Internet of Things equipment resource scheduling system based on multi-source data fusion

Abstract

The invention discloses an Internet of Things equipment resource scheduling system based on multi-source data fusion, and relates to the technical field of Internet of Things equipment. The resource scheduling platform is in communication connection with a data acquisition module, a data preprocessing module, a data fusion module, a resource management module, a resource scheduling module and a decision support module. According to the Internet of Things equipment resource scheduling system based on multi-source data fusion, multi-source data in Internet of Things equipment is widely collected through the data acquisition module, and the collected data is subjected to comprehensive cleaning, standardization and denoising processing by using the data preprocessing module, so that invalid, wrong and noise parts in the data are effectively eliminated; and the data fusion module adopts a particle swarm optimization algorithm to fuse multi-source data, so that the accuracy and robustness of a data set are further improved, the efficiency and quality of data analysis are remarkably improved, and a solid data basis is provided for subsequent resource scheduling and optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things devices, and in particular to an Internet of Things device resource scheduling system based on multi-source data fusion. Background Art

[0002] With the rapid development of information technology and network technology, the speed of data generation has shown explosive growth. There are more and more data sources in different fields, industries and regions. The rapid development of Internet of Things technology has enabled more and more devices to access the network and achieve interconnection. Inevitably, a large amount of real-time data will be generated during the operation of the equipment. In the resource scheduling process of Internet of Things devices, the data needs to be effectively processed to achieve effective resource management and scheduling.

[0003] Existing resource scheduling systems often rely on a single data source and are difficult to adapt to the complex and changeable IoT environment. In addition, the large amount of data generated by IoT devices will cause response delays to a certain extent. Therefore, how to process multi-source data in the IoT resource scheduling process and realize effective management and resource scheduling of IoT devices to optimize resource utilization and improve system efficiency is the problem we need to solve. To this end, this paper proposes an IoT device resource scheduling system based on multi-source data fusion. Summary of the invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: an IoT device resource scheduling system based on multi-source data fusion, comprising a resource scheduling platform, wherein the resource scheduling platform is communicatively connected with a data acquisition module, a data preprocessing module, a data fusion module, a resource management module, a resource scheduling module and a decision support module, wherein the modules are connected by electrical signals;

[0005] The data acquisition module is used to collect multi-source data of sensor data and device status information from IoT devices to ensure the diversity and comprehensiveness of the data and provide raw data support for subsequent data fusion and resource scheduling. The multi-source raw data includes real-time data and historical data;

[0006] The data preprocessing module is used to perform preprocessing operations such as cleaning, standardization and denoising on the collected data, including removing invalid or erroneous data points, standardizing data to have a uniform dimension and range, and applying filters to remove noise, thereby improving the accuracy and reliability of the data and reducing analysis errors caused by data quality issues;

[0007] The data fusion module uses a particle swarm optimization algorithm to fuse multi-source data and determine the optimal weights and parameters for data fusion to improve the accuracy and robustness of the overall data set. By fusing multi-source data and integrating the advantages of each data source, the comprehensiveness and accuracy of data analysis are improved, providing strong support for resource scheduling.

[0008] The resource management module predicts the time series changes and complex resource requirements of equipment resource requirements based on the time series prediction model and the machine learning prediction model, wherein the time series prediction model analyzes historical data to predict the periodic changes of future resource requirements, and the machine learning prediction model is based on the historical data training model to predict the nonlinear changes of equipment resource requirements and usage, accurately predict resource requirements, provide a scientific basis for resource scheduling, and avoid resource waste or shortage;

[0009] The resource scheduling module uses the queuing theory model to design resource scheduling strategies based on the prediction results, and applies the particle swarm optimization algorithm to optimize the queuing model parameters to achieve optimal resource allocation, reduce resource waste, and improve the overall operation efficiency of the system;

[0010] The decision support module is used to provide a user interface to display prediction results, scheduling strategies and system status, assist users in making decisions, and help users quickly understand the system status and resource scheduling through intuitive display and convenient operation interface, so as to make more scientific and reasonable decisions.

[0011] Preferably, in the data acquisition module, the process of collecting multi-source raw data includes:

[0012] Deploy various sensors and different types of IoT devices in the IoT environment to cover the areas that need to be monitored and managed, and configure corresponding communication interfaces for IoT devices. Sensors include temperature sensors, pressure sensors, and flow sensors.

[0013] Collect multi-source raw data from sensors and IoT devices, including sensor data of temperature, pressure, and flow, and IoT device status information of CPU usage and memory usage, wherein the multi-source raw data includes real-time data and historical data;

[0014] Perform preliminary verification on the collected data to ensure the integrity and validity of the data. Mark and preliminarily process the abnormal data to distinguish between normal and abnormal sensor data and IoT device status information so that the subsequent data preprocessing module can further process it.

[0015] The marked data is classified and labeled to facilitate data fusion and subsequent processing, and the data is temporarily stored in the data warehouse.

[0016] Preferably, in the data preprocessing module, the data preprocessing process includes:

[0017] Call the data in the data warehouse, analyze the integrity and continuity of the data set, determine the location of missing data, and fill the missing data points. Fill the missing values ​​through mean filling, median filling, mode filling and interpolation. Mean filling: use the average value of the corresponding data set or data series to fill. Median filling: use the median as the filling value, which is suitable for data sets with many outliers. Mode filling: use the value with the most occurrences to fill. Interpolation method: estimate the missing values ​​based on the internal relationship of the data, such as the previous and next values ​​of the time series;

[0018] Identify data values ​​that are significantly different from other data points as outliers, delete the identified outliers, mark them in the log, record the encountered outlier data values ​​for special processing, and identify and remove duplicate data to ensure data uniqueness;

[0019] The Z-score standardization method is used to convert data of different dimensions and ranges into a unified format, so that the data can be analyzed and processed under a unified framework. The expression of the Z-score standardization method is: x is the original data value, μ is the mean of the data, and τ is the standard deviation of the data;

[0020] Select the filter type according to the data characteristics and analysis requirements, apply different types of filters to smooth and denoise the data, adjust the filter parameters to optimize the denoising effect, and remove short-term fluctuations and noise by smoothing the data. The filters are low-pass filtering, high-pass filtering, and band-pass filtering. Low-pass filtering: used to remove high-frequency noise. For time series data, short-term fluctuations and noise can be removed by smoothing the data. High-pass filtering: In contrast to low-pass filtering, high-pass filtering is used to remove low-frequency noise and retain the high-frequency changes in the data. Band-pass filtering: selects data within a specific frequency range and removes noise at other frequencies.

[0021] Preferably, in the data fusion module, the process of fusing multi-source data includes:

[0022] Define the problem space, analyze the pre-processed multi-source data, and determine the types of data sources and data types in the Internet of Things;

[0023] Initialize the particle swarm and randomly generate a set of particles. Each particle represents a potential data fusion solution. The position of each particle represents a specific set of data fusion parameters, such as the weight of different data sources, time window size, fusion rules, etc.

[0024] Design a fitness function to evaluate the quality of each particle solution. The fitness function is based on the performance of the fused data. The fitness function includes prediction accuracy function, data consistency function and information gain function. Prediction accuracy function: if data fusion is used for prediction, the accuracy can be used as a measure of fitness. Data consistency function: evaluates the consistency and reliability of the fused data. Information gain function: measures the amount of information added by the fused data relative to a single data source.

[0025] Calculate the fitness value for each particle, update the individual optimal solution, and determine the global optimal solution, which is the solution with the highest fitness among all particles;

[0026] According to the current particle position, speed, individual optimal solution and global optimal solution, the particle swarm optimization algorithm is used to update the formula to adjust the speed and position of each particle, adjust the inertia weight w, cognitive factor c1, social factor c2, and introduce speed limits to prevent particles from flying out of the search space;

[0027] A maximum number of iterations or fitness threshold is set. When the number of iterations is reached or the fitness of a particle is no longer significantly improved, the iteration is stopped. After the iteration is completed, the weights and parameters corresponding to the global optimal particle are finally output as the best solution for data fusion.

[0028] Preferably, the multi-source data particle swarm algorithm is as follows:

[0029] Each particle represents a potential solution, and the particle's position vector x is determined i , whose position vector x i Contains: Data source weight: w1, w2, ...w n , where n is the number of data sources. The standard particle swarm update formula is used to update the speed and position of the particles. The influence of data update time is added to the fitness function. When the data update time difference is large, a penalty term is introduced to reduce the dependence on outdated data. The weight is adjusted by weighted average or other methods combined with the timeliness of the data;

[0030] Fitness function f(x i ) can be defined as the negative value of the data fusion error, and the formula is:

[0031]

[0032] Among them, E is the error function, y is the reference data, It is the data obtained by fusion based on the current particle parameters;

[0033] The particle swarm optimization algorithm update formula is used to adjust the speed and position of each particle. The expression of the particle swarm optimization algorithm update formula is:

[0034]

[0035]

[0036] in, and are the velocity and position of particle i at the tth iteration in dimension d, and are the velocity and position of particle i at the t+1th iteration in dimension d, is the individual optimal position of particle i in dimension d, is the global optimal position, r1 and r2 are random numbers between [0,1], w is the inertia weight, c1 is the cognitive factor, and c2 is the social factor.

[0037] Preferably, in the resource management module, the process of predicting the time series changes of device resource requirements and complex resource requirements includes:

[0038] Use the preprocessed historical data to train the ARIMA model and determine the parameters (p, d, q) of the ARIMA model, where: the ARIMA model is an autoregressive integrated moving average model, p is the number of autoregressive terms, d is the number of differences, which is used to make the non-stationary series stationary, and q is the number of moving average terms;

[0039] The particle swarm optimization algorithm is used to find the optimal ARIMA parameters (p, d, q) to minimize the prediction error of the model;

[0040] The trained ARIMA model is used to predict the periodic changes in future resource demand. The prediction results reflect the seasonal or periodic trend of resource demand. The prediction formula of the ARIMA model is:

[0041] in, is the predicted value at the future time point t, c is a constant term, is the autoregressive coefficient, is the d-order difference, θ k is the moving average coefficient, ∈ t-k is the forecast error at a past time point;

[0042] Extract the features of temperature, pressure, flow, CPU usage, and memory usage from the preprocessed historical data, and build a composite prediction model by combining decision tree and deep learning models. The deep learning model can handle more complex nonlinear relationships, while the decision tree model provides better interpretability. The model parameters are adjusted through the cross-validation method to optimize the prediction performance.

[0043] Based on the constructed composite prediction model and the extracted sensor data features, the environmental interference index is analyzed to evaluate the deviation of sensor data from the normal state. The composite prediction model and the extracted CPU usage and memory usage features are used to obtain the device utilization index and resource load index. The device resource prediction index is obtained through comprehensive analysis to evaluate the device's demand for resources.

[0044] Use the trained composite prediction model to predict nonlinear changes in device resource demand and usage, and combine the prediction results to capture sudden changes and non-periodic fluctuations in resource demand;

[0045] Combine the prediction results of the time series prediction model and the machine learning prediction model to form a comprehensive prediction of device resource requirements.

[0046] Preferably, the expression of the environmental interference index is:

[0047]

[0048] Where SFI is the environmental interference index, T h is the observation value of the hth sensor, T ref is the standard value of the corresponding sensor, w h is the weight of the hth sensor data, n is the total number of sensors, and ε is a very small constant used to ensure that the denominator is not zero. The smaller the SFI value, the higher the prediction index. The smaller the SFI, the closer the sensor reading is to the ideal state, indicating that the IoT device is less affected by temperature, pressure, and flow.

[0049] Preferably, the device resource prediction index is obtained by weighting the device utilization index and resource load index based on CPU utilization and memory utilization, and its expression is:

[0050] RFI=w U ·RFI U +w L ·RFI L ;

[0051]

[0052] Among them, RFI is the equipment resource forecast index, RFI U is the equipment utilization index, w U is the weight of the equipment utilization index, RFI L is the resource load index, w L is the weight of the resource load index, w U +w L =1,U r is the utilization of the rth resource (CPU or memory), U max and Umin are the minimum and maximum resource utilization, m is the number of resource types, L r is the load value of the rth resource, L avg is the average resource load value, L max It is the maximum value of the resource load value, reflecting the degree of fluctuation of the resource load. A smaller RFI value indicates that the equipment resources are in a relatively ideal state, that is, the resource utilization efficiency is high and stable. A larger RFI value may indicate that there are problems in resource utilization, such as over-utilization or large load fluctuations, which require attention and optimization.

[0053] Preferably, in the resource scheduling module, the process of designing a resource scheduling strategy using a queuing theory model includes:

[0054] The resource scheduling module receives the prediction results from the time series prediction model and the machine learning prediction model in the resource management module. The prediction results include the time series changes, periodic changes, and nonlinear and complex resource demand predictions of the equipment resource requirements;

[0055] According to the prediction results, the resource scheduling module uses the queuing theory model to simulate the resource request and allocation process, applies the particle swarm optimization algorithm to optimize the parameters of the queuing model, finds the optimal solution, and designs the fitness function based on multiple indicators such as resource scheduling efficiency, response time, and resource utilization;

[0056] In the particle swarm optimization algorithm, each particle searches for the optimal solution by iteratively updating its position and velocity. The particle position and velocity are updated based on the information of the individual optimal solution and the global optimal solution. By continuously adjusting the particle position and velocity, the particle swarm optimization algorithm finds the queuing model parameters that make the fitness function value optimal.

[0057] According to the optimized queuing model parameters, the resource scheduling module formulates a specific resource scheduling strategy and adjusts the resource allocation order, allocation amount, and priority to ensure that resource utilization and overall system operation efficiency are maximized while meeting resource requirements;

[0058] The resource scheduling module applies the formulated scheduling strategy to the actual IoT device resource scheduling process and adjusts resource allocation in real time to achieve optimal resource utilization and system performance.

[0059] Preferably, in the decision support module, the process of displaying the prediction results, scheduling strategy and system status includes:

[0060] Integrate forecast results, scheduling strategies, and system status data from the resource management module and resource scheduling module, and use charts and graphs to display time series forecasts, resource demand changes, and resource utilization indicators;

[0061] Clearly display the current resource scheduling strategy, including resource allocation order, allocation amount, priority, and real-time monitoring and display of system status, including resource usage, system load, and response time;

[0062] Based on the monitoring results of the system status, it identifies abnormal situations that occur during system operation, such as equipment failure, insufficient resources, etc., issues alarm information, and provides corresponding solutions and suggestions to help users respond to and solve problems quickly.

[0063] The present invention provides an Internet of Things device resource scheduling system based on multi-source data fusion. It has the following beneficial effects:

[0064] 1. The IoT device resource scheduling system based on multi-source data fusion widely collects multi-source data from IoT devices through the data acquisition module, and uses the data preprocessing module to comprehensively clean, standardize and denoise the collected data, effectively eliminating the invalid, erroneous and noisy parts in the data, ensuring the accuracy and reliability of the data. Subsequently, the data fusion module uses the particle swarm optimization algorithm to fuse multi-source data, further improving the accuracy and robustness of the data set, significantly improving the efficiency and quality of data analysis, and providing a solid data foundation for subsequent resource scheduling and optimization.

[0065] 2. The IoT device resource scheduling system based on multi-source data fusion uses the queuing theory model to simulate the resource request and allocation process based on the prediction results of the time series prediction model and the machine learning prediction model provided by the resource management module. By applying the particle swarm optimization algorithm to optimize the parameters of the queuing model, the system can adjust the resource allocation strategy in real time to ensure that resources are optimally configured, effectively avoid over-allocation or under-allocation of resources, significantly optimize resource utilization, reduce resource waste, and improve the overall economic benefits of the system.

[0066] 3. The IoT device resource scheduling system based on multi-source data fusion adopts multi-source data fusion technology and can process various types of data from different IoT devices, so that the system can adapt to the complex and changing IoT environment and meet the resource scheduling needs in different scenarios. At the same time, the system also has self-learning and self-adaptation capabilities, and can continuously optimize the prediction model and scheduling strategy based on historical data and real-time feedback. This dynamic adjustment and optimization mechanism enables the system to maintain high efficiency and stability, providing users with more reliable and high-quality services. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a structural module diagram of an Internet of Things device resource scheduling system based on multi-source data fusion according to the present invention;

[0068] Figure 2A flow chart of fusing multi-source data according to the present invention;

[0069] Figure 3 A flow chart for predicting time series changes in device resource requirements and complex resource requirements for the present invention. DETAILED DESCRIPTION

[0070] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0071] The first embodiment, as Figure 1 , Figure 2 As shown, the present invention provides a technical solution: an Internet of Things device resource scheduling system based on multi-source data fusion, including a resource scheduling platform, the resource scheduling platform is communicatively connected with a data acquisition module, a data preprocessing module, a data fusion module, a resource management module, a resource scheduling module and a decision support module, wherein the modules are connected by electrical signals;

[0072] The data acquisition module is used to collect multi-source data of sensor data and device status information from IoT devices to ensure the diversity and comprehensiveness of the data and provide original data support for subsequent data fusion and resource scheduling. The multi-source original data includes real-time data and historical data. Various sensors and different types of IoT devices are deployed in the IoT environment to cover the areas that need to be monitored and managed, and the corresponding communication interfaces are configured for the IoT devices. Among them, the sensors include temperature sensors, pressure sensors, and flow sensors. Multi-source original data are collected from sensors and IoT devices, including sensor data of temperature, pressure, and flow, as well as IoT device status information of CPU usage and memory usage. Among them, the multi-source original data includes real-time data and historical data. The collected data is preliminarily verified to ensure the integrity and validity of the data. For abnormal data, it is marked and preliminarily processed to distinguish normal and abnormal sensor data and IoT device status information so that the subsequent data preprocessing module can further process it. The marked data is classified and labeled to facilitate data fusion and subsequent processing, and the data is temporarily stored in the data warehouse.

[0073] The data preprocessing module is used to perform preprocessing operations such as cleaning, standardization and denoising on the collected data, including removing invalid or erroneous data points, standardizing data to make it have a unified dimension and range, and applying filters to remove noise, improve data accuracy and reliability, reduce analysis errors caused by data quality problems, call data in the data warehouse, analyze the integrity and continuity of the data set, determine the location of missing data, and fill in the missing data points. Missing values ​​are filled by mean filling, median filling, mode filling and interpolation. Mean filling: use the average value of the corresponding data set or data series to fill, median filling: use the median as the filling Value, suitable for data sets with many outliers, mode filling: use the value with the most occurrences to fill, interpolation method: estimate missing values ​​based on the intrinsic relationship of the data, such as the previous and next values ​​of the time series, identify data values ​​that are significantly different from other data points as outliers, delete the identified outliers, mark them in the log, record the encountered abnormal data values, so as to perform special processing, and identify and remove duplicate data to ensure the uniqueness of the data, use Z-score standardization to convert data of different dimensions and ranges into a unified format, so that the data can be analyzed and processed under a unified framework, where the expression of Z-score standardization is: x is the original data value, μ is the mean of the data, σ is the standard deviation of the data, select the filter type according to the data characteristics and analysis requirements, apply different types of filters to smooth and denoise the data, adjust the filter parameters to optimize the denoising effect, and remove short-term fluctuations and noise by smoothing the data. The filters are low-pass filtering, high-pass filtering, and band-pass filtering. Low-pass filtering: used to remove high-frequency noise. For time series data, short-term fluctuations and noise can be removed by smoothing the data. High-pass filtering: In contrast to low-pass filtering, high-pass filtering is used to remove low-frequency noise and retain the high-frequency changes in the data. Band-pass filtering: selects data within a specific frequency range and removes noise at other frequencies.

[0074] The data fusion module uses the particle swarm optimization algorithm to fuse multi-source data and determine the optimal weights and parameters for data fusion to improve the accuracy and robustness of the overall data set. By fusing multi-source data and integrating the advantages of each data source, the comprehensiveness and accuracy of data analysis can be improved, providing strong support for resource scheduling. It defines the problem space, analyzes the preprocessed multi-source data, determines the types of data sources and data types in the Internet of Things, initializes the particle swarm, and randomly generates a group of particles. Each particle represents a potential data fusion solution. The position of each particle represents a specific set of data fusion parameters, such as the weights of different data sources, the size of the time window, the fusion rules, etc. The fitness function is designed to evaluate the quality of each particle solution. The fitness function is based on the performance of the fused data. The fitness function includes the prediction accuracy function, the data consistency function, and the information gain function. The prediction accuracy function: If Data fusion is used for prediction, and accuracy can be used as a measure of fitness. Data consistency function: evaluates the consistency and reliability of fused data. Information gain function: measures the amount of information added to the fused data relative to a single data source. The fitness value is calculated for each particle, and the individual optimal solution is updated to determine the global optimal solution, that is, the solution with the highest fitness among all particles. According to the current particle's position, speed, individual optimal solution, and global optimal solution, the particle swarm optimization algorithm is used to update the formula to adjust the speed and position of each particle, adjust the inertia weight w, cognitive factor c1, and social factor c2, and introduce a speed limit to prevent particles from flying out of the search space. A maximum number of iterations or a fitness threshold is set. When the number of iterations is reached or the fitness of a particle is no longer significantly improved, the iteration is stopped. After the iteration is completed, the weights and parameters corresponding to the global optimal particle are finally output as the best solution for data fusion.

[0075] Furthermore, the multi-source data particle swarm algorithm is as follows:

[0076] Each particle represents a potential solution, and the particle's position vector x is determined i , whose position vector x i Contains: Data source weight: w1, w2, ...w n , where n is the number of data sources. The standard particle swarm update formula is used to update the speed and position of the particles. The influence of data update time is added to the fitness function. When the data update time difference is large, a penalty term is introduced to reduce the dependence on outdated data. The weight is adjusted by weighted average or other methods combined with the timeliness of the data;

[0077] Fitness function f(x i ) can be defined as the negative value of the data fusion error, and the formula is:

[0078]

[0079] Among them, E is the error function, y is the reference data, It is the data obtained by fusion based on the current particle parameters;

[0080] The particle swarm optimization algorithm update formula is used to adjust the speed and position of each particle. The expression of the particle swarm optimization algorithm update formula is:

[0081]

[0082]

[0083] in, and are the velocity and position of particle i at the tth iteration in dimension d, and are the velocity and position of particle i at the t+1th iteration in dimension d, p i t d is the individual optimal position of particle i in dimension d, is the global optimal position, r1 and r2 are random numbers between [0,1], w is the inertia weight, c1 is the cognitive factor, and c2 is the social factor;

[0084] The resource management module predicts the time series changes and complex resource requirements of equipment resource requirements based on the time series prediction model and the machine learning prediction model. The time series prediction model analyzes historical data to predict the periodic changes in future resource requirements. The machine learning prediction model trains the model based on historical data to predict the nonlinear changes in equipment resource requirements and usage. It accurately predicts resource requirements, provides a scientific basis for resource scheduling, and avoids resource waste or shortage.

[0085] The resource scheduling module uses the queuing theory model to design resource scheduling strategies based on the prediction results, and applies the particle swarm optimization algorithm to optimize the queuing model parameters to achieve optimal resource allocation, reduce resource waste, and improve the overall operation efficiency of the system;

[0086] The decision support module is used to provide a user interface to display prediction results, scheduling strategies and system status, assist users in making decisions, and help users quickly understand the system status and resource scheduling through intuitive display and convenient operation interface, so as to make more scientific and reasonable decisions.

[0087] The second embodiment is based on the first embodiment. Figure 3 As shown in the figure, in the resource management module, the process of predicting the time series changes and complex resource requirements of equipment resource requirements includes:

[0088] Use the preprocessed historical data to train the ARIMA model and determine the parameters (p, d, q) of the ARIMA model, where: the ARIMA model is an autoregressive integrated moving average model, p is the number of autoregressive terms, d is the number of differences, which is used to make the non-stationary series stable, and q is the number of moving average terms. The particle swarm optimization algorithm is used to find the optimal ARIMA parameters (p, d, q) to minimize the prediction error of the model. The trained ARIMA model is used to predict the periodic changes in future resource demand. The prediction results reflect the seasonal or periodic trend of resource demand. The prediction formula of the ARIMA model is:

[0089] in, is the predicted value at the future time point t, c is a constant term, is the autoregressive coefficient, is the d-order difference, θ k is the moving average coefficient, ∈ t-k is the forecast error at a past time point;

[0090] The features of temperature, pressure, flow, CPU usage, and memory usage are extracted from the preprocessed historical data, and a composite prediction model is constructed by combining the decision tree and deep learning model. The deep learning model can handle more complex nonlinear relationships, while the decision tree model provides better interpretability. The model parameters are adjusted through the cross-validation method to optimize the prediction performance. Based on the constructed composite prediction model and the extracted sensor data features, the environmental interference index is analyzed to evaluate the deviation of sensor data from the normal state. The composite prediction model and the extracted CPU usage and memory usage features are used to obtain the device utilization index and resource load index. The device resource prediction index is obtained through comprehensive analysis to evaluate the device's demand for resources. The trained composite prediction model is used to predict the nonlinear changes in the device resource demand and usage. The sudden changes and non-periodic fluctuations in resource demand are captured by combining the prediction results. The prediction results of the time series prediction model and the machine learning prediction model are combined to form a comprehensive prediction of the device resource demand.

[0091] Furthermore, the expression of the environmental interference index is:

[0092]

[0093] Where SFI is the environmental interference index, T h is the observation value of the hth sensor, T ref is the standard value of the corresponding sensor, w his the weight of the hth sensor data, n is the total number of sensors, and ε is a small constant used to ensure that the denominator is not zero. The smaller the SFI value, the higher the prediction index. The smaller the SFI, the closer the sensor reading is to the ideal state, indicating that the IoT device is less affected by temperature, pressure, and flow.

[0094] The device resource prediction index is obtained by weighting the device utilization index and resource load index based on CPU usage and memory usage. The expression is:

[0095] RFI=w U ·RFI U +w L ·RFI L ;

[0096]

[0097] Among them, RFI is the equipment resource forecast index, RFI U is the equipment utilization index, w U is the weight of the equipment utilization index, RFI L is the resource load index, w L is the weight of the resource load index, w U +w L =1,U r is the utilization of the rth resource (CPU or memory), U max and U min are the minimum and maximum resource utilization, m is the number of resource types, L r is the load value of the rth resource, L avg is the average resource load value, L max It is the maximum value of the resource load value, reflecting the degree of fluctuation of the resource load. A smaller RFI value indicates that the equipment resources are in a relatively ideal state, that is, the resource utilization efficiency is high and stable. A larger RFI value may indicate that there are problems in resource utilization, such as over-utilization or large load fluctuations, which require attention and optimization.

[0098] In the resource scheduling module, the process of designing resource scheduling strategies using the queuing theory model includes:

[0099] The resource scheduling module receives the prediction results of the time series prediction model and the machine learning prediction model in the resource management module. The prediction results include the time series changes, periodic changes, and nonlinear and complex resource demand predictions of the device resource requirements. According to the prediction results, the resource scheduling module uses the queuing theory model to simulate the resource request and allocation process, and applies the particle swarm optimization algorithm to optimize the parameters of the queuing model to find the optimal solution. The fitness function is designed based on multiple indicators such as resource scheduling efficiency, response time, and resource utilization. In the particle swarm optimization algorithm, each particle searches for the optimal solution by iteratively updating its position and speed. The position and speed of the particle are updated based on the information of the individual optimal solution and the global optimal solution. By continuously adjusting the position and speed of the particle, the particle swarm optimization algorithm finds the queuing model parameters that make the fitness function value optimal. According to the optimized queuing model parameters, the resource scheduling module formulates a specific resource scheduling strategy, adjusts the allocation order, allocation amount, and priority of resources to ensure that while meeting resource requirements, the resource utilization and the overall operation efficiency of the system are maximized. The resource scheduling module applies the formulated scheduling strategy to the actual resource scheduling process of IoT devices, and adjusts resource allocation in real time to achieve optimal resource utilization and system performance.

[0100] In the decision support module, the process of displaying prediction results, scheduling strategies, and system status includes:

[0101] The resource management module and resource scheduling module integrate prediction results, scheduling strategies and system status data, and use charts and graphs to display time series predictions, resource demand changes, and resource utilization indicators. The current resource scheduling strategy is clearly displayed, including resource allocation order, allocation amount, and priority. The system status is monitored and displayed in real time, including resource usage, system load, and response time. According to the monitoring results of the system status, abnormal situations in system operation, such as equipment failure and insufficient resources, are identified, alarm information is issued, and corresponding solutions and suggestions are given to help users respond to and solve problems quickly.

[0102] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. An IoT device resource scheduling system based on multi-source data fusion, including a resource scheduling platform, characterized in that: The resource scheduling platform is communicatively connected with a data acquisition module, a data preprocessing module, a data fusion module, a resource management module, a resource scheduling module and a decision support module, wherein the modules are connected by electrical signals; The data acquisition module is used to collect multi-source data of sensor data and device status information from IoT devices, wherein the multi-source raw data includes real-time data and historical data; The data preprocessing module is used to perform preprocessing operations such as cleaning, standardization and denoising on the collected data; The data fusion module uses a particle swarm optimization algorithm to fuse multi-source data and determine the optimal weights and parameters for data fusion; The resource management module predicts the time series changes and complex resource requirements of equipment resource requirements based on the time series prediction model and the machine learning prediction model, wherein the time series prediction model analyzes historical data to predict the periodic changes of future resource requirements, and the machine learning prediction model is based on the historical data training model to predict the nonlinear changes of equipment resource requirements and usage; The resource scheduling module uses the queuing theory model to design resource scheduling strategies based on the prediction results, and applies the particle swarm optimization algorithm to optimize the queuing model parameters to achieve optimal resource allocation; The decision support module is used to provide a user interface to display prediction results, scheduling strategies and system status to assist users in making decisions.

2. The IoT device resource scheduling system based on multi-source data fusion according to claim 1 is characterized in that: In the data acquisition module, the process of collecting multi-source raw data includes: Deploy various sensors and different types of IoT devices in the IoT environment to cover the areas that need to be monitored and managed, and configure corresponding communication interfaces for IoT devices. Sensors include temperature sensors, pressure sensors, and flow sensors. Collect multi-source raw data from sensors and IoT devices, including sensor data of temperature, pressure, and flow, and IoT device status information of CPU usage and memory usage, wherein the multi-source raw data includes real-time data and historical data; Perform preliminary verification on the collected data, mark and preliminarily process the abnormal data, and distinguish between normal and abnormal sensor data and IoT device status information; The marked data is classified and labeled, and the data is temporarily stored in the data warehouse.

3. The IoT device resource scheduling system based on multi-source data fusion according to claim 2 is characterized in that: In the data preprocessing module, the data preprocessing process includes: Call the data in the data warehouse, analyze the integrity and continuity of the data set, determine the location of missing data, and fill in the missing data points by filling in the missing values ​​through mean filling, median filling, mode filling and interpolation; Identify data values ​​that are significantly different from other data points as outliers, delete the identified outliers, mark them in the log, record the encountered outlier data values, and identify and remove duplicate data; The Z-score standardization method is used to convert data of different dimensions and ranges into a unified format, where the expression of the Z-score standardization method is: x is the original data value, μ is the mean of the data, and σ is the standard deviation of the data; Select the filter type according to the data characteristics and analysis requirements, apply different types of filters to smooth and denoise the data, adjust the filter parameters to optimize the denoising effect, and remove short-term fluctuations and noise by smoothing the data. The filters are low-pass filtering, high-pass filtering, and band-pass filtering.

4. The IoT device resource scheduling system based on multi-source data fusion according to claim 3 is characterized in that: In the data fusion module, the process of fusing multi-source data includes: Define the problem space, analyze the pre-processed multi-source data, and determine the types of data sources and data types in the Internet of Things; Initialize the particle swarm and randomly generate a set of particles, each particle represents a potential data fusion solution, and the position of each particle represents a specific set of data fusion parameters; Design a fitness function to evaluate the quality of each particle solution. The fitness function is based on the performance of the fused data and includes prediction accuracy function, data consistency function and information gain function. Calculate the fitness value for each particle, update the individual optimal solution, and determine the global optimal solution, which is the solution with the highest fitness among all particles; According to the current particle position, speed, individual optimal solution and global optimal solution, the particle swarm optimization algorithm is used to update the formula to adjust the speed and position of each particle, adjust the inertia weight w, cognitive factor c1, social factor c2, and introduce speed limits to prevent particles from flying out of the search space; A maximum number of iterations or fitness threshold is set. When the number of iterations is reached or the fitness of a particle is no longer significantly improved, the iteration is stopped. After the iteration is completed, the weights and parameters corresponding to the global optimal particle are finally output as the best solution for data fusion.

5. The IoT device resource scheduling system based on multi-source data fusion according to claim 4 is characterized in that: The multi-source data particle swarm algorithm is as follows: Each particle represents a potential solution, and the particle's position vector x is determined i , whose position vector x i Contains: Data source weight: w1, w2, ...w n , where n is the number of data sources. The standard particle swarm update formula is used to update the speed and position of the particles. The influence of data update time is added to the fitness function. When the data update time difference is large, a penalty term is introduced to reduce the dependence on outdated data. The weight is adjusted by weighted average or other methods combined with the timeliness of the data; Fitness function f(x i ) can be defined as the negative value of the data fusion error, and the formula is: Among them, E is the error function, y is the reference data, It is the data obtained by fusion based on the current particle parameters; The particle swarm optimization algorithm update formula is used to adjust the speed and position of each particle. The expression of the particle swarm optimization algorithm update formula is: in, and are the velocity and position of particle i at the tth iteration in dimension d, and are the velocity and position of particle i at the t+1th iteration in dimension d, is the individual optimal position of particle i in dimension d, p g t is the global optimal position, r1 and r2 are random numbers between [0,1], w is the inertia weight, c1 is the cognitive factor, and c2 is the social factor.

6. The IoT device resource scheduling system based on multi-source data fusion according to claim 5 is characterized in that: In the resource management module, the process of predicting the time series changes of device resource requirements and complex resource requirements includes: Use the preprocessed historical data to train the ARIMA model and determine the parameters (p, d, q) of the ARIMA model, where: the ARIMA model is an autoregressive integrated moving average model, p is the number of autoregressive terms, d is the number of differences, which is used to make the non-stationary series stationary, and q is the number of moving average terms; The particle swarm optimization algorithm is used to find the optimal ARIMA parameters (p, d, q) to minimize the prediction error of the model; Use the trained ARIMA model to predict the cyclical changes in future resource demand. The prediction results reflect the seasonal or cyclical trend of resource demand. Extract the features of temperature, pressure, flow, CPU usage, and memory usage from the preprocessed historical data, and build a composite prediction model by combining decision trees and deep learning models. Adjust the model parameters through cross-validation methods to optimize the prediction performance. Based on the constructed composite prediction model and the extracted sensor data features, the environmental interference index is analyzed to evaluate the deviation of sensor data from the normal state. The composite prediction model and the extracted CPU usage and memory usage features are used to obtain the device utilization index and resource load index. The device resource prediction index is obtained through comprehensive analysis to evaluate the device's demand for resources. Use the trained composite prediction model to predict nonlinear changes in device resource demand and usage, and combine the prediction results to capture sudden changes and non-periodic fluctuations in resource demand; Combine the prediction results of the time series prediction model and the machine learning prediction model to form a comprehensive prediction of device resource requirements.

7. The IoT device resource scheduling system based on multi-source data fusion according to claim 6 is characterized in that: The expression of the environmental interference index is: Where SFI is the environmental interference index, T h is the observation value of the hth sensor, T ref is the standard value of the corresponding sensor, w h is the weight of the hth sensor data, n is the total number of sensors, and ε is a constant used to ensure that the denominator is not zero. The smaller the SFI, the closer the sensor reading is to the ideal state, indicating that the IoT device is less affected by temperature, pressure, and flow.

8. The IoT device resource scheduling system based on multi-source data fusion according to claim 7 is characterized in that: The device resource prediction index is obtained by weighting the device utilization index and resource load index based on CPU utilization and memory utilization, and its expression is: <h2 style=";text-align:left;direction:ltr">RFI=w<h2 style=";text-align:left;direction:ltr"> U <h2 style=";text-align:left;direction:ltr"> RFI<h2 style=";text-align:left;direction:ltr"> U <h2 style=";text-align:left;direction:ltr"> +w<h2 style=";text-align:left;direction:ltr"> L <h2 style=";text-align:left;direction:ltr"> RFI<h2 style=";text-align:left;direction:ltr"> L <h2 style=";text-align:left;direction:ltr"> ; Among them, RFI is the equipment resource forecast index, RFI U is the equipment utilization index, w U is the weight of the equipment utilization index, RFI L is the resource load index, w L is the weight of the resource load index, w U +w L =1,U r is the utilization rate of the rth resource, U max and U min are the minimum and maximum resource utilization, m is the number of resource types, L r is the load value of the rth resource, L avg is the average resource load value, L max It is the maximum value of the resource load, reflecting the degree of fluctuation of the resource load.

9. The IoT device resource scheduling system based on multi-source data fusion according to claim 8 is characterized in that: In the resource scheduling module, the process of designing a resource scheduling strategy using a queuing theory model includes: The resource scheduling module receives the prediction results from the time series prediction model and the machine learning prediction model in the resource management module. The prediction results include the time series changes, periodic changes, and nonlinear and complex resource demand predictions of the equipment resource requirements; According to the prediction results, the resource scheduling module uses the queuing theory model to simulate the resource request and allocation process, applies the particle swarm optimization algorithm to optimize the parameters of the queuing model, finds the optimal solution, and designs the fitness function based on multiple indicators such as resource scheduling efficiency, response time, and resource utilization; In the particle swarm optimization algorithm, each particle searches for the optimal solution by iteratively updating its position and velocity. The particle position and velocity are updated based on the information of the individual optimal solution and the global optimal solution. By continuously adjusting the particle position and velocity, the particle swarm optimization algorithm finds the queuing model parameters that make the fitness function value optimal. According to the optimized queuing model parameters, the resource scheduling module formulates a specific resource scheduling strategy to adjust the resource allocation order, allocation amount, and priority; The resource scheduling module applies the formulated scheduling strategy to the actual IoT device resource scheduling process and adjusts resource allocation in real time.

10. The IoT device resource scheduling system based on multi-source data fusion according to claim 9, characterized in that: In the decision support module, the process of displaying the prediction results, scheduling strategies and system status includes: Integrate forecast results, scheduling strategies, and system status data from the resource management module and resource scheduling module, and use charts and graphs to display time series forecasts, resource demand changes, and resource utilization indicators; Display the current resource scheduling strategy, including resource allocation order, allocation amount, and priority, and monitor and display the system status in real time, including resource usage, system load, and response time; Based on the monitoring results of the system status, identify abnormal situations in system operation, issue alarm information, and provide corresponding solutions and suggestions.

Images