distribution characteristics of historical forecast error.

[0041] To ensure the accuracy of the model's predictions, the model The internal network weight parameters are determined by offline training on historical datasets of at least one full quarter of park operation. The training objective is to minimize the mean square error between historical predicted demand and actual demand with physical dimensions.

[0042] The risk assessment unit aims to quantitatively assess the potential risks faced by the system from two dimensions: service quality and user behavior, providing a basis for adaptive control decisions. This unit is based on historical service status datasets. and user behavior pattern dataset Calculate the service reputation index separately and system interference index Based on the two indices mentioned above, this unit generates a comprehensive system stability risk factor through weighted fusion. ;

[0043] The resource scheduling unit aims to allocate park resources in an optimal manner based on predicted service demand, thereby minimizing operating costs. This unit employs a pre-defined robust resource scheduling optimization model. The core idea of ​​this model is to generate an initial optimal resource allocation vector that can still meet demand in the worst-case scenario, even when there is uncertainty in the predicted demand. Its objective function and constraints are as follows:

[0044] in, : refers to The initial optimal resource allocation vector at time t is the resource unit vector obtained by optimizing the solution of this model; : Refers to the unit resource cost vector, the value of which comes from energy consumption and loss data provided by equipment suppliers or cost accounting by the finance department; : Refers to the resource service conversion matrix, which describes the amount of service that a unit of resource can provide. It is determined by calibration testing of the park's facilities. : Refers to the service demand vector predicted at time t, whose dimensions have a product term with . Consistent service dimensions are calculated by the demand forecasting unit; : Refers to the uncertainty deviation in demand forecasting, the range of which is defined by the uncertainty set U; : Refers to an uncertain set, the size and shape of which are set according to the statistical distribution of historical prediction errors. For example, it can be defined as an ellipsoid that can cover 95% of historical prediction errors.

[0045] The adaptive control unit aims to construct a closed-loop feedback mechanism that dynamically adjusts resource scheduling strategies based on risk assessment results, ensuring that the system prioritizes the stability of core services when facing different risk levels. This unit utilizes the system stability risk factors generated by the risk assessment unit. Compared with the preset first risk threshold Second risk threshold Real-time comparison is performed, and the robust optimization model for resource scheduling is dynamically corrected based on the comparison results to output the final optimal resource allocation vector;

[0046] Through the collaborative work of the above five units, this system constructs a complete closed loop from data collection, demand forecasting, risk assessment to resource scheduling and adaptive control. It not only accurately predicts service demand based on AI models and optimizes resource allocation, but more importantly, by introducing a two-dimensional risk assessment of service reputation and user behavior interference, combined with adaptive control logic, the system can dynamically identify and respond to potential risks. While ensuring overall service quality, it prioritizes the service experience of core users, greatly improving the intelligence level, operational efficiency, and system stability of the park's services. To ensure the effective execution of control commands, the system can further include a resource execution unit. This unit is responsible for receiving the final optimal resource allocation vector and parsing it into control commands for specific physical facilities within the park, such as servers and network switches, thus completing a complete closed loop from decision-making to execution.

[0047] To achieve this function, the final optimal resource allocation vector is defined as a structured data format, such as a JSON object, where each key-value pair specifies the resource type, target device ID, and allocation value. After receiving the JSON object, the resource execution unit will query the internal device driver mapping table, convert it into a specific API call instruction, and send it to the corresponding physical facility management interface through the preset control bus.

[0048] This system was validated on a simulation platform driven by historical data from park operations. Compared with traditional resource scheduling strategies based on fixed thresholds and passive responses, in terms of operational efficiency, through accurate demand forecasting and robust optimization, resource utilization was improved by approximately 20%, and overall operating costs were reduced by approximately 15%. In terms of service quality, the average response time for service requests from core users was shortened by 30%, and the service reputation index remained consistently above 0.9. In terms of system stability, under the stress scenario of a 200% instantaneous increase in concurrent requests from edge users, the default rate of the core service SLA of the traditional system exceeded 10%, while this system controlled the default rate below 1% through adaptive control, effectively ensuring the stability of the core service.

[0049] Example 2:

[0050] Multi-dimensional service status data includes: facility load data, user behavior data, and environmental parameter data; the user terminal identification process is as follows: user terminals accessing the service are divided into core users and edge users, and service request data of the two types of users are classified and stored.

[0051] This embodiment is a specific implementation of the status acquisition unit described in Embodiment 1;

[0052] In the implementation of this unit, the multi-dimensional service status data is further defined as three core data categories: facility load data, user behavior data, and environmental parameter data. Facility load data refers to a direct indicator reflecting the operational pressure on the park's infrastructure. Its function is to assess resource consumption, and its source is sensors installed on relevant equipment, such as real-time power consumption in various areas. User behavior data refers to direct data generated from user service requests. Its purpose is to quantify the intensity of demand for specific services. Its source is service management system logs, such as the frequency of security requests. Environmental parameter data refers to parameters describing the physical environment of the park. Their purpose is to provide contextual information for understanding and predicting service demand. These parameters originate from environmental sensors distributed throughout the park, such as indoor and outdoor temperature data. These three types of data together constitute a comprehensive description of the park's service status and are integrated into a historical service status dataset. ;

[0053] Meanwhile, the user terminal's identity recognition process is specifically implemented as follows: when a user terminal accesses the service, the system uses an authentication server to accurately classify it into core users and peripheral users based on its pre-registered identity information, and stores the service request data generated by the two types of users in a tagged classification.

[0054] By specifying service status data into three dimensions—facilities, users, and environment—the comprehensiveness and granularity of data collection are greatly improved, providing richer and higher-quality inputs for subsequent AI model training and prediction, thereby improving the accuracy of demand forecasting. At the same time, by dividing and classifying users into core and peripheral categories for storage, a key data foundation is laid for accurately identifying system interference caused by peripheral users and implementing differentiated service grading strategies.

[0055] Example 3:

[0056] The calculation process of the service reputation index is as follows: the minimum threshold of key service indicators is set according to the preset service level agreement and the service reputation index is initialized; based on the collected actual service quality performance data, and combined with the preset ideal service target value and the maximum tolerable deviation, the service reputation index is dynamically quantified using a preset iterative update model.

[0057] This embodiment focuses on the service reputation index within the risk assessment unit described in Embodiment 1. The concretization of the calculation process;

[0058] Service Reputation Index Its purpose is to quantify the erosion effect on users' long-term trust caused by small, continuous fluctuations in service quality, thereby capturing the chronic, long-term risks of the system; the calculation of the index includes the initialization of evaluation parameters and dynamic quantitative evaluation of service reputation.

[0059] During the parameter initialization phase of the assessment, minimum thresholds for key service indicators (KPIs) are set according to the park's Service Level Agreement (SLA), and the service reputation index is initialized. Its initial value Setting it to 1.0 means the system has perfect reputation in its initial state;

[0060] During the dynamic quantitative evaluation phase, a pre-defined iterative update model is used to track service reputation based on the collected actual service quality performance data. The technical idea of ​​this model originates from the exponential moving average method in signal processing, which can smoothly reflect changes in user-perceived service quality. The iterative calculation of this exponent is defined by the following formula:

[0061] in, : refers to the service reputation index at the current time t, which is a dimensionless value in the range [0,1], and is calculated iteratively by this formula; : Refers to the credit index at the previous moment, which is calculated from the previous period; : Refers to the update rate, which is a constant in the range (0,1) used to balance the impact of historical reputation and current service performance; : Refers to the importance weight of the i-th type of service, which is a dimensionless value pre-set by the park manager based on the importance of the service, and satisfies ; : Refers to the actual quality performance of the i-th type of service at time t, which is data with specific physical dimensions collected in real time by the IoT sensors of the state acquisition unit; : Refers to the ideal target value for the i-th type of service, which is set according to the SLA or best practices and is related to Values ​​of the same dimension; : Refers to the maximum tolerable deviation of the i-th type of service, which is based on the SLA and... Values ​​of the same dimension;

[0062] This formula quantifies each deviation in service quality using a normalized penalty term, when the actual service quality... Deviation from ideal value At that time, the penalty will increase according to the degree of deviation, leading to decline;

[0063] By introducing this iteratively updated quantitative model, the abstract service reputation is transformed into a calculable and traceable dynamic indicator. This method not only considers the immediate performance of service quality, but also reflects the historical cumulative effect through exponential moving average. It can more realistically and sensitively reflect the long-term risks caused by the continuous small decline in service quality, and provide a more in-depth dimension for the assessment of system stability.

[0064] Example 4:

[0065] The calculation process of the system interference index is as follows: Based on the user behavior pattern dataset, estimate the overall behavior pattern probability distribution of the current user group; obtain the preset baseline behavior pattern probability distribution model trained based on the historical data of core users; use the preset Kullback-Leibler divergence operator to calculate the difference between the overall behavior pattern probability distribution of the current user group and the baseline behavior pattern probability distribution model, and set the difference as the system interference index.

[0066] This embodiment focuses on the system interference index in the risk assessment unit described in Embodiment 1. The concretization of the calculation process;

[0067] System Interference Index Its purpose is to quantify system state deviations caused by users with abnormal behavioral patterns, thereby capturing the acute, short-term shock risks to the system; to achieve this purpose, it is based on user behavior pattern datasets. By using statistical methods such as kernel density estimation, the probability distribution of the overall behavioral patterns of the current user group is estimated. Simultaneously, a pre-defined baseline behavioral pattern probability distribution model, trained offline based on massive amounts of core user historical data, is obtained from the system. This model represents the typical service usage patterns within the park; it uses a pre-defined Kullback-Leibler divergence operator to calculate the probability distribution of the overall behavioral patterns of the current user group. Probability distribution model of baseline behavior patterns The difference between them is set as the system interference index. The calculation formula is as follows:

[0068] in, : Refers to the system disturbance index at time t, which is a non-negative dimensionless value calculated by this formula; This refers to the probability distribution of the overall behavioral patterns of the current user group, based on a user behavior pattern dataset. The probability distribution function estimated in real time; : Refers to the baseline behavior pattern probability distribution model, which is a probability distribution function obtained by offline training on historical big data of core users; : Refers to the Kullback-Leibler divergence operator, used to measure the difference between two probability distributions;

[0069] When a large number of peripheral users with behavioral patterns that are very different from those of the core users flood in... It will deviate significantly This leads to the KL divergence value Increased rapidly;

[0070] This embodiment creatively introduces KL divergence from information theory to quantify anomalies in user behavior patterns. Compared with traditional rule-based or threshold-based anomaly detection, this method can capture deviations from the overall distribution level, making it more comprehensive and robust. It enables the system to quantify the short-term impact risk caused by unconventional user behavior in real time and accurately, providing timely and reliable trigger signals for the system to activate emergency plans and implement service grading.

[0071] Example 5:

[0072] The process of generating the system stability risk factor is as follows: obtain the reciprocal of the service reputation index and the system interference index; perform a weighted summation of the reciprocal of the service reputation index and the system interference index, and set the sum as the system stability risk factor.

[0073] This embodiment focuses on the system stability risk factor in the risk assessment unit described in Embodiment 1. The concretization of the generation process;

[0074] System stability risk factors Its purpose is to reflect the service credit index, which reflects long-term risk. and the system disturbance index reflecting short-term shocks Effective integration is achieved to form a unified and comprehensive system risk measurement index; its calculation logic lies in obtaining the service reputation index calculated by the implementation method of Example 3. reciprocal The reciprocal operation aims to transform a decrease in credibility into an increase in risk, aligning it with the meaning of risk indicators; the system interference index calculated by the implementation method in Example 4 is obtained. The reciprocal of the service credibility index With system interference index Perform a weighted summation and set the resulting summation as the system stability risk factor. The calculation formula is as follows:

[0075] in, : Refers to the system stability risk factor at time t, which is a dimensionless value calculated by this formula; : Refers to the service reputation index at time t, which is calculated by the implementation method of Example 3; : refers to the system interference index at time t, which is calculated by the implementation method of Example 4; This refers to dimensionless weighting coefficients, whose values ​​are calibrated on a simulation platform through regression analysis of historical system failure scenarios. The goal is to ensure that... The changing trend has the highest correlation with the actual system failure probability;

[0076] This embodiment unifies the endogeneous chronic risks caused by service quality degradation and the exogenous acute risks caused by abnormal user behavior into a single evaluation framework through a concise and efficient weighted summation model. This allows system administrators to assess risks using a single indicator. A comprehensive understanding of the overall risk level currently faced by the system greatly simplifies the decision-making process and provides a scientific and unified quantitative basis for the subsequent implementation of precise and tiered response strategies by the adaptive control unit.

[0077] Example 6:

[0078] The process by which the adaptive control unit corrects itself based on the comparison results includes:

[0079] If the system stability risk factor is greater than the second risk threshold, then a second-level correction is performed on the resource scheduling robust optimization model;

[0080] If the system stability risk factor is greater than the first risk threshold and less than or equal to the second risk threshold, then the resource scheduling robust optimization model is modified in the first stage.

[0081] If the system stability risk factor is less than or equal to the first risk threshold, the resource scheduling robust optimization model will not be modified, and the initial optimal resource allocation vector will be used as the final optimal resource allocation vector.

[0082] The first-level correction is: to introduce an additional cost term for edge users that is related to the system stability risk factor into the objective function of the resource scheduling robust optimization model;

[0083] The second-level correction is to execute the first-level correction and add a hard upper limit constraint on the service supply for edge users to the resource scheduling robust optimization model.

[0084] This embodiment is a concretization of the adaptive control unit and its correction process described in Embodiment 1; this unit is based on the system stability risk factor generated in Embodiment 5. The value, compared with the preset first risk threshold. Second risk threshold The comparison automatically triggers a hierarchical dynamic correction of the robust optimization model for resource scheduling; the threshold and The technical principle behind this design is to combine risk quantification management with operational objectives. As a high-risk boundary, it can be statistically defined based on the risk factor values ​​that led to explicit defaults on Service Level Agreements (SLAs) in historical data, for example, by taking the 95th percentile of the distribution of such risk factors; As an early warning line, it can lead to a service reputation index based on risk factors. A threshold value is set for the inflection point where a continuous downward trend emerges, thereby ensuring the timeliness and foresight of the correction strategy;

[0085] If the system stability risk factor Less than or equal to the first risk threshold This indicates that the system is in a safe and stable state. At this point, the robust optimization model for resource scheduling is not modified, and the initial optimal resource allocation vector output by the resource scheduling unit is directly applied. The final optimal resource allocation vector will be output.

[0086] If the system stability risk factor Greater than the first risk threshold And less than or equal to the second risk threshold This indicates that the system has entered an early warning state, at which point a first-level correction is performed on the resource scheduling robust optimization model. The first-level correction involves introducing a risk factor related to system stability into the objective function of the resource scheduling robust optimization model. The revised objective function for the related additional costs for edge users becomes:

[0087] in, : Refers to the edge user resource mask vector, which is a... Two vectors of the same dimension, if If a certain component is a resource allocated to edge users, then... The corresponding component is 1, otherwise it is 0; : Refers to the Hadamard product operator, which represents the element-wise multiplication of vectors; : Refers to the risk penalty function, which is a function relating to risk factors Monotonically increasing functions, for example Its output is a dimensionless penalty coefficient;

[0088] This correction has been passed. The function artificially increases the nominal cost of allocating resources to peripheral users, so the optimization algorithm will prioritize using limited resources to ensure the service experience of core users.

[0089] If the system stability risk factor Greater than the second risk threshold This indicates that the system has entered an emergency state. At this time, a second-level correction is performed on the resource scheduling robust optimization model. The second-level correction is as follows: based on the first-level correction, a hard service supply ceiling constraint for edge users is further added to the resource scheduling robust optimization model. The new constraint is:

[0090] in, : Refers to the maximum service supply capacity vector of the park, which is a constant vector representing the physical limit that the park can provide for various services; : Refers to the resource quota function, which is a function related to risk factors The monotonically decreasing function, whose output is a dimensionless resource quota ratio in the range (0,1);

[0091] This constraint is passed. Function, based on risk factors The size of the limit dynamically and forcibly sets a constantly shrinking hard cap on the total amount of service allocated to edge users;

[0092] This embodiment establishes a clear, orderly, and progressively advancing closed-loop control strategy. The first-level correction, as a soft adjustment, guides resources towards core users through economic levers, achieving a smooth adjustment of service priorities. The second-level correction is a hard constraint based on this, ensuring the survivability of core services under extreme risks by imposing service caps. This strategy of combining soft and hard measures and responding in stages achieves a delicate balance between service quality and operating costs under different risk levels, maximizing the service experience for core users and the overall robustness of the system.

[0093] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart integrated service system for industrial parks based on the Internet of Things, characterized in that: include: The status acquisition unit is used to collect multi-dimensional service status data to build a historical service status dataset, and to perform user terminal identification to build a user behavior pattern dataset. The demand forecasting unit is used to generate service demand vectors based on historical service status datasets and using a pre-defined generative AI model. The risk assessment unit is used to calculate the service reputation index and the system interference index based on the historical service status dataset and the user behavior pattern dataset, respectively, and to generate a system stability risk factor by combining the service reputation index and the system interference index. The resource scheduling unit is used to generate an initial optimal resource allocation vector based on the service demand vector and using a preset resource scheduling robust optimization model. The adaptive control unit is used to compare the system stability risk factor with the preset first risk threshold and second risk threshold, and to modify the resource scheduling robust optimization model according to the comparison result, so as to output the final optimal resource allocation vector.

2. The IoT-based smart integrated service system for industrial parks according to claim 1, characterized in that, The multi-dimensional service status data includes: facility load data, user behavior data, and environmental parameter data; the user terminal identification process is as follows: the user terminals accessing the service are divided into core users and edge users, and the service request data of the two types of users are classified and stored.

3. The IoT-based smart integrated service system for industrial parks according to claim 1, characterized in that, The calculation process of the service reputation index is as follows: the minimum threshold of key service indicators is set according to the preset service level agreement and the service reputation index is initialized; based on the collected actual service quality performance data, and combined with the preset ideal service target value and the maximum tolerable deviation, the service reputation index is dynamically quantified using a preset iterative update model.

4. The IoT-based smart integrated service system for industrial parks according to claim 1, characterized in that, The calculation process of the system interference index is as follows: Based on the user behavior pattern dataset, estimate the overall behavior pattern probability distribution of the current user group; obtain a preset benchmark behavior pattern probability distribution model trained based on the historical data of core users; use the preset Kullback-Leibler divergence operator to calculate the difference between the overall behavior pattern probability distribution of the current user group and the benchmark behavior pattern probability distribution model, and set the difference as the system interference index.

5. The IoT-based smart integrated service system for industrial parks according to claim 1, characterized in that, The process of generating the system stability risk factor is as follows: obtain the reciprocal of the service reputation index and obtain the system interference index; perform a weighted summation of the reciprocal of the service reputation index and the system interference index, and set the sum as the system stability risk factor.

6. The IoT-based smart integrated service system for industrial parks according to claim 1, characterized in that, The process by which the adaptive control unit corrects itself based on the comparison results includes: If the system stability risk factor is greater than the second risk threshold, then a second-level correction is performed on the resource scheduling robust optimization model; If the system stability risk factor is greater than the first risk threshold and less than or equal to the second risk threshold, then the resource scheduling robust optimization model is modified in the first stage. If the system stability risk factor is less than or equal to the first risk threshold, the resource scheduling robust optimization model will not be modified, and the initial optimal resource allocation vector will be used as the final optimal resource allocation vector.

7. The IoT-based smart integrated service system for industrial parks according to claim 6, characterized in that, The first-level correction is to introduce an additional cost term for edge users that is related to the system stability risk factor into the objective function of the resource scheduling robust optimization model.

8. The IoT-based smart integrated service system for industrial parks according to claim 6, characterized in that, The second-level correction is to perform the first-level correction and add a hard upper limit constraint on the service supply for edge users to the resource scheduling robust optimization model.

Images