Data regulation and control method and device in intelligent building data platform, equipment and medium

Abstract

A data regulation and control method, device, equipment and medium in an intelligent building data platform relates to the technical field of intelligent buildings, and comprises the steps of performing feature extraction on target multi-source data to generate a target feature vector, the target feature vector comprising a time sequence feature, an equipment state feature, an environment feature and an event feature; data demand prediction is carried out on the target feature vector based on a preset target demand prediction model to output a target data demand, and the target data demand comprises data demand occurrence time, a data demand scale and a data demand type; determining a target load balancing algorithm according to the target data demand and the real-time system load state; and performing data regulation and control through the target load balancing algorithm, the real-time system load state and the target data demand. Through the application, reasonable allocation and optimization of resources can be effectively realized, the user experience is improved, manual intervention is not needed, and the operation and maintenance cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent buildings, and in particular to a data regulation method and device in an intelligent building data platform, equipment and a medium. BACKGROUND

[0002] With the complication of intelligent building data systems, the demand for data regulation is increasingly diversified and dynamically changing, so that the system needs to complete data collection, decision-making and execution in a short time, and quickly adjust the resource allocation strategy. In related technologies, the traditional intelligent building data platform mainly relies on manual experience to realize data regulation, which not only has high operation and maintenance cost, but also lacks dynamic adjustment capability, and cannot cope with dynamic data demand, so that it cannot reasonably allocate and optimize resources, which may cause local overload or resource idling and high delay, etc., thereby causing resource waste or affecting user experience.

[0003] Therefore, how to effectively regulate data in an intelligent building data platform to reasonably allocate resources, improve user experience and reduce operation and maintenance cost is a problem to be solved at present. SUMMARY

[0004] The present application provides a data regulation method, device, equipment and medium in an intelligent building data platform, which can solve the technical problems of high operation and maintenance cost, resource waste and poor user experience caused by manual experience in the prior art.

[0005] In a first aspect, the embodiments of the present application provide a data regulation method in an intelligent building data platform, which comprises: performing feature extraction on target multi-source data to generate a target feature vector, the target feature vector comprising time sequence features, device state features, environment features and event features; performing data demand prediction on the target feature vector based on a preset target demand prediction model to output a target data demand, the target data demand comprising data demand occurrence time, data demand size and data demand type; determining a target load balancing algorithm according to the target data demand and a real-time system load state; performing data regulation through the target load balancing algorithm, the real-time system load state and the target data demand.

[0006] In combination with the first aspect, in an implementation mode, the determination of the target load balancing algorithm according to the target data demand and the real-time system load state comprises: determining target node information based on the data demand type, the target node information comprising node type and node quantity; Based on the target node information and the real-time system load status, a target load balancing algorithm is selected from a preset set of load balancing algorithms. The load balancing algorithm set includes static load balancing algorithms and dynamic load balancing algorithms. The dynamic load balancing algorithms include the least connections algorithm and the dynamic feedback weighted algorithm.

[0007] In conjunction with the first aspect, in one implementation, the real-time system load status includes the real-time load of nodes, and the data demand scale includes the future load of nodes. When the target load balancing algorithm is a dynamic feedback weighted algorithm, the data regulation through the target load balancing algorithm, the real-time system load status, and the target data demand includes: The target weight of each node is calculated based on the node’s future load, the node’s real-time load, and the node’s theoretical maximum processing capacity. Target bandwidth is allocated to each node according to the target weight, so as to perform data regulation through the target bandwidth.

[0008] In conjunction with the first aspect, in one implementation method, the formula for calculating the target weight is:

[0009] In the formula, This represents the target weight of the i-th node. This represents the theoretical maximum processing capacity of the i-th node. Indicates the load adjustment coefficient. This represents the real-time load of the i-th node. This represents the future load of the i-th node.

[0010] In conjunction with the first aspect, in one implementation, the target data requirement further includes a prediction result confidence level. After the step of predicting the data requirement of the target feature vector based on a preset target requirement prediction model to output the target data requirement, the method further includes: The target confidence interval of the prediction result is determined based on the preset confidence interval. The data requirement scale in the target data requirement is updated according to the target adjustment amount corresponding to the target confidence interval to obtain a new target data requirement, and the step of determining the target load balancing algorithm based on the target data requirement and the real-time system load status is executed based on the new target data requirement.

[0011] In conjunction with the first aspect, in one implementation method, the data control process further includes: Obtain performance metrics and user feedback information for the target node; The target load balancing algorithm is updated based on the performance metrics and user feedback information to achieve data regulation updates.

[0012] In conjunction with the first aspect, in one implementation, the target demand prediction model includes an input layer, a first LSTM layer, a second LSTM layer, an attention layer, and a fully connected layer connected in sequence. The input layer is used to receive the target feature vector, the first LSTM layer is used to capture short-term temporal dependencies based on the target feature vector, the second LSTM layer is used to extract long-term periodic patterns based on the target feature vector, the attention layer is used to calculate the feature weights at each time step, and the fully connected layer is used to output the target data demand.

[0013] Secondly, embodiments of this application provide a data control device in a smart building data platform, the data control device in the smart building data platform comprising: The data processing module is used to extract features from the target multi-source data to generate a target feature vector, which includes time-series features, device status features, environmental features, and event features. The demand forecasting module is used to predict the data demand of the target feature vector based on a preset target demand forecasting model, so as to output the target data demand, which includes the data demand occurrence time, data demand scale and data demand type. The algorithm filtering module is used to determine the target load balancing algorithm based on the target data requirements and the real-time system load status. The data regulation module is used to regulate data based on the target load balancing algorithm, the real-time system load status, and the target data demand.

[0014] In conjunction with the second aspect, in one implementation, the algorithm filtering module is specifically used for: Based on the data requirement type, target node information is determined, including node type and number of nodes; Based on the target node information and the real-time system load status, a target load balancing algorithm is selected from a preset set of load balancing algorithms. The load balancing algorithm set includes static load balancing algorithms and dynamic load balancing algorithms. The dynamic load balancing algorithms include the least connections algorithm and the dynamic feedback weighted algorithm.

[0015] In conjunction with the second aspect, in one implementation, the real-time system load status includes the real-time load of nodes, and the data demand scale includes the future load of nodes. When the target load balancing algorithm is a dynamic feedback weighted algorithm, the data regulation module is specifically used for: The target weight of each node is calculated based on the node’s future load, the node’s real-time load, and the node’s theoretical maximum processing capacity. Target bandwidth is allocated to each node according to the target weight, so as to perform data regulation through the target bandwidth.

[0016] In conjunction with the second aspect, in one implementation method, the formula for calculating the target weight is:

[0017] In the formula, This represents the target weight of the i-th node. This represents the theoretical maximum processing capacity of the i-th node. Indicates the load adjustment coefficient. This represents the real-time load of the i-th node. This represents the future load of the i-th node.

[0018] In conjunction with the second aspect, in one implementation, the target data requirement further includes a prediction result confidence level, and the requirement prediction module is specifically used for: The target confidence interval of the prediction result is determined based on the preset confidence interval. The target data requirement scale is updated according to the target adjustment amount corresponding to the target confidence interval to obtain a new target data requirement, and the algorithm screening module executes the step of determining the target load balancing algorithm based on the target data requirement and the real-time system load status based on the new target data requirement.

[0019] In conjunction with the second aspect, in one implementation, during the data regulation process, the data regulation module is further used for: Obtain performance metrics and user feedback information for the target node; The target load balancing algorithm is updated based on the performance metrics and user feedback information to achieve data regulation updates.

[0020] In conjunction with the second aspect, in one embodiment, the target demand prediction model includes an input layer, a first LSTM layer, a second LSTM layer, an attention layer, and a fully connected layer connected in sequence. The input layer is used to receive the target feature vector, the first LSTM layer is used to capture short-term temporal dependencies based on the target feature vector, the second LSTM layer is used to extract long-term periodic patterns based on the target feature vector, the attention layer is used to calculate the feature weights at each time step, and the fully connected layer is used to output the target data demand.

[0021] Thirdly, embodiments of this application provide a data control device in a smart building data platform. The data control device in the smart building data platform includes a processor, a memory, and a data control program in the smart building data platform stored in the memory and executable by the processor. When the data control program in the smart building data platform is executed by the processor, it implements the steps of the aforementioned data control method in the smart building data platform.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium storing a data control program for a smart building data platform. When the data control program for the smart building data platform is executed by a processor, it implements the steps of the data control method for the smart building data platform as described above.

[0023] The beneficial effects of the technical solutions provided in this application include: By extracting features from target multi-source data, a target feature vector is generated, including time-series features, device status features, environmental features, and event features. Based on a target demand prediction model, the target feature vector is dynamically predicted to improve dynamic response capabilities. The output includes target data demand, including the time of data demand occurrence, data demand scale, and data demand type. Then, a target load balancing algorithm is determined based on the target data demand and the real-time system load status. Data regulation is then performed based on the target load balancing algorithm, the real-time system load status, and the target data demand to achieve reasonable allocation and optimization of resources, improve user experience, and reduce operation and maintenance costs without manual intervention. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an embodiment of the data control method in the smart building data platform of this application; Figure 2 For this application Figure 1 A detailed flowchart of step S30; Figure 3 For this application Figure 1 A detailed flowchart of step S40; Figure 4 This is a schematic diagram of the functional modules of the data control device embodiment in the smart building data platform of this application; Figure 5 This is a schematic diagram of the hardware structure of the data control device in the smart building data platform involved in the embodiments of this application. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0027] In a first aspect, embodiments of this application provide a data control method in a smart building data platform.

[0028] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the data control method in the smart building data platform of this application. Figure 1 As shown, the data control methods in the smart building data platform include: Step S10: Extract features from the target multi-source data to generate a target feature vector, which includes time-series features, device status features, environmental features, and event features.

[0029] In this exemplary embodiment, the smart building data platform first collects multi-source raw data, including data records, equipment operating status, and user behavior patterns, from various smart building systems (such as security, environmental monitoring, and energy management). Then, a data preprocessing module cleans, aligns, and standardizes the multi-source raw data to eliminate noise and unify the data format, yielding the target multi-source data. The data preprocessing module includes a data cleaning unit, a time alignment unit, and a normalization unit. The data cleaning unit primarily handles missing values ​​and outliers (such as sudden traffic spikes). For example, it uses a sliding window method to detect traffic outliers (e.g., a data point with a Z-score > 3 is considered an outlier; the Z-score measures the deviation of a data point from the mean) to identify outliers and fills in missing data using the mean of adjacent time periods. The time alignment unit aligns asynchronous data from different systems (such as security / energy) using a unified timestamp. The normalization unit normalizes multi-dimensional data (such as data volume and equipment load) to eliminate dimensional differences.

[0030] Then, feature engineering techniques are used to extract key features from the integrated multi-source data. That is, the feature engineering module is used to extract features from the target multi-source data to extract key features that are strongly related to the data requirements (such as data transmission frequency, data volume, time period distribution, equipment failure rate, etc.), thereby constructing the model input vector (i.e. the target feature vector). It should be noted that the key features strongly related to data requirements include, but are not limited to, the following four types of key features: (1) Time series features, such as historical traffic periodicity (hour / day / week), sliding window statistics (mean, variance), etc.; (2) Equipment status features, such as equipment online rate, CPU / memory utilization, fault history records, etc.; (3) Environmental features, such as building traffic (from access control system), energy consumption patterns (from electricity meter data), etc.; (4) Event features, such as preset event labels (such as holidays, large conferences); Therefore, by performing One-Hot encoding on all the extracted key features, the target feature vector can be obtained. That is, the target feature vector is encoded by key features such as time series features, equipment status features, environmental features and event features. For example, the target feature vector includes the average traffic of the past hour, the traffic of the same period last week, the current equipment load rate, the predicted traffic value and whether it is a working day, etc.

[0031] Step S20: Based on the preset target demand prediction model, perform data demand prediction on the target feature vector to output the target data demand, which includes the data demand occurrence time, data demand scale, and data demand type.

[0032] In this exemplary embodiment, the target demand prediction model refers to a demand prediction model that has already been trained. This model can employ time series analysis algorithms, machine learning algorithms (such as LSTM (Long Short-Term Memory), GRU (Gated Recurrent Unit), and other RNN (Recurrent Neural Network) models), or deep learning algorithms. For instance, a hybrid model combining LSTM and attention mechanisms can be used. Alternatively, the architecture of the demand prediction model can be determined based on actual needs. Then, features are extracted from historical multi-source data obtained through the smart building data platform to obtain historical feature vectors, including time-series features, equipment status features, environmental features, and event features. These historical feature vectors are then used as input to the demand prediction model, i.e., the model is trained based on these historical feature vectors to enable it to learn the complex relationships between data demands and various features. Once training is complete, the target demand prediction model can be generated.

[0033] Therefore, after obtaining the target feature vector, the target feature vector is input into the target demand prediction model. The target demand prediction model can then automatically predict future data demand based on the target feature vector and convert the prediction results into executable business indicators. The conversion process includes at least two key operations: (1) denormalization, which restores the predicted value to the actual data volume; (2) demand type classification, which classifies the demand type according to the source IP and port of the data traffic (such as security video stream / sensor data / energy control instructions, etc.). Through the above conversion operations, the target data demand, including the time of data demand occurrence, the scale of data demand, and the type of data demand, can be output.

[0034] It should be noted that the data demand occurrence time refers to the future time period corresponding to the occurrence of the data demand. The data demand scale includes the total data demand volume, the data demand volume corresponding to different data demand types (i.e., the future load of nodes), and their proportions. The data demand type refers to the demand type corresponding to the data stream type (e.g., security video stream data corresponds to security demand, sensor data corresponds to environmental monitoring demand, etc.). For example, if the target data demand is a total traffic value of 150GB within the next 30 minutes, with security demand accounting for 80% and environmental monitoring demand accounting for 20%, then the data demand occurrence time is the time period formed by adding 30 minutes to the current time. The data demand scale includes the total traffic value of 150GB, the security demand accounting for 80%, and the environmental monitoring demand accounting for 20%. The data demand type includes both security demand and environmental monitoring demand. It should be noted that the target data demand includes, but is not limited to, the data demand occurrence time, data demand scale, and data demand type. These parameters can also be adjusted according to actual needs, and are not limited here.

[0035] Furthermore, in one embodiment, the target demand prediction model includes an input layer, a first LSTM layer, a second LSTM layer, an attention layer, and a fully connected layer connected in sequence. The input layer is used to receive the target feature vector, the first LSTM layer is used to capture short-term temporal dependencies based on the target feature vector, the second LSTM layer is used to extract long-term periodic patterns based on the target feature vector, the attention layer is used to calculate the feature weights at each time step, and the fully connected layer is used to output the target data demand.

[0036] In this exemplary embodiment, a hybrid model combining LSTM and attention mechanisms is preferably used as the target demand prediction model. The target demand prediction model has a four-layer hierarchical structure, comprising an input layer, two stacked LSTM layers (i.e., a first LSTM layer and a second LSTM layer), an attention layer, and a fully connected layer. The input layer receives the target feature vector (its dimension equals the number of features; for example, if the number of features is 20, the dimension is 20). The two stacked LSTM layers are: the first LSTM layer preferably consists of 64 neurons, used for capturing short-term temporal dependencies (e.g., hourly fluctuations) based on the target feature vector; the second LSTM layer preferably consists of 32 neurons, used for extracting long-term periodic patterns (weekly patterns) based on the target feature vector to prevent overfitting. The attention layer calculates the feature weights at each time step to highlight key influencing factors (e.g., sudden surges in pedestrian traffic). The fully connected layer maps the outputs of the LSTM and attention layers to the prediction target, outputting the target data demand (e.g., traffic volume for the next 30 minutes and the proportion of traffic for different data demand types).

[0037] Specifically, let's take predicting data demand for the next 30 minutes as an example: Step N1: Multi-source data access, i.e., real-time collection of historical logs (in JSON format, etc.) from security cameras, temperature and humidity sensors, and energy management systems, as well as device status (using protocols such as SNMP (Simple Network Management Protocol)). For example, the following is an example of raw data in JSON format: { "Timestamp": "2024-07-20 14:00:00", Device ID: "CAM_001", Data size: 2.3GB CPU utilization: 65% }

[0038] Step N2: Data preprocessing, namely, the data cleaning unit removes invalid data (such as null value records when sensors are disconnected), the time alignment unit aggregates data from various systems to a unified time axis with a 5-minute time window, and the normalization unit compresses the data to the [0,1] interval. For example: normalized data result = (maximum value - minimum value) / (data size - minimum value), where the maximum value refers to the largest data within the time window, the minimum value refers to the smallest data within the time window, and the data size refers to the actual amount of data of the data object that needs to be normalized.

[0039] Step N3: Feature extraction and concatenation, which involves extracting four key features (time series / device / environment / event) from the preprocessed data to generate a 20-dimensional feature vector. For example, the feature vector is [0.72, 0.85, 0.61, 0.0, 1.0,...]. Each item in the feature vector corresponds to features such as traffic in the past hour, traffic in the same period last week, CPU load, non-holiday, and weekday.

[0040] Step N4: Model inference, which involves inputting the feature vector into the target demand prediction model. At this point, the first LSTM layer will capture short-term temporal dependencies, such as capturing the short-term upward trend of traffic caused by "current surge in traffic". The second LSTM layer 2 will extract long-term periodic patterns, such as identifying the long-term pattern of "periodic increase in traffic every Friday afternoon". The attention layer will calculate the feature weights at each time step, such as giving higher weights to the features "traffic" and "equipment load" (for example, if the average weight is 0.2, then the weight values ​​of "traffic" and "equipment load" will be set to 0.6). Finally, the fully connected layer will output the predicted values ​​for the next 6 time points (one point every 5 minutes).

[0041] Step N5: Demand analysis and output, which involves inversely normalizing the predicted value to obtain the actual traffic prediction value (e.g., the total traffic demand in the next 30 minutes is 150GB). The classifier identifies 80% of the data as video stream data (corresponding to security needs) and 20% as sensor data (corresponding to environmental monitoring needs) based on the port number, and then outputs the following structured prediction result: { "Time window": "14:30-15:00", Total bandwidth requirement: 150GB Data Requirement Type Distribution: {"Security Requirement": 80%, "Environmental Monitoring Requirement": 20%} } This completes the automatic prediction of data demand for the next 30 minutes.

[0042] Furthermore, in one embodiment, the target data requirement further includes a prediction result confidence level. After the step of predicting the data requirement of the target feature vector based on a preset target requirement prediction model to output the target data requirement, the method further includes: The target confidence interval of the prediction result is determined based on the preset confidence interval. The data requirement scale in the target data requirement is updated according to the target adjustment amount corresponding to the target confidence interval to obtain a new target data requirement, and the step of determining the target load balancing algorithm based on the target data requirement and the real-time system load status is executed based on the new target data requirement.

[0043] As an example, in this embodiment, during the process of converting the prediction results into actionable business metrics, a confidence assessment can also be performed to output the confidence level of the prediction results. The prediction results can then be fine-tuned based on the confidence level of the prediction results, thereby further improving the accuracy of the prediction results.

[0044] Specifically, first, confidence intervals are divided. For example, the overall confidence interval [0, 100%] is divided into 5 confidence intervals: confidence interval 1 is [0, 60%), confidence interval 2 is [60%, 70%), confidence interval 3 is [70%, 80%), confidence interval 4 is [80%, 90%), and confidence interval 5 is [90%, 100%]. Then, different adjustment amounts are set for different confidence intervals according to the level of confidence. For example, the adjustment amount corresponding to confidence interval 2 is ±20G. The adjustment amount corresponding to confidence interval 3 is ±15G, the adjustment amount corresponding to confidence interval 4 is ±10G, and the adjustment amount corresponding to confidence interval 5 is 0G. It should be noted that the above is only a presentation of an example, and the adjustment amount of different confidence intervals can be adaptively adjusted according to specific needs, which is not limited here. In addition, since the confidence level corresponding to confidence interval 1 is relatively low, it indicates that the accuracy or reliability of its prediction result is poor, so the prediction result needs to be discarded and re-predicted. Therefore, it is not necessary to set an adjustment amount for confidence interval 1.

[0045] Next, determine the target confidence interval where the prediction result's confidence level falls. For example, if the prediction result's confidence level is 80%, then confidence interval 4 is taken as its target confidence interval, and the adjustment amount ±10G corresponding to confidence interval 4 is taken as the target adjustment amount. Then, based on ±10G, the data requirement size in the target data requirement is updated to obtain a new target data requirement. For example, if the data requirement size is 150G, then the data requirement size in the new target data requirement will become 150G±10G. It should be noted that if the prediction result's confidence level is in confidence interval 1, then the data requirement size in the target data requirement is not adjusted, but the target data requirement is directly discarded, and the data requirement is predicted again.

[0046] Step S30: Determine the target load balancing algorithm based on the target data requirements and the real-time system load status.

[0047] As an example, this embodiment will utilize an adaptive strategy generation engine to automatically generate data control strategies. A data control strategy template can be built into the smart building data platform, which may include different data routing schemes, load balancing algorithms, data compression and encryption methods, etc., to adapt to different data control scenarios. Specifically, the load balancing algorithm, as a core strategy library component of the adaptive strategy generation engine, can dynamically adjust the load allocation strategy according to real-time data demands to optimize resource utilization and reduce latency. For example, in multi-node data control scenarios, data traffic can be dynamically allocated to the optimal node based on real-time load conditions (such as server CPU / memory utilization, network bandwidth, request queue length, etc.) to avoid single-point overload. Applicable scenarios include high-concurrency requests (such as burst transmission of security video streams), heterogeneous server clusters (i.e., mixed deployment of devices with different performance levels), and resource elastic scaling requirements (such as cloud-edge collaborative environments). Therefore, the algorithm selector in the adaptive strategy generation engine dynamically filters load balancing algorithms based on the timing, scale, and type of future data demands, as well as the current real-time system load status, to obtain the target load balancing algorithm. This addresses dynamic data demands, enabling reasonable resource allocation and optimization, thereby avoiding issues such as localized overload, resource idleness, and high latency, ultimately improving user experience. It should be noted that real-time system load status includes, but is not limited to, load volatility and node heterogeneity; these can be determined based on actual needs and are not limited here.

[0048] Further, see Figure 2 As shown, the step of determining the target load balancing algorithm based on the target data requirements and real-time system load status includes: Step S301: Determine the target node information based on the data requirement type, wherein the target node information includes node type and number of nodes; Step S302: Select a target load balancing algorithm from a preset set of load balancing algorithms based on the target node information and the real-time system load status; The load balancing algorithm set includes static load balancing algorithms and dynamic load balancing algorithms. The dynamic load balancing algorithms include the least connections algorithm and the dynamic feedback weighted algorithm.

[0049] For example, in this embodiment, the load balancing algorithm includes a static load balancing algorithm and a dynamic load balancing algorithm. The static load balancing algorithm includes round-robin and weighted round-robin, while the dynamic load balancing algorithm includes the least connection algorithm and the dynamic feedback weighted algorithm, thus forming a preset load balancing algorithm set. Among them, the templates corresponding to different load balancing algorithms are also different. Specifically, the round-robin template includes: (1) Rule: Distribute requests in the order of the node list. The node refers to the IoT gateway device responsible for the verification work in the area; (2) Applicable scenarios: Including scenarios such as uniform node performance (such as processing power, storage capacity, etc.), stable load, and low dynamic change requirements (i.e., low concurrency and simple structure (such as few nodes and types)). The weighted round-robin template includes: (1) Rule: Distribute requests according to the preset weight, i.e., the node with higher weight handles more traffic; Applicable scenarios: Including scenarios such as heterogeneous clusters with significant differences in node performance, stable load, and low dynamic change requirements.

[0050] The least connections template includes: (1) Rule: Assign new requests to the node with the fewest current connections; Applicable scenarios: Long connection services (such as video streaming). The dynamic feedback weighted template includes: (1) Rule: Calculate node weights dynamically based on the traffic characteristics predicted by the target demand prediction model and the real-time node load; (2) Applicable scenarios: Unstable load, high dynamic change demand (i.e., high concurrency and complex structure (such as many nodes and types)).

[0051] Based on this, this embodiment determines target node information, including node type and number, according to the data requirement type. This target node information reflects the dynamic changes in the current scenario's requirements; more node types and numbers indicate a more complex structure and higher concurrency, while fewer nodes indicate a simpler structure and lower concurrency. Real-time system load status reflects the current system's load stability. Therefore, by combining real-time system status and target node information, the optimal target load balancing algorithm can be dynamically and accurately selected from the set of load balancing algorithms. It should be noted that if a static load balancing algorithm is selected based on real-time system status and target node information, it can be randomly selected from round-robin and weighted round-robin. Further selection can also be based on node performance uniformity or other reference criteria, which are not limited here. If a dynamic load balancing algorithm is selected based on real-time system status and target node information, the choice between the least connection algorithm and the dynamic feedback weighted algorithm can be further determined based on whether the service is a long-connection service. That is, if it is a long-connection service, the least connection algorithm is selected; otherwise, the dynamic feedback weighted algorithm is selected.

[0052] Step S40: Perform data regulation based on the target load balancing algorithm, the real-time system load status, and the target data demand.

[0053] As an example, in this embodiment, after the target load balancing algorithm is determined, it will be pushed to the data execution layer to guide the actual data regulation process. Specifically, data regulation is carried out by taking into account the target load balancing algorithm, real-time system load status, data demand occurrence time, data demand scale, and data demand type to achieve reasonable allocation and optimization of resources, improve user experience, and effectively reduce operation and maintenance costs without manual intervention.

[0054] Further, in one embodiment, the real-time system load status includes the real-time load of nodes, and the data demand scale includes the future load of nodes. When the target load balancing algorithm is a dynamic feedback weighted algorithm, see [reference needed]. Figure 3 As shown, the data regulation based on the target load balancing algorithm, the real-time system load status, and the target data demand includes: Step S401: Calculate the target weight of each node based on the node's future load, the node's real-time load, and the node's theoretical maximum processing capacity; wherein, the formula for calculating the target weight is:

[0055] In the formula, This represents the target weight of the i-th node. This represents the theoretical maximum processing capacity of the i-th node. Indicates the load adjustment coefficient. This represents the real-time load of the i-th node. This represents the future load of the i-th node; Step S402: Allocate target bandwidth to each node according to the target weight, so as to perform data regulation through the target bandwidth.

[0056] As an example, it should be noted that the real-time system load status includes, but is not limited to, the real-time load of each node. When the target load balancing algorithm is a dynamic feedback weighted algorithm, for each node i, the weight calculation unit in the adaptive policy generation engine will calculate the weight based on the node's future load. (For example, the future load capacity of a node in a security system is 120GB), and the real-time load of the node. (For example, CPU utilization or bandwidth utilization is 70%), the node's theoretical maximum processing capacity. The target weight of node i is calculated using the following weight formula. :

[0057] It should be noted that the load adjustment coefficient This can be a default value, the specific value of which can be determined according to actual needs and is not limited here. For example, setting... Set to 0.7; in addition, the theoretical maximum processing capacity of each node is a known value that can be obtained directly, for example, the theoretical maximum processing capacity of node i is 10Gbps bandwidth.

[0058] After calculating the target weight for each node, the traffic allocation executor in the adaptive strategy generation engine determines the target bandwidth for the corresponding node based on the target weight, and allocates the corresponding target bandwidth to each node so as to distribute data requests to the corresponding node through the target bandwidth, thereby realizing data regulation of each node.

[0059] Based on this, the data control process is illustrated using the example of a sudden surge in video stream requests during the morning peak hours in a smart building security system: First, AI prediction is performed, automatically predicting a 200% increase in security video traffic to 130GB within the next 10 minutes using a target demand prediction model; second, strategy generation is performed, such as an adaptive strategy generation engine selecting a dynamic feedback weighting algorithm based on the AI ​​prediction results and calculating node weights, for example, increasing the weight of node A in the edge server from 0.5 to 0.8 to prioritize its processing of security video streams; based on the above data control, the node load balancing is optimized from 0.6 (unbalanced) to 0.9 (nearly balanced) and the average request response time is reduced from 120ms to 45ms.

[0060] It's worth noting that during data manipulation, strategies can be adjusted in real time. The feedback adjustment module in the adaptive strategy generation engine monitors the strategy's execution effect and dynamically adjusts it based on feedback. Specifically, when the feedback adjustment module detects that a node's CPU utilization exceeds a threshold (e.g., 85%), it triggers a dynamic feedback weighting algorithm to reduce the node's weight, thereby reducing new request allocation. If the node remains overloaded, the algorithm selector will switch the dynamic feedback weighting algorithm to the least connections algorithm to avoid a cascading failure effect.

[0061] Furthermore, this embodiment can also control the adaptive strategy generation engine to automatically and dynamically quantify and evaluate the applicability of the current target load balancing algorithm from multiple dimensions based on the output of the target demand prediction model, and fine-tune or recombine the target load balancing algorithm through optimization algorithms (such as genetic algorithms, simulated annealing, etc.) to generate the optimal load balancing algorithm. For applicability assessment, a comprehensive evaluation index system encompassing three core dimensions—timeliness, resource utilization, and fault tolerance—can be constructed by collecting real-time network status (e.g., link bandwidth, node load), data transmission performance indicators (e.g., latency, packet loss rate, throughput), and expected load characteristics output by the target demand prediction model. This system includes real-time data collection of network status (e.g., link bandwidth, node load), data transmission performance indicators (e.g., latency, resource utilization, fault tolerance), and entropy-weighted TOPSIS (a distance-based comprehensive evaluation method) algorithms to quantify the target load balancing algorithm. The weights of each index are dynamically allocated using the entropy-weighted method (e.g., increased latency weights in bursty traffic scenarios). Simultaneously, TOPSIS is used to calculate the closeness between the target load balancing algorithm and the ideal solution, quantifying the algorithm's suitability for the current environment. Finally, an algorithm performance feedback loop is established, using the execution results of the target load balancing algorithm as new training data to continuously optimize the parameter weights of the evaluation model, thus evolving the evaluation process from static rule-driven to dynamic experience-based learning. This mechanism, through the integration of data-driven and operations-based optimization, ensures that algorithm evaluation considers both real-time scenario characteristics and long-term system performance.

[0062] For load balancing algorithm optimization, optimization algorithms can achieve fine-tuning and recombination through dynamic iteration and algorithm evolution mechanisms. Specifically, genetic algorithms encode the parameters of the target load balancing algorithm as gene sequences (such as routing rule priorities, load thresholds, etc.), and generate new load balancing algorithms through population initialization, fitness evaluation (e.g., based on TOPSIS proximity or objective function weighting), gene crossover (exchanging algorithm fragments), and mutation (randomly perturbing parameters), and approach the global optimum through multiple generations of evolution. At the same time, simulated annealing uses the performance of the load balancing algorithm as "energy" (such as latency, energy consumption), generates algorithm variants through neighborhood search (e.g., adjusting compression ratio), and gradually converges to the local optimum according to the Metropolis criterion (i.e., probabilistically accepting inferior solutions) and annealing scheduling (e.g., temperature gradual decrease), so as to balance exploration and development. It can be seen that both genetic algorithms and simulated annealing, through a closed loop of generation, evaluation, and optimization, combined with reinforcement learning feedback on the historical algorithm effects, dynamically adjust the search direction, and finally output a combination of efficient load balancing algorithms that adapt to real-time network conditions and prediction needs, thereby achieving a deep integration of data-driven and operations research optimization, supporting the self-evolution capability of the load balancing algorithm library.

[0063] Furthermore, in one embodiment, the data manipulation process further includes: Obtain performance metrics and user feedback information for the target node; The target load balancing algorithm is updated based on the performance metrics and user feedback information to achieve data regulation updates.

[0064] As an example, in this embodiment, the availability of each node can be periodically checked through a health check mechanism to automatically identify potential problems. Specifically, the smart building data platform can achieve performance monitoring by deploying monitoring agents at key nodes, that is, collecting performance indicators such as latency, throughput, and error rate in real time during the data regulation process. At the same time, it can collect user satisfaction and feedback on the data regulation service through the user interface or API interface. Then, the performance indicators and user feedback information are input into the comprehensive analysis module to use artificial intelligence algorithms for multi-dimensional data mining and correlation analysis, thereby accurately identifying potential problems and improvement points. For example, this embodiment will analyze network latency, user complaints, and other data in real time and combine unsupervised learning (such as clustering algorithms) to detect abnormal patterns: if a node is detected to frequently experience a surge in latency during a specific period of time, accompanied by user feedback of "security screen lag", it will be correlated and inferred to be regional network congestion and traced back to defects in the bandwidth allocation strategy. In addition, traffic trends can be predicted during the data regulation process through time series models (such as LSTM). When the actual traffic data deviates significantly from the predicted value (such as the utilization rate of a link being consistently lower than expected), hidden problems such as routing configuration errors can be identified.

[0065] It is worth noting that, for the detected problems, this embodiment can combine causal reasoning (such as Bayesian networks) to analyze the root cause and update the target load balancing algorithm to ensure continuous optimization of the data regulation process. It should be noted that this update of the load balancing algorithm can be the selection of a new load balancing algorithm or triggering the target load balancing algorithm to recalculate weights; no limitation is made here. For example, if a resource conflict such as "high data compression rate causing CPU overload" is detected, an optimization scheme that dynamically adjusts the compression strategy or switches the load balancing algorithm will ultimately be generated. Furthermore, upon detecting a node failure through the health check mechanism, it will be automatically removed from the allocation list and an alarm notification will be triggered to the operation and maintenance system. Specifically, real-time monitoring and log analysis technologies will be used to automatically identify abnormal situations during data regulation, such as network interruptions, equipment failures, and data anomalies. Then, artificial intelligence technologies (such as decision trees and random forests) will be used to quickly locate and diagnose the fault to determine its root cause. In intelligent fault diagnosis, artificial intelligence technology will be used to perform feature matching and multi-dimensional correlation reasoning to quickly pinpoint the root cause of the fault. Finally, based on the fault diagnosis results, appropriate recovery measures will be automatically implemented or suggested, such as restarting the device, switching routes, and adjusting the data compression ratio, to restore data services as quickly as possible.

[0066] In summary, the entire process forms a closed loop of "monitoring-diagnosis-decision-making," enabling the smart building data platform to function as a "fault microscope" and a "strategy navigator." It can not only see the correlations behind complex data but also automatically plan the optimal improvement path, thereby achieving a leap from manual firefighting to intelligent prevention.

[0067] The following examples illustrate the diagnostic process using a video stream stuttering problem in a building security system as an example: (1) Data acquisition, i.e., real-time capture of multi-dimensional data during the fault period, such as network latency of camera nodes (e.g., 200ms), CPU load of adjacent switches (e.g., 90%), packet loss rate of transmission links (e.g., 15%), and user complaint tags (e.g., screen stuttering). (2) Feature mapping, i.e., converting the raw data into structured feature vectors and inputting them into a pre-trained random forest model (e.g., trained based on a historical fault database, which includes tags such as "device overload", "network congestion", and "configuration error"). (3) Decision reasoning, i.e., the model performs feature weight analysis, i.e., if high CPU load (e.g., weight 30%) and high packet loss (e.g., weight 25%) occur simultaneously, it is inferred that the queue overflow is caused by switch overload; for example, if packet loss is concentrated on a certain physical link (e.g., fiber optic F01) and accompanied by a surge in CRC (Cyclic Redundancy Check) error count, it is inferred that the hardware link is aging or the interface is not in good contact. (4) Root cause verification, which involves eliminating possibilities layer by layer through a decision tree rule chain. For example, if the delay returns to normal after switching to the backup link, it is confirmed that the original link is faulty, rather than an error in the upper-layer strategy. It can be seen that the above diagnostic process transforms fragmented data into interpretable diagnostic conclusions through feature correlation analysis and probabilistic classification. Compared with traditional manual investigation, it is more than 10 times more efficient and can discover implicit correlations (such as unconventional faults such as "encryption algorithm causing chip overheating"), and can reduce operation and maintenance costs.

[0068] In summary, this embodiment achieves dynamic optimization of smart building data control through a four-layer closed-loop architecture: First, it integrates multi-source data such as security and energy, and uses a target data demand prediction model to automatically predict future data demands; based on this, it adaptively filters and generates load balancing algorithms to dynamically adjust parameters such as routing and compression; furthermore, it monitors indicators such as latency / throughput in real time during execution and combines user feedback to implement closed-loop optimization strategies; when an anomaly is detected, it quickly locates the fault and triggers a self-healing mechanism (such as link switching / load adjustment), ultimately forming a complete intelligent closed loop of "data perception - AI decision-making - strategy execution - anomaly repair" to achieve high efficiency, adaptability, and fault tolerance in building data control. It is evident that this embodiment ensures optimal selection of data transmission paths through adaptive strategy generation and real-time monitoring, improving data transmission speed and efficiency, and enhancing the operational efficiency and service quality of various business operations within the building by optimizing the data transmission process, thereby improving user experience; moreover, the application of artificial intelligence technology enables the platform to automatically adjust data control strategies according to real-time demands, enhancing system flexibility and adapting to different business scenarios; in addition, intelligent fault diagnosis and recovery mechanisms reduce the need for manual intervention, thus lowering operation and maintenance costs.

[0069] Secondly, embodiments of this application also provide a data control device in a smart building data platform.

[0070] In one embodiment, reference is made to Figure 4 , Figure 4 This is a schematic diagram of the functional modules of an embodiment of the data control device in the smart building data platform of this application. Figure 4 As shown, the data control device in the smart building data platform includes: The data processing module is used to extract features from the target multi-source data to generate a target feature vector, which includes time-series features, device status features, environmental features, and event features. The demand forecasting module is used to predict the data demand of the target feature vector based on a preset target demand forecasting model, so as to output the target data demand, which includes the data demand occurrence time, data demand scale and data demand type. The algorithm filtering module is used to determine the target load balancing algorithm based on the target data requirements and the real-time system load status. The data regulation module is used to regulate data based on the target load balancing algorithm, the real-time system load status, and the target data demand.

[0071] Furthermore, in one embodiment, the algorithm filtering module is specifically used for: Based on the data requirement type, target node information is determined, including node type and number of nodes; Based on the target node information and the real-time system load status, a target load balancing algorithm is selected from a preset set of load balancing algorithms. The load balancing algorithm set includes static load balancing algorithms and dynamic load balancing algorithms. The dynamic load balancing algorithms include the least connections algorithm and the dynamic feedback weighted algorithm.

[0072] Further, in one embodiment, the real-time system load status includes the real-time load of nodes, and the data demand scale includes the future load of nodes. When the target load balancing algorithm is a dynamic feedback weighted algorithm, the data regulation module is specifically used for: The target weight of each node is calculated based on the node’s future load, the node’s real-time load, and the node’s theoretical maximum processing capacity. Target bandwidth is allocated to each node according to the target weight, so as to perform data regulation through the target bandwidth.

[0073] Furthermore, in one embodiment, the formula for calculating the target weight is:

[0074] In the formula, This represents the target weight of the i-th node. This represents the theoretical maximum processing capacity of the i-th node. Indicates the load adjustment coefficient. This represents the real-time load of the i-th node. This represents the future load of the i-th node.

[0075] Furthermore, in one embodiment, the target data requirement further includes a prediction result confidence level, and the requirement prediction module is specifically used for: The target confidence interval of the prediction result is determined based on the preset confidence interval. The target data requirement scale is updated according to the target adjustment amount corresponding to the target confidence interval to obtain a new target data requirement, and the algorithm screening module executes the step of determining the target load balancing algorithm based on the target data requirement and the real-time system load status based on the new target data requirement.

[0076] Furthermore, in one embodiment, during the data regulation process, the data regulation module is also used for: Obtain performance metrics and user feedback information for the target node; The target load balancing algorithm is updated based on the performance metrics and user feedback information to achieve data regulation updates.

[0077] Furthermore, in one embodiment, the target demand prediction model includes an input layer, a first LSTM layer, a second LSTM layer, an attention layer, and a fully connected layer connected in sequence. The input layer is used to receive the target feature vector, the first LSTM layer is used to capture short-term temporal dependencies based on the target feature vector, the second LSTM layer is used to extract long-term periodic patterns based on the target feature vector, the attention layer is used to calculate the feature weights at each time step, and the fully connected layer is used to output the target data demand.

[0078] The functions of each module in the data control device of the aforementioned smart building data platform correspond to the steps in the data control method embodiment of the aforementioned smart building data platform, and their functions and implementation processes will not be described in detail here.

[0079] Thirdly, embodiments of this application provide a data control device in a smart building data platform. The data control device in the smart building data platform can be a personal computer (PC), a laptop, a server, or other device with data processing capabilities.

[0080] Reference Figure 5 , Figure 5 This is a schematic diagram of the hardware structure of the data control device in the smart building data platform involved in the embodiments of this application. In this embodiment, the data control device in the smart building data platform may include a processor, a memory, a communication interface, and a communication bus.

[0081] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0082] Communication interfaces include input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal components of data control equipment within the smart building data platform, as well as interfaces used for interconnecting data control equipment with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0083] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0084] The processor can be a general-purpose processor, which can call the data control program in the smart building data platform stored in the memory and execute the data control method in the smart building data platform provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the data control program in the smart building data platform is called can refer to the various embodiments of the data control method in the smart building data platform of this application, and will not be repeated here.

[0085] Those skilled in the art will understand that Figure 5 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0086] Fourthly, embodiments of this application also provide a computer-readable storage medium.

[0087] The present application stores a data control program in a smart building data platform on a readable storage medium, wherein when the data control program in the smart building data platform is executed by a processor, it implements the steps of the data control method in the smart building data platform as described above.

[0088] The method implemented when the data control program in the smart building data platform is executed can be referred to in various embodiments of the data control method in the smart building data platform of this application, and will not be repeated here.

[0089] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0090] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0091] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0092] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0093] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0094] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0095] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A data control method in a smart building data platform, characterized in that, The data control methods in the smart building data platform include: Feature extraction is performed on the target multi-source data to generate a target feature vector, which includes time-series features, device status features, environmental features, and event features; Based on a preset target demand prediction model, the target feature vector is used to predict data demand, so as to output the target data demand, which includes the time of data demand occurrence, the scale of data demand, and the type of data demand. The target load balancing algorithm is determined based on the target data requirements and the real-time system load status. Data regulation is performed based on the target load balancing algorithm, the real-time system load status, and the target data requirements.

2. The data control method in the smart building data platform as described in claim 1, characterized in that, The step of determining the target load balancing algorithm based on the target data requirements and real-time system load status includes: Based on the data requirement type, target node information is determined, including node type and number of nodes; Based on the target node information and the real-time system load status, a target load balancing algorithm is selected from a preset set of load balancing algorithms. The load balancing algorithm set includes static load balancing algorithms and dynamic load balancing algorithms. The dynamic load balancing algorithms include the least connections algorithm and the dynamic feedback weighted algorithm.

3. The data control method in the smart building data platform as described in claim 2, characterized in that, The real-time system load status includes the real-time load of nodes, and the data demand scale includes the future load of nodes. When the target load balancing algorithm is a dynamic feedback weighted algorithm, the data regulation through the target load balancing algorithm, the real-time system load status, and the target data demand includes: The target weight of each node is calculated based on the node’s future load, the node’s real-time load, and the node’s theoretical maximum processing capacity. Target bandwidth is allocated to each node according to the target weight, so as to perform data regulation through the target bandwidth.

4. The data control method in the smart building data platform as described in claim 3, characterized in that, The formula for calculating the target weight is: In the formula, This represents the target weight of the i-th node. This represents the theoretical maximum processing capacity of the i-th node. Indicates the load adjustment coefficient. This represents the real-time load of the i-th node. This represents the future load of the i-th node.

5. The data control method in the smart building data platform as described in claim 1, characterized in that, The target data requirement also includes the confidence level of the prediction result. After the step of predicting the data requirement of the target feature vector based on the preset target data requirement prediction model to output the target data requirement, the method further includes: The target confidence interval of the prediction result is determined based on the preset confidence interval. The data requirement scale in the target data requirement is updated according to the target adjustment amount corresponding to the target confidence interval to obtain a new target data requirement, and the step of determining the target load balancing algorithm based on the target data requirement and the real-time system load status is executed based on the new target data requirement.

6. The data control method in the smart building data platform as described in claim 1, characterized in that, The process of data regulation also includes: Obtain performance metrics and user feedback information for the target node; The target load balancing algorithm is updated based on the performance metrics and user feedback information to achieve data regulation updates.

7. The data control method in the smart building data platform as described in claim 1, characterized in that: The target demand prediction model includes an input layer, a first LSTM layer, a second LSTM layer, an attention layer, and a fully connected layer connected in sequence. The input layer is used to receive the target feature vector. The first LSTM layer is used to capture short-term temporal dependencies based on the target feature vector. The second LSTM layer is used to extract long-term periodic patterns based on the target feature vector. The attention layer is used to calculate the feature weights at each time step. The fully connected layer is used to output the target data demand.

8. A data control device in a smart building data platform, characterized in that, The data control device in the smart building data platform includes: The data processing module is used to extract features from the target multi-source data to generate a target feature vector, which includes time-series features, device status features, environmental features, and event features. The demand forecasting module is used to predict the data demand of the target feature vector based on a preset target demand forecasting model, so as to output the target data demand, which includes the data demand occurrence time, data demand scale and data demand type. The algorithm filtering module is used to determine the target load balancing algorithm based on the target data requirements and the real-time system load status. The data regulation module is used to regulate data based on the target load balancing algorithm, the real-time system load status, and the target data demand.

9. A data control device in a smart building data platform, characterized in that, The data control device in the smart building data platform includes a processor, a memory, and a data control program in the smart building data platform stored in the memory and executable by the processor. When the data control program in the smart building data platform is executed by the processor, it implements the steps of the data control method in the smart building data platform as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a data control program for a smart building data platform, wherein when the data control program for the smart building data platform is executed by a processor, it implements the steps of the data control method for the smart building data platform as described in any one of claims 1 to 7.

Images