Intelligent control method and system of intelligent computing center combined with environmental health information monitoring

Abstract

This invention discloses an intelligent control method and system for intelligent computing centers that integrates environmental health information monitoring. It relates to the field of data processing technology and includes: acquiring the intelligent computing resource node space of the intelligent computing center; collecting multi-source environmental health monitoring data and generating multiple computing task links; combining intelligent computing resource nodes for each computing task link to generate candidate collaborative intelligent computing resource units; calling a fusion readiness evaluation model for analysis to obtain fusion readiness indicators for multi-source output data; and selecting multiple preferred collaborative intelligent computing resource units to execute calculations on the multi-source environmental health monitoring data. This invention solves the technical problems in existing technologies where neglecting synchronous processing consistency during the fusion analysis of multi-source environmental health monitoring data leads to temporal offsets, semantic inconsistencies, or feature space misalignments. It achieves the technical effects of improving the synchronous consistency, accuracy, effectiveness of intelligent computing resource collaborative scheduling, and reliability of fusion analysis results in cross-source data fusion analysis.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for intelligent control of a smart computing center that combines environmental health information monitoring. Background Technology

[0002] Environmental health monitoring systems typically include air quality sensors, temperature and humidity monitors, noise monitoring terminals, video surveillance equipment, and meteorological data sources, which vary significantly in sampling frequency, time reference, data format, and semantic representation. Existing intelligent computing centers primarily use resource scheduling mechanisms based on computing power scale, such as the number of CPU / GPU cores and memory size, as well as current load status, such as task queue length and resource utilization. They employ "computing power priority" or "load balancing" strategies to allocate resource nodes for computing tasks. However, this approach presents numerous challenges for the fusion analysis of multi-source, heterogeneous environmental health monitoring data. Asynchronous clocks, data transmission delays, and inconsistent task queue processing orders among different data sources affect the accuracy of pollutant diffusion trend predictions and anomaly timing tracing during fusion analysis. Furthermore, data from different devices uses different units, encoding rules, or health status classification standards, and semantic inconsistencies can lead to logical conflicts or invalid comparisons during fusion. The varying feature dimensions, distribution characteristics, and noise levels of different data sources reduce the performance of subsequent classification, regression, or anomaly detection models.

[0003] Therefore, in the current related technologies, there are technical problems such as neglecting the consistency of synchronous processing during the fusion analysis of multi-source environmental health monitoring data, resulting in temporal offset, semantic inconsistency, or feature space misalignment. Summary of the Invention

[0004] This application provides an intelligent control method and system for intelligent computing centers that combines environmental health information monitoring. This solves the technical problems in the prior art where neglecting the consistency of synchronous processing during the fusion analysis of multi-source environmental health monitoring data leads to temporal offsets, semantic inconsistencies, or feature space misalignments. It achieves the technical effect of improving the synchronous consistency, accuracy, effectiveness of intelligent computing resource collaborative scheduling, and reliability of fusion analysis results in cross-source data fusion analysis.

[0005] This application provides an intelligent control method for a smart computing center that combines environmental health information monitoring. The method includes: acquiring a space of smart computing resource nodes that can be used for control within the smart computing center; collecting multi-source environmental health monitoring data and generating multiple corresponding computing task links based on the multi-source environmental health monitoring data; combining smart computing resource nodes for each of the multiple computing task links through the smart computing resource node space to generate candidate collaborative smart computing resource units; calling a fusion readiness evaluation model to analyze the multiple candidate collaborative smart computing resource units corresponding to the multiple computing task links and obtaining a fusion readiness index for the multi-source output data obtained after processing the multi-source environmental health monitoring data based on each candidate collaborative smart computing resource unit; selecting multiple preferred collaborative smart computing resource units corresponding to the multiple computing task links according to the fusion readiness index, and the smart computing center performing calculations on the multi-source environmental health monitoring data based on the multiple preferred collaborative smart computing resource units.

[0006] In one possible implementation, the method for obtaining the intelligent computing resource node space that can be used for regulation in the intelligent computing center includes: obtaining the basic resource parameters and operating performance parameters of each intelligent computing resource node in the intelligent computing center; analyzing the historical output consistency parameters of the same type of intelligent computing resource nodes based on the basic resource parameters; encoding the basic resource parameters, operating performance parameters, and historical output consistency parameters to construct the node state vector corresponding to each intelligent computing resource node; and constructing the intelligent computing resource node space according to each intelligent computing resource node and its corresponding node state vector.

[0007] In a possible implementation, the multi-source environmental health monitoring data is analyzed according to a DAG (Directed Acyclic Graph) to generate multiple corresponding computational task links, wherein each computational task link includes at least a data preprocessing stage, a feature extraction stage, a correlation analysis stage, and a fusion computation stage.

[0008] In a possible implementation, the intelligent computing resource node combination is performed on each of the multiple computing task links through the intelligent computing resource node space. The method includes: obtaining the resource requirement feature vector of each computing node in each computing task link; matching the resource requirement feature vector with the intelligent computing resource node space to obtain candidate intelligent computing resource nodes for each computing node; combining the candidate intelligent computing resource nodes of each computing node according to each computing task link to output candidate collaborative intelligent computing resource units; and performing conflict optimization on the multiple candidate collaborative intelligent computing resource units of the multiple computing task links to output optimized candidate collaborative intelligent computing resource units.

[0009] In a possible implementation, the method for obtaining the fusion readiness index of the multi-source output data obtained after processing the multi-source environmental health monitoring data based on each candidate collaborative intelligent computing resource unit includes: extracting the time synchronization characteristics, feature distribution characteristics, and standardized state characteristics of the obtained multi-source output data; calculating the time alignment degree based on the time synchronization characteristics, which is used to characterize the degree of alignment of the output results from different data sources on the time axis; calculating the spatial alignment degree based on the feature distribution characteristics, which is used to characterize the degree of closeness of the output results from different data sources in statistical distribution; calculating the standardized state characteristics based on the standardized state characteristics, which is used to characterize the degree of uniformity of the output results from different data sources in terms of units and coding rules; and calling a fusion readiness evaluation model to evaluate the time alignment degree, spatial alignment degree, and standardized state characteristics, and outputting the fusion readiness index.

[0010] In a possible implementation, after generating multiple computing task links corresponding to the multi-source environmental health monitoring data, the method further includes: obtaining the fusion computing task nodes of the multiple computing task links; inputting the computing task links of the fusion readiness evaluation model according to the fusion link object tags of the fusion computing task nodes; and reading the marked candidate collaborative intelligent computing resource units corresponding to the marked computing task links.

[0011] In a possible implementation, a fusion readiness evaluation model is invoked to evaluate the time alignment, spatial alignment, and standardized state features. The fusion readiness evaluation model includes learning parameters, which are obtained by training on historical data samples of multi-source environmental health monitoring of the intelligent computing center and the label information of the corresponding actual fusion readiness evaluation results. The fusion readiness evaluation model converges when the fusion loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result is less than a preset loss threshold.

[0012] In a possible implementation, the label loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result is calculated. The method includes: obtaining the temporal alignment loss, spatial alignment loss, and normalized state feature loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result; and performing fusion loss analysis based on the temporal alignment loss, spatial alignment loss, and normalized state feature loss to obtain the fusion loss.

[0013] In a possible implementation, multiple preferred collaborative intelligent computing resource units corresponding to the multiple computing task links are selected according to the fusion readiness index. The method includes: sorting the candidate collaborative intelligent computing resource units corresponding to each computing task link according to the fusion readiness index, obtaining the preferred collaborative intelligent computing resource units for each computing task link; and obtaining multiple preferred collaborative intelligent computing resource units corresponding to multiple computing task links.

[0014] This application also provides an intelligent control system for a smart computing center that integrates environmental health information monitoring. The system includes: a resource node space acquisition module for acquiring intelligent computing resource node space available for control within the smart computing center; a monitoring data acquisition module for collecting multi-source environmental health monitoring data and generating multiple corresponding computing task links based on the multi-source environmental health monitoring data; an intelligent computing resource node combination module for combining intelligent computing resource nodes in each of the multiple computing task links through the intelligent computing resource node space to generate candidate collaborative intelligent computing resource units; a fusion readiness index acquisition module for calling a fusion readiness evaluation model to analyze the multiple candidate collaborative intelligent computing resource units corresponding to the multiple computing task links and obtain a fusion readiness index for the multi-source output data obtained after processing the multi-source environmental health monitoring data based on each candidate collaborative intelligent computing resource unit; and a health monitoring data calculation module for selecting multiple preferred collaborative intelligent computing resource units corresponding to the multiple computing task links according to the fusion readiness index, and the smart computing center performing calculations on the multi-source environmental health monitoring data based on the multiple preferred collaborative intelligent computing resource units.

[0015] This application proposes a method and system for intelligent control of intelligent computing centers that integrates environmental health information monitoring. The method involves: acquiring the intelligent computing resource node space of the intelligent computing center; collecting multi-source environmental health monitoring data to generate multiple computing task chains; combining intelligent computing resource nodes for each computing task chain to generate candidate collaborative intelligent computing resource units; calling a fusion readiness evaluation model for analysis to obtain fusion readiness indicators for multi-source output data; and selecting multiple preferred collaborative intelligent computing resource units to execute calculations on the multi-source environmental health monitoring data. This addresses the technical problems in existing technologies where neglecting synchronous processing consistency during the fusion analysis of multi-source environmental health monitoring data leads to temporal shifts, semantic inconsistencies, or feature space misalignments. It achieves the technical effect of improving the synchronous consistency, accuracy, effectiveness of intelligent computing resource collaborative scheduling, and reliability of fusion analysis results in cross-source data fusion analysis. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1This is a schematic diagram of the intelligent control method for a computing center that combines environmental health information monitoring, provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the intelligent control system structure of the intelligent computing center that combines environmental health information monitoring, provided in an embodiment of this application.

[0019] Figure labeling: Resource node space acquisition module 10, monitoring data acquisition module 20, intelligent computing resource node combination module 30, fusion readiness index acquisition module 40, health monitoring data calculation module 50. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0021] This application provides an intelligent control method for a smart computing center that combines environmental health information monitoring, such as... Figure 1 As shown, the method includes: Step S100: Obtain the intelligent computing resource node space that the intelligent computing center can use for regulation.

[0022] Step S100 further includes: obtaining the basic resource parameters and operating performance parameters of each intelligent computing resource node in the intelligent computing center; analyzing the historical output consistency parameters of the same type of intelligent computing resource nodes based on the basic resource parameters; encoding the basic resource parameters, operating performance parameters, and historical output consistency parameters to construct the node state vector corresponding to each intelligent computing resource node; and constructing the intelligent computing resource node space according to each intelligent computing resource node and its corresponding node state vector.

[0023] Preferably, the processor type, number of computing cores, storage capacity, video memory capacity, network bandwidth, and current load rate of each intelligent computing resource node in the intelligent computing center are obtained to determine the basic resource parameters. Among them, the processor type is the instruction set architecture and microarchitecture identifier of the computing chip used by the intelligent computing resource node, such as different categories such as x86, ARM, GPU, NPU, DPU, etc.; the number of computing cores is the total number of processing units that independently execute computing tasks within the processor; the storage capacity refers to the total number of available bytes of solid-state drives and hard disk drives equipped in the node; the video memory capacity is the total number of bytes of high-bandwidth memory dedicated to GPUs or accelerator cards in the node; the network bandwidth refers to the maximum number of data bits that the node's network card or network interface can transmit per unit time; and the current load rate refers to the percentage of computing resources occupied by the node at the current point in time relative to the total available resources. Next, obtain the average response latency, throughput, task queue length, and resource utilization volatility of each intelligent computing resource node in the intelligent computing center to determine the operating performance parameters. Among them, response latency is the time interval between a node receiving a computing task and outputting a result. The average response latency is obtained by taking the arithmetic mean of multiple historical tasks. Throughput is the number of computing tasks that a node successfully completes and outputs results per unit time. Task queue length is the number of computing tasks that have not yet started execution in the node's current waiting queue. Resource utilization volatility refers to the ratio of the standard deviation to the mean of the node's computing resource utilization over a period of time, which is used to measure the stability of resource usage.

[0024] Preferably, based on the basic resource parameters, statistical calculations are performed on the historical operating data of the same type of intelligent computing resource nodes to obtain standardized output deviation, feature distribution difference degree, and timestamp alignment error, thus determining historical output consistency parameters. Specifically, standardized output deviation refers to the degree of difference between the output result of the same type of node under the same input conditions and the mean output result of the same type of node; feature distribution difference degree refers to the difference between the feature vector distribution of the output result of the same type of node processing the same data distribution and the output distribution of the same type of node, calculated using KL divergence or JS divergence; and timestamp alignment error is the maximum absolute value of the deviation between the timestamps carried by the output results of different nodes in a multi-source data collaborative processing scenario. Then, the basic resource parameters, operating performance parameters, and historical output consistency parameters are converted into unified numerical vector elements through normalization, discretization, or embedding mapping. These elements are then concatenated in a fixed order to form a node state vector, with each intelligent computing resource node corresponding to a unique node state vector. Finally, all controllable intelligent computing resource nodes and their corresponding node state vectors together constitute the intelligent computing resource node space, which is indexed by node identifiers, with each index pointing to the state vector of that node.

[0025] Step S200: Collect multi-source environmental health monitoring data and generate multiple corresponding computing task links according to the multi-source environmental health monitoring data.

[0026] Step S200 further includes generating multiple corresponding computational task links by analyzing the multi-source environmental health monitoring data according to the DAG directed acyclic graph, wherein each computational task link includes at least a data preprocessing stage, a feature extraction stage, a correlation analysis stage, and a fusion computation stage.

[0027] Preferably, a Directed Acyclic Graph (DAG) is a graph structure composed of vertices and directed edges. Each edge has a direction, and it is impossible to return to a vertex by following the direction of an edge from any vertex. In the graph structure, vertices represent computational operations, directed edges represent data dependencies between operations, and the direction of the edges indicates the order of computation execution. The processing flow of multi-source environmental health monitoring data is represented as a DAG structure, where each vertex represents a computational operation performed on the data, such as filtering, normalization, and feature extraction. Each directed edge represents the computation result of the upstream operation as input data for the downstream operation. The data source types, data formats, sampling frequencies, data volumes, and logical relationships between the data sources in the multi-source environmental health monitoring data are analyzed. Based on the analysis results, a corresponding DAG structure is constructed for each data source or each group of data processing flows with dependencies. Each DAG structure corresponds to a computational task chain, where each computational task chain includes at least a data preprocessing stage, a feature extraction stage, a correlation analysis stage, and a fusion computation stage.

[0028] Preferably, the data preprocessing stage is used to perform cleaning, transformation, and normalization operations on the raw multi-source environmental health monitoring data. Specifically, this includes missing value handling (filling in or deleting empty values ​​in the data); outlier detection and correction (identifying and correcting values ​​exceeding the sensor's range or statistical threshold); noise filtering (using methods such as low-pass filtering and median filtering to eliminate sensor noise); data normalization (converting data of different dimensions to a unified numerical range); and time alignment (resampling or interpolating data from different sampling frequencies according to a unified time reference). The feature extraction stage is used to extract key numerical features that characterize the environmental health status from the preprocessed raw data. Specifically, this includes time-domain feature extraction, such as calculating statistics like mean, variance, peak value, and root mean square; frequency-domain feature extraction, such as extracting spectral energy distribution and dominant frequency components through Fourier transform; and time-series feature extraction, such as extracting rate of change, trend slope, and autocorrelation coefficient. The correlation analysis stage is used to perform correlation analysis on different features, identify the dependencies and influence patterns among different environmental indicators, specifically including calculating the Pearson correlation coefficient or Spearman rank correlation coefficient between different features, determining the time lag between changes in one environmental indicator and the impact of another, determining the causal direction between data sources using methods such as Granger causality tests, and identifying anomalous data combinations. The fusion calculation stage is used to comprehensively calculate and generate environmental health status assessment results from multi-source features, specifically including concatenating or weighting multiple feature vectors into a single feature vector, performing joint inference on the judgment results from each data source, and outputting the final environmental health index, risk level, or predicted pollution concentration value.

[0029] Step S300: Combine the intelligent computing resource nodes of each of the multiple computing task links through the intelligent computing resource node space to generate candidate collaborative intelligent computing resource units.

[0030] Step S300 further includes: obtaining the resource requirement feature vector of each computing node in each computing task link; matching the resource requirement feature vector with the intelligent computing resource node space to obtain candidate intelligent computing resource nodes for each computing node; combining the candidate intelligent computing resource nodes of each computing node according to each computing task link to output candidate collaborative intelligent computing resource units; and performing conflict optimization on the multiple candidate collaborative intelligent computing resource units of the multiple computing task links to output optimized candidate collaborative intelligent computing resource units.

[0031] Preferably, each vertex in the DAG task chain is considered a computing node, such as "missing value imputation" in the data preprocessing stage or "mean calculation" in the feature extraction stage. Each computing node in each computing task chain is traversed, and its resource requirement feature vector is calculated based on the node's operation type, input data volume, algorithm complexity, and time constraints. This vector represents the computing resource requirements for performing a specific operation, including at least processor type requirements, number of computing cores requirements, storage capacity requirements, GPU memory capacity requirements, network bandwidth requirements, and computation time constraints. The resource requirement feature vector of each computing node is then compared with the intelligent computing resources... The state vectors of each intelligent computing resource node in the node space are compared and calculated. For example, the dot product of two vectors is used to calculate the inner product of two vectors, and the larger the value, the better the match. The Euclidean distance between two vectors is used to calculate the Euclidean distance between two vectors, and the smaller the value, the better the match. For each computing node, the top K intelligent computing resource nodes that are selected from the intelligent computing resource node space, whose processor type is consistent with the processor type requirement in the resource requirement feature vector, and whose number of computing cores, storage capacity, video memory capacity, and network bandwidth are all greater than or equal to the corresponding requirements, are selected. K is a preset number, such as 3 or 5. In this way, multiple candidate intelligent computing resource nodes that can meet its resource requirements are filtered.

[0032] Preferably, for a computing task link containing N computing nodes, each computing node i corresponds to multiple candidate intelligent computing resource nodes. One candidate intelligent computing resource node is selected to form an ordered combination of N nodes. The intelligent computing resource node at position i is used to execute the computing operation corresponding to the i-th computing node in the link, thereby obtaining a specific resource allocation scheme, i.e., a candidate collaborative intelligent computing resource unit. This unit specifically includes the computing task link identifier, the intelligent computing resource node identifier allocated to each computing node, the type of computing operation undertaken by each resource node in the link, and the execution order of each computing node. For each computing task link, multiple different candidate collaborative intelligent computing resource units may be generated based on the size of the candidate set of each computing node. When different candidate collaborative intelligent computing resource units of multiple computing task links contain the same intelligent computing resource node, a conflict occurs if two links need to use the same node to perform computing operations simultaneously within the same time period.

[0033] Preferably, conflicts are categorized into: node exclusivity conflict (where a single intelligent computing resource node can only execute one computing task at a time and cannot simultaneously serve two different computing task chains); resource overload conflict (where, although an intelligent computing resource node can process tasks in a time-sharing manner, the task time windows of two chains overlap, and the cumulative resource demand exceeds the node's capacity); and sequence dependency conflict (where two chains require the use of the same node, but the execution order does not conform to the DAG dependency constraint). Candidate collaborative intelligent computing resource units for multiple computing task chains are jointly screened and optimized. If the intelligent computing resource nodes involved in the selected collaborative intelligent computing resource unit do not have any type of conflict, the optimized candidate collaborative intelligent computing resource unit is retained. If multiple chains compete for the same intelligent computing resource node, the allocation decision is made based on the priority weight of each chain.

[0034] Step S400: Invoke the fusion readiness evaluation model to analyze multiple candidate collaborative intelligent computing resource units corresponding to the multiple computing task links, and obtain the fusion readiness index of the multi-source output data obtained after processing the multi-source environmental health monitoring data based on each candidate collaborative intelligent computing resource unit.

[0035] Preferably, the fusion readiness evaluation model is used to simulate or predict the quality characteristics of the output data generated by each candidate collaborative intelligent computing resource unit after executing the multi-source environmental health monitoring data processing task. Specifically, the intelligent computing resource nodes allocated in a candidate collaborative intelligent computing resource unit are used to process the multi-source environmental health monitoring data on their corresponding computing nodes. Each computing node generates multi-source output data, including the result data generated by each computing node after performing computational operations on its allocated intelligent computing resource nodes, the output results generated by the fusion computation stage after all computing nodes in a computational task chain have been executed sequentially, and the set of multiple output results generated after multiple computational task chains have been executed independently. The fusion readiness evaluation model evaluates the time synchronization characteristics, feature distribution characteristics, and standardized state characteristics of the multi-source output data based on the configuration parameters of the candidate collaborative intelligent computing resource unit and the characteristics of the multi-source environmental health monitoring data. Among them, the time synchronization characteristics represent the degree of alignment of each data stream in the multi-source output data on the time axis, the feature distribution characteristics represent the degree of similarity of each data stream in the statistical distribution of the multi-source output data, and the standardized state characteristics represent the degree of uniformity of the multi-source output data in terms of units, numerical range, and coding rules.

[0036] Furthermore, step S400 also includes extracting the time synchronization features, feature distribution features, and standardized state features of the obtained multi-source output data; calculating the time alignment degree based on the time synchronization features, which is used to characterize the degree of alignment of the output results from different data sources on the time axis; calculating the spatial alignment degree based on the feature distribution features, which is used to characterize the degree of similarity in the statistical distribution of the output results from different data sources; calculating the standardized state features based on the standardized state features, which is used to characterize the degree of uniformity in the units and coding rules of the output results from different data sources; and calling the fusion readiness evaluation model to evaluate the time alignment degree, spatial alignment degree, and standardized state features, and outputting the fusion readiness index.

[0037] Preferably, the header or metadata fields of the multi-source output data of each intelligent computing resource node are parsed to extract time synchronization features, such as the timestamps carried by the output results of each data source, the timestamp difference sequence between the output results of different data sources, the sampling interval or output interval of each data source, and the point density distribution on the time axis in the multi-source output data; statistical calculations are performed on the numerical sequence output by each intelligent computing resource node to generate first-order, second-order, and higher-order statistics describing the distribution characteristics of the sequence, and feature distribution features are obtained, such as the mean, median, mode, variance and standard deviation, skewness, kurtosis, and the numerical range formed by the minimum and maximum values ​​of the numerical output results of each data source; the unit and encoding information in the data format description field or metadata of the output results of each intelligent computing resource node are parsed, and the numerical sequence is scanned to obtain standardized state features, such as the physical quantity unit identifier, numerical scale range, and encoding mapping table of categorical data of the output results of each data source.

[0038] Preferably, the data source with the highest timestamp precision or sampling frequency is selected from the output results of all data sources as the time reference benchmark. The time offset of each data source relative to the time reference benchmark is calculated, and the time alignment is calculated using an inverse normalization formula. This is used to quantitatively characterize the alignment degree of the output results of different data sources on the time axis. The higher the value, the better the alignment degree, reaching its maximum when perfectly aligned. For the output feature value sequence of each data source, kernel density estimation or histogram statistics are used to calculate its probability density function, estimating the probability distribution of the output results of each data source. The distribution difference between pairwise data sources is calculated using KL divergence or JS divergence. The average or maximum value of the distribution difference measure for all pairwise combinations of data sources is taken, and then converted into alignment degree, i.e., spatial alignment degree = 1 - (average distribution difference / maximum possible distribution difference). When the output distribution of all data sources is exactly the same, the spatial alignment degree = 1; when the distribution difference reaches its maximum, the spatial alignment degree = 0.

[0039] Preferably, for the output results of all data sources, check whether the physical quantity units belong to the same dimensional system, calculate the unit consistency score, and then check whether the numerical range has been uniformly normalized. The scale uniformity score = 1 - (numerical range difference of each data source / maximum allowable range). The numerical range difference is the difference between the maximum value among the maximum values ​​of each data source and the minimum value among the minimum values ​​of each data source. Next, for the categorical data fields in the output results of all data sources, check whether their encoding mapping tables are consistent. The encoding consistency score = (number of fields with completely consistent encoding rules) / (total number of categorical fields). The unit consistency score, scale uniformity score, and encoding consistency score are calculated by weighting, and the weight coefficients are summed to 1. Then, the temporal alignment, spatial alignment, and standardized state characteristics are input into the fusion readiness evaluation model for analysis. The three alignments are weighted and summed or multiplied to output a fusion readiness index. A high value indicates that after using this candidate collaborative intelligent computing resource unit to process multi-source environmental health monitoring data, the multi-source output data performs well in terms of temporal alignment, statistical distribution consistency, and standardization uniformity, and has a high degree of fusion readiness. A low value indicates that the output data has serious inconsistencies in one or more dimensions, has a low degree of fusion readiness, and it is not recommended to use this resource unit.

[0040] Furthermore, step S400 also includes calling a fusion readiness evaluation model to evaluate the time alignment, spatial alignment, and standardized state features. The fusion readiness evaluation model includes learning parameters, which are obtained by training on historical data samples of multi-source environmental health monitoring of the intelligent computing center and label information corresponding to the actual fusion readiness evaluation results. The fusion readiness evaluation model converges when the fusion loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result is less than a preset loss threshold.

[0041] Preferably, the learning parameters are adjustable variables within the fusion readiness evaluation model. Their values ​​are continuously updated during model training and are used to control the mapping relationship between inputs and outputs. For a linear weighted model, the learning parameters include three weight coefficients and possible bias terms. Before training begins, the learning parameters are set to random values ​​or initialized to all zeros. Specifically, the training data includes historical data samples from multi-source environmental health monitoring and label information corresponding to the actual fusion readiness evaluation results. The historical data samples from multi-source environmental health monitoring include multi-source environmental health monitoring data actually collected and processed by the intelligent computing center during its past operation, such as the inputs and intermediate results of each stage of data preprocessing, feature extraction, correlation analysis, and fusion calculation. The label information corresponding to the actual fusion readiness evaluation results refers to the multi-source output data processed by the resource allocation scheme relative to the historical data samples, which is determined through manual annotation or third-party verification, and the standard fusion readiness index values.

[0042] Preferably, for each historical data sample, its corresponding time alignment, spatial alignment, and normalized state alignment are calculated and concatenated into a three-dimensional feature vector as the input sample of the model. For the same historical data sample, its corresponding actual fusion readiness evaluation result label is obtained through expert annotation and known effective fusion verification. The input feature vector is paired with the label to form a training sample. Then, the input sample is input into the fusion readiness evaluation model under the current parameter state. The model calculates and outputs the predicted fusion readiness evaluation result. The predicted value and the actual label are substituted into the loss function to calculate the loss value of the current sample. According to the gradient direction of the loss value relative to each learning parameter, the learning parameters are adjusted according to gradient descent optimization and Adam optimization to reduce the loss value. Multiple rounds of training are performed using all samples in the training dataset, and the learning parameters are gradually updated after each iteration. The fusion loss is a weighted sum of temporal alignment loss, spatial alignment loss, and standardized state feature loss. Weighting coefficients control the contribution of different alignment dimensions to the total loss. The fusion loss quantifies the difference between the predicted and actual fusion readiness evaluation results. A smaller loss value indicates more accurate model predictions, while a loss of zero indicates complete consistency between the predicted and actual values. A preset loss threshold is established based on business accuracy requirements and validation set performance, serving as the decision boundary for determining whether the model is sufficiently trained. When the fusion loss between the predicted and actual fusion readiness evaluation results is less than the preset loss threshold, the fusion readiness evaluation model is considered to have converged.

[0043] Furthermore, step S400 also includes obtaining the temporal alignment loss, spatial alignment loss, and normalized state feature loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result; and performing fusion loss analysis based on the temporal alignment loss, spatial alignment loss, and normalized state feature loss to obtain the fusion loss.

[0044] Preferably, the squared loss function is used to calculate the temporal alignment loss, spatial alignment loss, and normalized state feature loss between the predicted and actual fusion readiness evaluation results. These losses are used to quantify the differences between the predicted and actual temporal alignment, spatial alignment, and normalized state alignment, respectively, with values ​​ranging from [0, 1]. The temporal alignment loss, spatial alignment loss, and normalized state feature loss are then weighted and summed to obtain the fusion loss, which comprehensively characterizes the model's prediction performance. A fusion loss of 0 indicates that the predicted result is completely consistent with the actual result, with no error in the alignment of the three dimensions. A fusion loss greater than 0 indicates that there is an error in the prediction result, with a larger value indicating a higher degree of error. The weight coefficients are dynamically adjusted according to the application scenario type. For example, in time-sensitive applications, the weight coefficients for temporal alignment loss, spatial alignment loss, and normalized state feature loss are 0.6, 0.2, and 0.2, respectively; in balanced applications, the weight coefficients are 0.33, 0.33, and 0.34, respectively.

[0045] Furthermore, step S400 also includes obtaining the fusion computing task nodes of the multiple computing task links; inputting the computing task links of the fusion readiness evaluation model according to the fusion link object tags of the fusion computing task nodes; and reading the marked candidate collaborative intelligent computing resource units corresponding to the marked computing task links.

[0046] Preferably, the structure of each computing task link is analyzed, the nodes of the fusion computing stage are identified, and fusion computing task nodes are extracted for each computing task link. These are computing nodes that perform multi-source data fusion computing operations in a computing task link. Each computing task link includes at least a data preprocessing stage, a feature extraction stage, a correlation analysis stage, and a fusion computing stage. The fusion computing task node is one or more computing nodes in the fusion computing stage. Identification and extraction are performed on all computing task links, and complete fusion computing task nodes are output. Based on the information of the labeled computing task links as query conditions, identifiable tags are added to each computing task link that needs to be input into the fusion readiness evaluation model to distinguish different fusion link objects. All labeled candidate collaborative intelligent computing resource units corresponding to the link are retrieved and obtained. Then, labeled candidate collaborative intelligent computing resource units associated with the fusion computing task nodes in a specific computing task link are selected. A tag field associated with the fusion computing task node is added. The labeled candidate collaborative intelligent computing resource unit content includes at least a link identifier, a fusion node identifier, a resource allocation scheme, tag information, and a fusion readiness index.

[0047] Step S500: Select multiple preferred collaborative intelligent computing resource units corresponding to the multiple computing task links according to the fusion readiness index, and the intelligent computing center performs the calculation of the multi-source environmental health monitoring data based on the multiple preferred collaborative intelligent computing resource units.

[0048] Step S500 further includes sorting the candidate collaborative intelligent computing resource units corresponding to each computing task link according to the fusion readiness index, obtaining the preferred collaborative intelligent computing resource units for each computing task link; and obtaining multiple preferred collaborative intelligent computing resource units corresponding to multiple computing task links.

[0049] Preferably, candidate collaborative intelligent computing resource units corresponding to each computing task link and their corresponding fusion readiness index are obtained. Then, the fusion readiness index value is used as the sorting key, and the candidate collaborative intelligent computing resource units are sorted in descending order of value. The candidate collaborative intelligent computing resource units with larger fusion readiness index values ​​indicate that their output data after processing multi-source environmental health monitoring data performs better in terms of time alignment, spatial alignment, and standardized state alignment, and are ranked higher. Then, the first unit in the sorted list, that is, the unit with the largest fusion readiness index value, is selected as the preferred collaborative intelligent computing resource unit for each computing task link, or the first K units in the sorted list are selected as the preferred collaborative intelligent computing resource units for each computing task link, where K is a preset positive integer. The preferred collaborative intelligent computing resource unit acquisition operation is performed on all computing task links, and finally multiple preferred collaborative intelligent computing resource units corresponding to multiple computing task links are obtained.

[0050] Preferably, the intelligent computing center calculates multi-source environmental health monitoring data based on multiple optimized collaborative intelligent computing resource units. The intelligent computing center is a component integrating a large number of intelligent computing resource nodes (such as GPU servers, NPU nodes, and CPU clusters), possessing core functions such as resource scheduling, task distribution, and result aggregation. Specifically, the intelligent computing center converts each optimized collaborative intelligent computing resource unit into specific task execution instructions and distributes them to the corresponding intelligent computing resource nodes. Each intelligent computing resource node loads input data according to the task execution instructions, performs corresponding calculation operations within its allocated resource quota, and allocates independent resource partitions to each unit according to its resource quota, ensuring that units do not interfere with each other. The calculation operations include: in the data preprocessing stage, interpolating missing values, removing outliers, and standardizing Z-scores on sensor time-series data; in the feature extraction stage, calculating statistical features such as mean, variance, peak frequency, and rate of change within a sliding window; in the correlation analysis stage, calculating the Pearson correlation coefficient matrix and causality test between multi-source features; and in the fusion calculation stage, inputting multi-source features into a weighted fusion model and outputting a comprehensive environmental health index. Once all computing nodes in all computing task chains have completed their execution, the intelligent computing center collects the final output results of each chain, thereby improving the synchronization consistency, accuracy, effectiveness of intelligent computing resource collaborative scheduling, and reliability of cross-source data fusion analysis results.

[0051] In the above text, refer to Figure 1 This paper describes in detail an intelligent control method for a smart computing center that combines environmental health information monitoring according to an embodiment of the present invention. Next, we will refer to... Figure 2 This invention describes an intelligent control system for a computing center that incorporates environmental health information monitoring, according to an embodiment of the present invention.

[0052] The intelligent control system of the intelligent computing center combining environmental health information monitoring according to embodiments of the present invention is used to solve the technical problems existing in the prior art where neglecting the consistency of synchronous processing during the fusion analysis of multi-source environmental health monitoring data leads to temporal offsets, semantic inconsistencies, or feature space misalignments. It achieves the technical effect of improving the synchronous consistency, accuracy, effectiveness of intelligent computing resource collaborative scheduling, and reliability of fusion analysis results in cross-source data fusion analysis. Figure 2 As shown, the intelligent control system of the intelligent computing center combined with environmental health information monitoring includes: a resource node space acquisition module 10, a monitoring data acquisition module 20, an intelligent computing resource node combination module 30, a fusion readiness index acquisition module 40, and a health monitoring data calculation module 50.

[0053] The system comprises the following modules: a resource node space acquisition module 10, used to acquire intelligent computing resource node space available for regulation in the intelligent computing center; a monitoring data acquisition module 20, used to collect multi-source environmental health monitoring data and generate multiple corresponding computing task links based on the multi-source environmental health monitoring data; an intelligent computing resource node combination module 30, used to combine intelligent computing resource nodes for each of the multiple computing task links through the intelligent computing resource node space to generate candidate collaborative intelligent computing resource units; a fusion readiness index acquisition module 40, used to call a fusion readiness evaluation model to analyze the multiple candidate collaborative intelligent computing resource units corresponding to the multiple computing task links and obtain a fusion readiness index for the multi-source output data obtained after processing the multi-source environmental health monitoring data based on each candidate collaborative intelligent computing resource unit; and a health monitoring data calculation module 50, used to select multiple preferred collaborative intelligent computing resource units corresponding to the multiple computing task links according to the fusion readiness index, and the intelligent computing center performs calculations on the multi-source environmental health monitoring data based on the multiple preferred collaborative intelligent computing resource units.

[0054] The specific configuration of the resource node space acquisition module 10 will be described in detail below. The resource node space acquisition module 10 further includes: acquiring the basic resource parameters and operating performance parameters of each intelligent computing resource node in the intelligent computing center; analyzing the historical output consistency parameters of the same type of intelligent computing resource nodes based on the basic resource parameters; encoding the basic resource parameters, operating performance parameters, and historical output consistency parameters to construct a node state vector corresponding to each intelligent computing resource node; and constructing an intelligent computing resource node space according to each intelligent computing resource node and its corresponding node state vector.

[0055] The specific configuration of the monitoring data acquisition module 20 will be described in detail below. The monitoring data acquisition module 20 further includes: generating multiple corresponding computational task links based on the analysis of the multi-source environmental health monitoring data according to the DAG (Directed Acyclic Graph), wherein each computational task link includes at least a data preprocessing stage, a feature extraction stage, a correlation analysis stage, and a fusion computation stage.

[0056] The specific configuration of the intelligent computing resource node combination module 30 will be described in detail below. The intelligent computing resource node combination module 30 further includes: obtaining the resource requirement feature vector of each computing node in each computing task link; matching the resource requirement feature vector with the intelligent computing resource node space to obtain candidate intelligent computing resource nodes for each computing node; combining the candidate intelligent computing resource nodes of each computing node according to each computing task link to output candidate collaborative intelligent computing resource units; and performing conflict optimization on the multiple candidate collaborative intelligent computing resource units of the multiple computing task links to output optimized candidate collaborative intelligent computing resource units.

[0057] The specific configuration of the fusion readiness index acquisition module 40 will be described in detail below. The fusion readiness index acquisition module 40 further includes: extracting the time synchronization characteristics, feature distribution characteristics, and normalized state characteristics of the acquired multi-source output data; calculating the time alignment degree based on the time synchronization characteristics, which characterizes the degree of alignment of the output results from different data sources on the time axis; calculating the spatial alignment degree based on the feature distribution characteristics, which characterizes the degree of similarity in the statistical distribution of the output results from different data sources; calculating the normalized state characteristics based on the normalized state characteristics, which characterizes the degree of uniformity in the units and coding rules of the output results from different data sources; and calling the fusion readiness evaluation model to evaluate the time alignment degree, spatial alignment degree, and normalized state characteristics, and outputting the fusion readiness index.

[0058] The specific configuration of the fusion readiness index acquisition module 40 will be described in detail below. The fusion readiness index acquisition module 40 further includes: acquiring the fusion computing task nodes of the multiple computing task links; inputting the computing task links of the fusion readiness evaluation model according to the fusion link object tags of the fusion computing task nodes; and reading the marked candidate collaborative intelligent computing resource units corresponding to the marked computing task links.

[0059] The following will describe in detail the specific configuration of the fusion readiness index acquisition module 40. The fusion readiness index acquisition module 40 further includes: calling a fusion readiness evaluation model to evaluate the time alignment, spatial alignment, and standardized state characteristics. The fusion readiness evaluation model includes learning parameters, which are obtained by training on historical data samples of multi-source environmental health monitoring of the intelligent computing center and the label information corresponding to the actual fusion readiness evaluation results. The fusion readiness evaluation model converges when the fusion loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result is less than a preset loss threshold.

[0060] The following will describe in detail the specific configuration of the fusion readiness index acquisition module 40. The fusion readiness index acquisition module 40 further includes: acquiring the temporal alignment loss, spatial alignment loss, and normalized state feature loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result; and performing fusion loss analysis based on the temporal alignment loss, spatial alignment loss, and normalized state feature loss to obtain the fusion loss.

[0061] The specific configuration of the health monitoring data calculation module 50 will be described in detail below. The health monitoring data calculation module 50 further includes: sorting the candidate collaborative intelligent computing resource units corresponding to each computing task link according to the fusion readiness index, obtaining the preferred collaborative intelligent computing resource unit for each computing task link; and obtaining multiple preferred collaborative intelligent computing resource units corresponding to multiple computing task links.

[0062] The intelligent control system for the intelligent computing center that combines environmental health information monitoring, provided in the embodiments of the present invention, can execute the intelligent control method for the intelligent computing center that combines environmental health information monitoring provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A smart control method for intelligent computing centers that combines environmental health information monitoring, characterized in that: The method includes: Obtain the intelligent computing resource node space that can be used for regulation in the intelligent computing center; Collect multi-source environmental health monitoring data, and generate multiple corresponding computing task links based on the multi-source environmental health monitoring data; The intelligent computing resource node space is used to combine intelligent computing resource nodes for each of the multiple computing task links to generate candidate collaborative intelligent computing resource units. The fusion readiness evaluation model is invoked to analyze multiple candidate collaborative intelligent computing resource units corresponding to the multiple computing task links, and to obtain the fusion readiness index of the multi-source output data obtained after processing the multi-source environmental health monitoring data based on each candidate collaborative intelligent computing resource unit. According to the fusion readiness index, multiple preferred collaborative intelligent computing resource units corresponding to the multiple computing task links are selected, and the intelligent computing center performs the calculation of the multi-source environmental health monitoring data based on the multiple preferred collaborative intelligent computing resource units.

2. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 1, characterized in that, Methods for acquiring intelligent computing resource node space that can be controlled within an intelligent computing center include: Obtain the basic resource parameters and operational performance parameters of each intelligent computing resource node in the intelligent computing center; Based on the aforementioned basic resource parameters, analyze the historical output consistency parameters of the same type of intelligent computing resource nodes; The basic resource parameters, operating performance parameters, and historical output consistency parameters are encoded to construct the node state vector corresponding to each intelligent computing resource node. The intelligent computing resource node space is constructed according to each intelligent computing resource node and its corresponding node state vector.

3. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 1, characterized in that, The multi-source environmental health monitoring data is analyzed according to the DAG (Directed Acyclic Graph) to generate multiple corresponding computational task links. Each computational task link includes at least a data preprocessing stage, a feature extraction stage, a correlation analysis stage, and a fusion computation stage.

4. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 1, characterized in that, The method of combining intelligent computing resource nodes for each of the multiple computing task links through the intelligent computing resource node space includes: Obtain the resource requirement feature vector of each computing node in each computing task chain; The resource demand feature vector is matched with the intelligent computing resource node space to obtain candidate intelligent computing resource nodes for each computing node. Based on each computing task link, candidate intelligent computing resource nodes for each computing node are combined to output candidate collaborative intelligent computing resource units; Conflict optimization is performed on multiple candidate collaborative intelligent computing resource units in the multiple computing task links, and the optimized candidate collaborative intelligent computing resource units are output.

5. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 1, characterized in that, The method for obtaining a fusion readiness index of the multi-source output data obtained after processing the multi-source environmental health monitoring data based on each candidate collaborative intelligent computing resource unit includes: Extract the time synchronization characteristics, feature distribution characteristics, and standardized state characteristics of the multi-source output data; The time alignment degree is calculated based on the time synchronization feature, and the time alignment degree is used to characterize the degree of alignment of the output results of different data sources on the time axis. The spatial alignment is calculated based on the feature distribution characteristics, and the spatial alignment is used to characterize the degree of similarity in the statistical distribution of output results from different data sources. The standardized state features are calculated based on the standardized state features, and the spatial alignment is used to characterize the degree of uniformity in the units and coding rules of the output results from different data sources. The fusion readiness evaluation model is invoked to evaluate the time alignment, spatial alignment, and normalized state characteristics, and the fusion readiness index is output.

6. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 1, characterized in that, After generating multiple computational task chains based on the multi-source environmental health monitoring data, the method further includes: Obtain the fusion computing task node of the multiple computing task links; The fusion link object is input into the fusion readiness evaluation model according to the fusion computing task node's fusion link object tag, and the tag candidate collaborative intelligent computing resource unit corresponding to the tagged computing task link is read.

7. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 5, characterized in that, The fusion readiness evaluation model is invoked to evaluate the time alignment, spatial alignment, and standardized state characteristics. The fusion readiness evaluation model includes learning parameters, which are obtained by training on historical data samples of multi-source environmental health monitoring of the intelligent computing center and label information corresponding to the actual fusion readiness evaluation results. Specifically, the fusion readiness evaluation model converges when the fusion loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result is less than a preset loss threshold.

8. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 7, characterized in that, The method for calculating the labeled loss between the predicted fusion readiness evaluation result and the actual fusion readiness evaluation result includes: Obtain the temporal alignment loss, spatial alignment loss, and normalized state feature loss between the predicted fusion readiness evaluation results and the actual fusion readiness evaluation results; The fusion loss is obtained by performing fusion loss analysis based on the temporal alignment loss, spatial alignment loss, and normalized state feature loss.

9. The intelligent control method for a computing center combining environmental health information monitoring as described in claim 1, characterized in that, The method for selecting multiple preferred collaborative intelligent computing resource units corresponding to the multiple computing task links according to the aforementioned fusion readiness index includes: The candidate collaborative intelligent computing resource units corresponding to each computing task link are sorted according to the fusion readiness index to obtain the preferred collaborative intelligent computing resource unit for each computing task link. Multiple optimal collaborative intelligent computing resource units corresponding to multiple computing task links are obtained.

10. An intelligent control system for a computing center integrating environmental health information monitoring, characterized in that: The system is used to implement the intelligent control method for a smart computing center that combines environmental health information monitoring as described in any one of claims 1 to 9, and the system includes: The resource node space acquisition module is used to acquire the intelligent computing resource node space that the intelligent computing center can use for regulation. The monitoring data acquisition module is used to collect multi-source environmental health monitoring data and generate multiple corresponding computing task links according to the multi-source environmental health monitoring data; The intelligent computing resource node combination module is used to combine intelligent computing resource nodes for each computing task link of the multiple computing task links through the intelligent computing resource node space to generate candidate collaborative intelligent computing resource units. The fusion readiness index acquisition module is used to call the fusion readiness evaluation model to analyze multiple candidate collaborative intelligent computing resource units corresponding to the multiple computing task links, and obtain the fusion readiness index of the multi-source output data obtained after each candidate collaborative intelligent computing resource unit processes the multi-source environmental health monitoring data. The health monitoring data calculation module is used to select multiple preferred collaborative intelligent computing resource units corresponding to the multiple computing task links according to the fusion readiness index, and the intelligent computing center performs the calculation of the multi-source environmental health monitoring data based on the multiple preferred collaborative intelligent computing resource units.

Images