Artificial intelligence-based intelligent regulation and control method and system for building automation system

Abstract

The application provides an intelligent regulation and control method and system for a building automatic control system based on artificial intelligence, which collects real-time operation data of each device in the building automatic control system and real-time environment data of each functional area, constructs a target association cognitive network, and the target association cognitive network comprises device entity nodes, environment entity nodes and association relationships between entities. The network is used for situation recognition to generate a target situation feature set, and then the situation recognition is used to identify the current building operation situation type, a dynamic regulation and control scheme containing device regulation and control priority and parameter adjustment sequence is generated according to the situation type, the dynamic regulation and control scheme is converted into a control instruction and is executed, and the target association cognitive network is updated by collecting data after regulation and control, so that intelligent, dynamic and efficient regulation and control of the building automatic control system is realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to an intelligent control method and system for building automation systems based on artificial intelligence. Background Technology

[0002] In the field of traditional building automation systems, the operation of various devices within a building is typically managed using preset rules or simple threshold controls. For example, lighting equipment is switched on and off according to a fixed schedule, and air conditioning systems are started or stopped based on preset temperature thresholds. These control methods have many limitations.

[0003] On the one hand, preset rules and thresholds are often based on experience or specific scenarios, making it difficult to adapt to the complex and ever-changing environment and equipment status during actual building operation. The demands of buildings for environmental comfort and energy consumption vary greatly depending on the season, time of day, and the activities of different people; fixed rules cannot be flexibly adjusted to meet these dynamic needs.

[0004] On the other hand, traditional methods lack a deep understanding and analysis of the relationships between various devices within a building and between these devices and the environment. Each device operates independently, failing to form an organic whole, resulting in poor control effects, significant energy waste, and an inability to promptly identify and resolve potential building operation problems. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an intelligent control method for a building automation system based on artificial intelligence, the method comprising:

[0006] Collect real-time operating data of each device in the building automation system and real-time environmental data of each functional area;

[0007] A target association cognitive network is constructed based on the real-time operation data and real-time environment data. The target association cognitive network includes device entity nodes, environment entity nodes, and the relationship between entities.

[0008] The target association cognitive network is used to perform context recognition processing on the real-time operation data and real-time environment data to generate a target context feature set.

[0009] Based on the target context feature set and target association cognitive network, situational awareness processing is performed to identify the current building operation status type;

[0010] A dynamic control scheme is generated based on the current building operation status type and the target association cognitive network. The dynamic control scheme includes equipment control priority and parameter adjustment sequence.

[0011] The dynamic control scheme is converted into control commands for the building automation system and sent to the corresponding equipment to perform control operations. The control operation data and environmental data of the equipment after control are collected to update the target association cognitive network.

[0012] In another aspect, embodiments of the present invention also provide an intelligent control system for building automation systems based on artificial intelligence, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the machine-readable storage medium to implement the above-mentioned method.

[0013] Based on the above, this invention, through comprehensive collection of real-time operational data from various devices in the building automation system and real-time environmental data from various functional areas, constructs a target association cognitive network containing device entity nodes, environmental entity nodes, and inter-entity relationships. This enables comprehensive perception and correlation analysis of the building's operational status. The target association cognitive network is used for context recognition processing to generate a target context feature set, and further situational awareness processing is performed to accurately identify the current building operational status type. This allows for dynamic adaptation to the complex and ever-changing environment and device status during building operation, providing a reliable basis for precise control. Based on the current building operational status type, a dynamic control scheme containing device control priorities and parameter adjustment sequences is generated, enabling intelligent and orderly control of building equipment, effectively improving energy utilization efficiency and reducing energy consumption. Simultaneously, the collected and adjusted device operational data and environmental data update the target association cognitive network, allowing the system to continuously learn and optimize, continuously improving the accuracy and effectiveness of control, and ensuring the long-term stable, efficient, and comfortable operation of the building. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the execution flow of the intelligent control method for building automation systems based on artificial intelligence provided in this embodiment of the invention.

[0015] Figure 2 This is a schematic diagram of exemplary hardware and software components of the intelligent control system for building automation based on artificial intelligence provided in an embodiment of the present invention. Detailed Implementation

[0016] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating an intelligent control method for a building automation system based on artificial intelligence, provided in one embodiment of the present invention. The following is a detailed description of this intelligent control method for a building automation system based on artificial intelligence.

[0017] Step S110: Collect real-time operating data of each device in the building automation system and real-time environmental data of each functional area.

[0018] In this embodiment, a building automation system for a commercial complex is used as the application scenario. This complex includes multiple functional areas such as office areas, shopping mall areas, dining areas, and underground parking. The equipment involved includes central air conditioning units, fresh air exchangers, lighting controllers, elevators, water supply and drainage pumps, and fire-fighting linkage equipment. Data collection is completed through sensors deployed on each device and environmental monitoring terminals within the area. For real-time equipment operation data, the collected content includes device identification, operating status parameters, and operating time information. The operating status parameters vary depending on the device type. For example, the operating status parameters for central air conditioning units include compressor operating frequency, condenser temperature, evaporator temperature, supply air volume, and return air volume; the operating status parameters for fresh air exchangers include fan speed, fresh air volume, exhaust air volume, and filter pressure difference; the operating status parameters for lighting controllers include output voltage, output current, and lamp brightness; and the operating status parameters for elevators include operating speed, load, current floor, and door open / close status. For real-time environmental data of each functional area, the collected information includes area identification, environmental parameters, and collection time. Environmental parameters include temperature, relative humidity, carbon dioxide concentration, PM2.5 concentration, illuminance, and noise levels. During collection, data acquired by each sensor and monitoring terminal is transmitted to the data processing center via a data acquisition gateway according to a preset communication protocol. Encryption algorithms are used to encrypt the data during transmission to prevent data leakage and ensure data privacy and security.

[0019] Step S120: Construct a target association cognitive network based on the real-time operation data and real-time environment data. The target association cognitive network includes device entity nodes, environment entity nodes, and inter-entity relationships.

[0020] In this embodiment, the construction of the target association cognitive network needs to be carried out in multiple stages, including entity node extraction, association relationship analysis, association strength calculation, graph construction and verification, which is specifically achieved through the following sub-steps.

[0021] Step S121: Extract device identifier, operating status parameters and operating time information from the real-time operating data, treat each device identifier as a device entity node, and treat the operating status parameters and operating time information as attributes of the device entity node.

[0022] From the collected real-time operational data, the data parsing module extracts the device identifier for each device. This device identifier is a unique code used to distinguish different devices. Each device identifier is mapped to a device entity node in the target association cognitive network. Simultaneously, the extracted operating status parameters and operating time information of the device are stored as attributes of the device entity node. For example, the device entity node corresponding to the device identifier of a central air conditioning unit includes attributes such as the unit's current compressor operating frequency, condenser temperature, and operating time.

[0023] Step S122: Extract the area identifier, environmental parameters and collection time information from the real-time environmental data, treat each area identifier as an environmental entity node, and treat the environmental parameters and collection time information as attributes of the environmental entity node.

[0024] The collected real-time environmental data is analyzed to extract area identifiers for each functional area. These identifiers are unique codes corresponding to specific areas within the commercial complex, such as a specific room on a floor of an office building or a specific counter area in a shopping mall. Each area identifier is treated as an environmental entity node in a target-associative cognitive network, and the extracted environmental parameters and collection time information for that area are used as attributes of that environmental entity node. For example, the environmental entity node corresponding to the area identifier of a room in the office building would have attributes including the room's current temperature, relative humidity, carbon dioxide concentration, and collection time.

[0025] Step S123: Analyze the spatial relationship between device entity nodes and environment entity nodes, determine the environment entity node corresponding to each device entity node, and the device entity node associated with each environment entity node.

[0026] The system retrieves the equipment installation location information for each physical device node and the regional spatial range information for each environmental physical node from a spatial location database. It then matches the equipment installation location information with the regional spatial range information. If the installation location of a device falls within the regional spatial range corresponding to a certain environmental physical node, a spatial association is determined between the device physical node and that environmental physical node. For example, the physical device node corresponding to a fresh air exchanger installed in a room in an office area has a spatial association with the environmental physical node corresponding to that room; simultaneously, the physical device nodes associated with this environmental physical node also include nodes corresponding to other devices in that room, such as lighting controllers and indoor air conditioning units.

[0027] Step S124: Analyze the operational coordination relationship between equipment entity nodes, and identify pairs of equipment entity nodes that operate in the same functional area or have interlocking control relationships.

[0028] Based on the equipment's operational control logic and historical operation records, the operational coordination relationships between equipment entity nodes are analyzed. For equipment within the same functional area, if multiple devices start simultaneously, stop simultaneously, or exhibit synchronous changes in their operating status parameters, then a parallel operational coordination relationship exists between these equipment entity nodes. If a change in the operating status of one device triggers an adjustment in the operating status of another device, such as the start of a smoke exhaust fan in a fire-fighting linkage system triggering the opening of a corresponding fire damper, then an interlocking control coordination relationship exists between these two equipment entity nodes. For example, the central air conditioning indoor unit and its corresponding fan coil unit in a certain area of ​​a shopping mall typically operate simultaneously, with their operating status parameters changing synchronously; therefore, their corresponding equipment entity nodes constitute a pair of parallel operating equipment entity nodes. Similarly, in a fire protection system, a fire alarm and a sprinkler pump are interlocked; when the fire alarm detects a fire, it triggers the sprinkler pump to start, thus their corresponding equipment entity nodes constitute a pair of interlocking control equipment entity nodes.

[0029] Step S125: Analyze the environmental impact relationships between environmental entity nodes and identify the mutual influence of environmental parameters between environmental entity nodes in adjacent areas.

[0030] The spatial relationships between environmental entity nodes are obtained to identify adjacent environmental entity nodes. By analyzing the changing patterns of environmental parameters in adjacent areas from historical environmental data, if a change in the environmental parameter of one area causes a corresponding change in the environmental parameter of an adjacent area, then an environmental impact relationship exists between the environmental entity nodes of these two adjacent areas. For example, in an office area, if the air conditioning temperature in one room is lowered, the temperature in the other room will also decrease to some extent, thus indicating an environmental impact relationship. Similarly, in a shopping mall area, if the fresh air volume in the passageway area affects the carbon dioxide concentration in the counter area, then an environmental impact relationship also exists between them.

[0031] Step S126: Set relationship type labels for spatial association, operational collaboration, and environmental impact, respectively. The relationship type labels include spatial association labels, collaborative operation labels, and environmental impact labels.

[0032] For the three different types of relationships identified above, corresponding relationship type labels are assigned: spatial relationships correspond to spatial association labels, operational collaboration relationships correspond to collaborative operation labels, and environmental impact relationships correspond to environmental impact labels. These labels clearly identify the relationship types, facilitating the differentiation and processing of different types of relationships in the target association cognitive network. For example, connections between device entity nodes and environmental entity nodes are labeled with spatial association labels; connections between device entity nodes are labeled with collaborative operation labels based on the collaboration type; and connections between environmental entity nodes are labeled with environmental impact labels.

[0033] Step S127: Calculate the association strength value between different entity nodes based on historical operation data and historical environment data. The association strength value is used to represent the tightness of the association relationship.

[0034] Calculating the correlation strength value requires multi-dimensional analysis based on historical data, which can be achieved through the following sub-steps.

[0035] Step S1271: Collect historical operating data and historical environmental data within a preset time period. The historical operating data includes a sequence of historical operating status parameters of the device entity nodes, and the historical environmental data includes a sequence of historical environmental parameters of the environmental entity nodes.

[0036] Historical operational data and historical environmental data stored in the building automation system of the commercial complex over a period of time are collected. The historical operational data is classified by device identifier, forming a historical operational status parameter sequence corresponding to each device entity node. This historical operational status parameter sequence is arranged in chronological order and includes the operational status parameters of the device at different time points within a preset time period. The historical environmental data is classified by area identifier, forming a historical environmental parameter sequence corresponding to each environmental entity node. This historical environmental parameter sequence is also arranged in chronological order and includes the environmental parameters of the area at different time points within a preset time period.

[0037] Step S1272: Perform time alignment processing on the historical operating status parameter sequence and the historical environment parameter sequence so that the parameter data at the same point in time are correlated.

[0038] Timestamp information is extracted from historical operating status parameter sequences and historical environmental parameter sequences. Using a preset time interval as a benchmark, the two sequences are time-calibrated. Missing time point data is supplemented using interpolation; for duplicate time point data, the latest data is retained. Time alignment ensures a one-to-one correspondence between equipment operating status parameters and environmental parameters at the same time point.

[0039] Step S1273: For spatial correlation, select device entity nodes and environmental entity nodes within the same spatial region, calculate the correlation coefficient between the device operating state parameter sequence and the environmental parameter sequence, and use the absolute value of the correlation coefficient as the initial correlation strength value of the spatial correlation.

[0040] For device entity nodes and environmental entity nodes with spatial relationships, historical operating status parameter sequences and historical environmental parameter sequences, aligned with their time intervals, are selected, and the correlation coefficient between the two sequences is calculated. The correlation coefficient is calculated using statistical methods to measure the degree of linear correlation between the two sequences, and its value ranges from -1 to 1. The absolute value of the calculated correlation coefficient is taken as the initial correlation strength value for this spatial relationship; a larger absolute value indicates a stronger spatial correlation.

[0041] Step S1274: For the operational coordination relationship, select a pair of device entity nodes with coordinated operation characteristics, calculate the synchronization change rate of the operational state parameter sequence of the two device entity nodes, and use the synchronization change rate as the initial association strength value of the operational coordination relationship.

[0042] For a pair of device entity nodes that constitute an operational coordination relationship, extract their historical operational state parameter sequences after time alignment, and calculate the changes in the two sequences within the same time interval. The synchronization rate of change is obtained by calculating the consistency of the changes in the two sequences. If the direction and magnitude of change of the two sequences are similar in most time intervals, the synchronization rate of change is higher; otherwise, it is lower. The calculated synchronization rate of change is used as the initial association strength value of the operational coordination relationship. The higher the synchronization rate of change, the stronger the operational coordination between the two.

[0043] Step S1275: For the environmental impact relationship, select environmental entity nodes in adjacent spatial regions, calculate the lag correlation coefficient between environmental parameter sequences, and use the absolute value of the lag correlation coefficient as the initial correlation strength value of the environmental impact relationship. The lag correlation coefficient takes into account the time delay effect of environmental parameter changes.

[0044] For adjacent environmental entity nodes with an environmental impact relationship, historical environmental parameter sequences aligned with their respective timeframes are selected. Considering the time delay in the transmission of environmental parameters from one region to adjacent regions, different lag time windows are set, and the correlation coefficient between the two environmental parameter sequences is calculated within each lag time window; this is the lag correlation coefficient. The largest lag correlation coefficient is selected, and its absolute value is taken as the initial association strength value for this environmental impact relationship. A larger absolute value indicates a more significant mutual influence between the environmental parameters of adjacent regions.

[0045] Step S1276: Based on the preset correlation strength attenuation factor, attenuate the initial correlation strength value of the correlation relationship that has not undergone state change within the preset attenuation monitoring time window.

[0046] Set an attenuation monitoring time window and an association strength attenuation factor, with the attenuation factor being a value less than 1. For associations where the parameters of the two corresponding entity nodes do not change significantly within the attenuation monitoring time window, multiply their initial association strength value by the attenuation factor to attenuate the association strength. For example, if the parameters of a device entity node and an environmental entity node remain stable within the attenuation monitoring time window, the spatial association strength value between them will be reduced according to the attenuation factor.

[0047] Step S1277: Based on the preset association strength enhancement factor, enhance the initial association strength value of the association relationship in which the number of times the cooperative change occurs within the preset enhanced monitoring time window exceeds the preset frequency threshold.

[0048] The enhanced monitoring time window, frequency threshold, and correlation strength enhancement factor are set, with the enhancement factor being a value greater than 1. Within the enhanced monitoring time window, the number of times the parameters of two correlated entity nodes undergo coordinated changes is counted. If this number exceeds the preset frequency threshold, their initial correlation strength value is multiplied by the enhancement factor to enhance the correlation strength. For example, if a pair of device entity nodes undergoes multiple simultaneous changes in operating status within the enhanced monitoring time window, and the change trends are consistent, and the number of coordinated changes exceeds the frequency threshold, then the operational coordination strength value between the two will be adjusted upwards according to the enhancement factor.

[0049] Step S1278: Normalize the processed association strength value, and assign the normalized association strength value to the corresponding association relationship between entity nodes as the attribute value of the association relationship.

[0050] A normalization method is used to transform the association strength values, after attenuation and enhancement processing, to the range of 0 to 1. Normalization is achieved by subtracting the minimum value from each association strength value, and then dividing by the difference between the maximum and minimum values. The normalized association strength values ​​are then added as attributes to the corresponding association relationships between entity nodes, completing the calculation and assignment of association strength.

[0051] Step S128: Integrate device entity nodes, environment entity nodes, relationship type labels, and association strength values ​​to construct a preliminary association graph structure.

[0052] Using equipment entity nodes and environmental entity nodes as nodes in the graph, and different types of association relationships as edges between nodes, a preliminary association graph structure is constructed by labeling the edges with corresponding relationship type tags and normalized association strength values. This association graph structure is stored and managed through a graph database, supporting quick querying, adding, and modifying of nodes and edges. For example, the graph may contain central air conditioning unit nodes and office room nodes, connected by edges labeled with spatial association tags and association strength values; similarly, central air conditioning unit nodes and fan coil unit nodes may be connected by edges labeled with collaborative operation tags and association strength values.

[0053] Step S129: Set a dynamic adjustment mechanism for the preliminary correlation map structure to update the correlation strength value according to changes in real-time operating data and real-time environmental data.

[0054] The dynamic adjustment mechanism comprises a real-time monitoring module and a strength update module. The real-time monitoring module continuously collects real-time operational data and real-time environmental data. When a change in the parameter of an entity node corresponding to a certain relationship is detected, the strength update module is triggered. Based on the parameter changes and referring to the attenuation and enhancement rules in steps S1276 and S1277, the strength update module adjusts the association strength value of the relationship in real time and synchronously updates the corresponding attribute values ​​in the graph. For example, when the operating status parameters of a device entity node associated with a certain environmental entity node change frequently, causing changes in the parameters of the environmental entity node accordingly, the dynamic adjustment mechanism will enhance the association strength value between the two.

[0055] Step S1210: Verify the integrity of the preliminary association graph structure using the graph verification rules, and determine the verified preliminary association graph structure as the initial version of the target association cognitive network.

[0056] The graph verification rules include node integrity rules, relationship integrity rules, and strength rationality rules. Node integrity rules check whether all collected devices and regions have been mapped to corresponding entity nodes. If any devices or regions are not mapped, the corresponding nodes need to be constructed. Relationship integrity rules check whether necessary relationships are missing between entity nodes. For example, whether spatial relationships are established between devices and environmental nodes within the same region, and whether environmental relationships are established between environmental nodes in adjacent regions. If any are missing, the relationships need to be added. Strength rationality rules check whether the association strength value is within the range of 0 to 1 and conforms to the actual association between entity nodes. If outliers are found, the values ​​need to be recalculated. After the above verifications confirm that there are no problems, this preliminary association graph structure is determined as the initial version of the target association cognitive network.

[0057] Step S130: Use the target association cognitive network to perform context recognition processing on the real-time running data and real-time environmental data to generate a target context feature set.

[0058] Based on the constructed target association cognitive network, context recognition is performed by combining real-time collected data, which is achieved through the following sub-steps.

[0059] Step S131: Extract the historical change patterns of environmental parameters of environmental entity nodes from the target association cognitive network, and determine the normal fluctuation range and focus change pattern of each environmental parameter.

[0060] For each environmental entity node, its environmental parameter sequence is extracted from historical data of the target-related cognitive network. The variation characteristics of this sequence are analyzed, including the amplitude, frequency, peak value, and trough value. Based on these characteristics, the normal fluctuation range of each environmental parameter is determined, that is, the numerical range in which the environmental parameter is located most of the time. At the same time, focused change patterns are identified, such as the pattern of temperature gradually rising before work in the morning, and the pattern of carbon dioxide concentration gradually rising during peak hours. For example, the normal fluctuation range of temperature for the environmental entity node in the office area is a specific range, and its change pattern is that it gradually rises to a set value in the morning on weekdays and remains stable, and then gradually decreases after get off work.

[0061] Step S132: Extract the historical change patterns of the operating status parameters of the device entity nodes from the target association cognitive network, and determine the normal operating range and focused operating mode of each operating status parameter.

[0062] For each device node, its historical operating status parameter sequence is extracted, and the variation pattern of this sequence is analyzed to determine the normal operating range of each operating status parameter, i.e., the value range of the parameter during normal operation of the equipment. Simultaneously, focused operating modes are identified, such as the high-frequency operation mode of the compressor in a central air conditioning unit during the high-temperature period in summer, or the high-frequency start-stop mode of an elevator during peak commuting hours. For example, the normal operating range of the fan speed of a fresh air exchanger is a specific range, and its operating mode is high speed operation during the day and low speed operation at night.

[0063] Step S133: Compare the environmental parameters in the real-time environmental data with the normal fluctuation range of the corresponding environmental entity nodes to identify abnormal environmental parameters and the degree of abnormality.

[0064] The environmental parameters of each environmental entity node collected in real time are compared one by one with the normal fluctuation range determined in step S131. If a real-time environmental parameter exceeds the normal fluctuation range, the parameter is determined to be an abnormal environmental parameter. The degree of abnormality is determined by calculating the ratio of the difference between the real-time parameter value and the boundary value of the normal fluctuation range to the normal fluctuation range; the larger the ratio, the higher the degree of abnormality. For example, if the real-time temperature of a room in an office area exceeds the upper limit of the normal fluctuation range, the ratio of this excess value to the normal fluctuation range is calculated, and the degree of abnormality is determined as mild, moderate, or severe based on the ratio.

[0065] Step S134: Compare the operating status parameters in the real-time operating data with the normal operating range of the corresponding device entity node to identify abnormal operating parameters and the degree of abnormality.

[0066] Using a method similar to step S133, the real-time collected operating status parameters of each device node are compared with the normal operating range determined in step S132. If a real-time operating status parameter exceeds the normal operating range, it is determined to be an abnormal operating parameter. The degree of abnormality is determined by calculating the proportion of the difference between the real-time parameter value and the boundary value of the normal operating range to the normal operating range; the larger the proportion, the higher the degree of abnormality. For example, if the compressor operating frequency of a central air conditioning unit exceeds the lower limit of the normal operating range, the degree of abnormality is determined by calculating the proportion of this difference to the normal operating range.

[0067] Step S135: Based on the spatial association relationship in the target association cognitive network, perform association queries on the environmental entity nodes corresponding to the abnormal environmental parameters and the associated device entity nodes to determine the set of device entity nodes that may affect the environmental entity node.

[0068] In the target association cognitive network, all device entity nodes connected to the environmental entity node corresponding to the abnormal environmental parameter are found according to the spatial association label. Since these device entity nodes are spatially associated with the environmental entity node, changes in their operating status may cause abnormal environmental parameters. Therefore, the above device entity nodes are combined into a set of device entity nodes that may affect the environmental entity node.

[0069] Step S136: Based on the operational collaboration relationship in the target association cognitive network, perform association queries on the device entity node corresponding to the abnormal operation parameter and other associated device entity nodes to determine the set of collaborative device entity nodes that may be affected by the device entity node.

[0070] In the target association cognitive network, the device entity node corresponding to abnormal operating parameters is located based on the collaborative operation tag, and all associated nodes labeled with the collaborative operation tag are queried. Since these associated nodes have operational collaborative relationships with the abnormal device entity node, changes in the operating status of the abnormal device may be transmitted to these associated nodes, causing their operating parameters to become abnormal. Therefore, these associated nodes are integrated into a set of potentially affected collaborative device entity nodes. For example, if the device entity node corresponding to a central air conditioning indoor unit is identified as an abnormal node due to abnormal compressor operating frequency, by querying its collaborative operation relationships, associated device entity nodes such as fan coil units and electric regulating valves can be found. These nodes together constitute the set of collaborative device entity nodes.

[0071] Step S137: Construct a preliminary scenario description vector based on the abnormal environment parameters, abnormal operating parameters and their associated set of device entity nodes and set of cooperating device entity nodes.

[0072] The process extracts the types, severity, and corresponding environmental entity node identifiers of abnormal environmental parameters; the types, severity, and corresponding device entity node identifiers of abnormal operating parameters; and the identifiers of each node in the device entity node set and the identifiers of each node in the cooperating device entity node set. This information is then converted into vector dimensions. Each dimension corresponds to a feature value for one piece of information. For example, the type of abnormal environmental parameter corresponds to one dimension, and the severity corresponds to another. Node identifiers are encoded and converted into feature values ​​for their respective dimensions, ultimately forming a multi-dimensional preliminary situation description vector. This preliminary situation description vector contains the core elements of the current abnormal situation and can comprehensively reflect the basic characteristics of the situation.

[0073] Step S138: Match the preliminary scenario description vector with a preset scenario template library, which contains feature vectors and scenario identifiers for various focused scenarios.

[0074] The matching process requires standardization and similarity calculation to achieve accurate matching, which is carried out through the following sub-steps.

[0075] Step S1381: Construct a preset scenario template library. The scenario template library contains a variety of focused scenarios. Each focused scenario corresponds to a scenario template. Each scenario template contains a scenario identifier, a feature vector dimension description, and a standard feature vector.

[0076] This study identifies common focal scenarios in the operation of commercial complex buildings, such as "air conditioning malfunction leading to abnormal regional temperatures," "insufficient fresh air volume leading to excessive carbon dioxide concentration," and "lighting controller malfunction leading to abnormal illuminance." A unique scenario identifier is assigned to each scenario. For each scenario, the dimensional composition of its feature vector is defined, including the types of abnormal parameters and associated nodes that must be included in the scenario. Standard feature vectors for corresponding scenarios are constructed based on historical typical case data. Scenario identifiers, dimensional descriptions, and standard feature vectors are integrated into scenario templates, and multiple scenario templates together form a scenario template library.

[0077] Step S1382: Extract feature values ​​of each dimension from the preliminary context description vector to determine the number of dimensions and the meaning of each dimension of the preliminary context description vector.

[0078] The initial scenario description vector is parsed, and the feature values ​​corresponding to each dimension are extracted one by one. At the same time, the total number of dimensions of the vector is recorded, and the specific meaning of each dimension is clarified. For example, one dimension represents the degree of abnormality of the abnormal temperature, and another dimension represents the identification code of the associated air conditioning equipment, etc., to ensure that the structure and content of the initial scenario description vector are clearly defined.

[0079] Step S1383: Select a subset of situation templates from the situation template library that are consistent with the number of dimensions and the meaning of the dimensions of the preliminary situation description vector. For each situation template in the situation template subset, obtain its standard feature vector and calculate the cosine similarity value between the preliminary situation description vector and the standard feature vector. The cosine similarity value is used to represent the degree of closeness between the directions of the two vectors.

[0080] The process iterates through all scenario templates in the scenario template library, selecting those with the same number of dimensions as the initial scenario description vector and whose dimensions correspond one-to-one, forming a scenario template subset. For each scenario template in this subset, its standard feature vector is extracted, and the similarity between the initial scenario description vector and the standard feature vector is calculated using a cosine similarity algorithm. The calculation is achieved by measuring the cosine value of the angle between the two vectors; the smaller the angle, the closer the cosine similarity value is to 1, indicating that the two vectors are more aligned and the corresponding scenario match is higher.

[0081] Step S1384: Record the cosine similarity value of each scenario template, sort them in descending order of similarity value, and select the scenario template at the top of the sort as the best matching template.

[0082] The cosine similarity value between the initial context description vector and each standard feature vector in the context template subset is recorded and stored. A sorting algorithm is used to sort all context templates in descending order of similarity value. The context template ranked first after sorting is the best matching template with the highest degree of matching with the current initial context description vector.

[0083] Step S1385: When the cosine similarity value of the best matching template is greater than the preset similarity threshold, the context identifier of the best matching template is taken as the current context type; when the cosine similarity value of the best matching template is less than or equal to the preset similarity threshold, the current context type is marked as an unknown context, and the preliminary context description vector is added to the list to be updated in the context template library.

[0084] A similarity threshold is set, based on historical matching data. The cosine similarity value of the best-matching template is compared with this threshold. If it is greater than the threshold, a match is considered successful, and the context of the best-matching template is identified as the current context type. If it is less than or equal to the threshold, the current context cannot match any existing contexts in the template library, and it is marked as an unknown context. The corresponding preliminary context description vector is stored in the context template library's pending update list for subsequent manual analysis or model self-learning to supplement new context templates.

[0085] Step S139: Based on the current situation type and associated entity node attributes, generate a target situation feature set containing situation identifier, related entity node list, abnormal parameter list, and abnormality level.

[0086] Based on the determined current scenario type, its corresponding scenario identifier is extracted. Simultaneously, the identifiers and attribute information of all relevant entity nodes involved in this scenario are collected to form a list of relevant entity nodes. All abnormal environmental parameters and abnormal operational parameters occurring in this scenario are summarized to form an abnormal parameter list. The abnormality level information corresponding to each abnormal parameter is then organized. Finally, the scenario identifier, the list of relevant entity nodes, the list of abnormal parameters, and the abnormality level information are integrated to generate a target scenario feature set. This target scenario feature set comprehensively and systematically presents the scenario characteristics of the current building operation.

[0087] Step S140: Based on the target context feature set and target association cognitive network, perform situational awareness processing to identify the current building operation status type.

[0088] By combining the target context feature set with the target association cognitive network, situational awareness is developed from multiple aspects such as context association, temporal changes, and equipment load. This is achieved through the following sub-steps.

[0089] Step S141: Extract each scenario identifier and its corresponding list of related entity nodes from the target scenario feature set to determine the distribution location and influence range of each scenario in the building space.

[0090] The target context feature set is analyzed to extract all context identifiers and a list of related entity nodes corresponding to each context identifier. Based on the device and environmental entity node identifiers in the list of related entity nodes, the spatial location attributes of these nodes in the target association cognitive network are queried to determine the specific building area corresponding to each context, i.e., its distribution location. Based on the spatial range information and association relationships of the nodes, the surrounding areas that the context may affect are analyzed to determine its scope of influence. For example, a context of "air conditioning malfunction causing abnormal temperature in the area" might be located in a room in an office area, and its scope of influence might include that room and two adjacent rooms.

[0091] Step S142: Based on the environmental influence relationships in the target association cognitive network, analyze the mutual influence paths and influence intensity between different scenarios.

[0092] In the target association cognitive network, the relationships between environmental entity nodes corresponding to different scenarios are found based on environmental impact labels. If environmental entity nodes of two scenarios have an environmental impact relationship, it is determined that there is a mutual influence between the two scenarios. Then, the association links from environmental entity nodes in one scenario to environmental entity nodes in another scenario are identified, i.e., the mutual influence paths. Based on the association strength value of the environmental impact relationship, the strength of the mutual influence between different scenarios is determined; the higher the association strength value, the stronger the influence.

[0093] Step S143: Construct a context association network based on the influence range and mutual influence paths of each context. The nodes of the context association network are context identifiers, and the edges are the influence relationships and influence strengths between contexts.

[0094] Each context identifier in the target context feature set serves as a node in the context association network. For two context nodes that have a mutual influence relationship, an edge is used to connect them, and the corresponding influence relationship type and influence strength value are labeled on the edge, thus constructing the context association network. This context association network can intuitively display the interaction relationships between various contexts, facilitating the analysis of the diffusion and transmission patterns of contexts.

[0095] Step S144: Perform time series analysis on the context association network to extract context sequence change features within a continuous time period. The sequence change features include the order of context occurrence, duration, and transition frequency.

[0096] A continuous time period is selected, and the order in which each situation node appears in the situation association network within that time period is recorded, i.e., the situation appearance order. The duration of each situation from its appearance to its disappearance is calculated, i.e., the duration. The ratio of the number of transitions between different situations to the duration is calculated, i.e., the transition frequency. The situation appearance order, duration, and transition frequency are integrated into a sequence variation feature, which can reflect the dynamic change pattern of situations in the time dimension.

[0097] Step S145: Based on the attributes of device entity nodes in the target association cognitive network, analyze the device operating load status corresponding to each scenario, and calculate the overall device load rate and load distribution balance.

[0098] The system extracts operational status parameters of device entity nodes corresponding to each scenario from the target association cognitive network, such as actual operating power, rated power, and operating time. The overall device load rate is obtained by calculating the ratio of the sum of the actual operating power of all relevant devices to the sum of their rated power. Load distribution balance is obtained by calculating the standard deviation of the device load rate in each region; a smaller standard deviation indicates a more balanced load distribution, and vice versa.

[0099] Step S146: Combine the context association network, sequence change features, overall device load rate and load distribution balance to construct a situation description matrix. The rows of the situation description matrix represent the situation type, the columns represent the situation assessment dimensions, and the element values ​​represent the assessment values ​​of the situation type in the corresponding dimension.

[0100] Define the dimensions for situation assessment, including the size of the impact, the intensity of the impact, the duration, the frequency of transitions, the load rate, and the quality of load balancing. Using the situation types in the situation association network as the rows of a matrix and the situation assessment dimensions as the columns, construct a situation description matrix by assigning the specific values ​​for each situation type across each assessment dimension as the element values. For example, a situation type might have a specific assessment value for the "size of the impact" dimension and another specific assessment value for the "load rate" dimension; these values ​​together constitute the corresponding elements in the matrix.

[0101] Step S147: Input the situation description matrix into the preset situation recognition model. The situation recognition model learns from historical situation samples to output the membership values ​​of multiple preset situation types, and selects the preset situation type with the highest membership value as the current building operation situation type.

[0102] The specific process of constructing, training, and applying the situation recognition model is as follows.

[0103] Step S1471: Construct a preset situation recognition model, which includes an input layer, a feature mapping layer, an association analysis layer, and an output layer. The input layer receives the situation description matrix and converts it into a matrix vector form that the model can process. The feature mapping layer includes multiple parallel feature extractors, each of which extracts key features from different dimensions of the situation description matrix to generate a high-dimensional feature vector. The association analysis layer includes an association processor based on a graph attention mechanism, which models the inter-entity relationships of the high-dimensional feature vector and strengthens the weight of important features. The output layer includes multiple neurons, each corresponding to a preset situation type, and outputs the membership value of that situation type.

[0104] The input layer employs a fully connected structure, flattening the input situation description matrix into a one-dimensional vector for subsequent model processing. The feature mapping layer uses multiple independent feature extractors, each extracting features from one evaluation dimension of the situation description matrix. Through convolution operations and activation functions, the basic features of each dimension are converted into high-dimensional features, which are then concatenated to form a high-dimensional feature vector. The graph attention mechanism in the association analysis layer distinguishes the importance of different features by calculating attention weights between features, assigning higher weights to important features to enhance their impact on situation recognition. The output layer uses a softmax activation function, with each neuron corresponding to a preset situation type, such as "stable operating situation," "mildly abnormal situation," "moderately abnormal situation," and "severely abnormal situation." The output value represents the membership value of the current situation to that type.

[0105] Step S1472: Collect historical situation sample data, which includes a historical situation description matrix and corresponding actual situation type labels.

[0106] Collect building operation data for the commercial complex over a past period. For each historical operation period, construct a corresponding historical situation description matrix according to steps S141 to S146. Based on the actual operation during that period, professionals label the data with corresponding actual situation type tags, forming historical situation sample data. The historical situation sample data must cover multiple different situation types to ensure the diversity and representativeness of the sample.

[0107] Step S1473: Divide the historical situation sample data into a training sample set and a validation sample set.

[0108] A random partitioning method is used to divide historical situational sample data into a training sample set and a validation sample set according to a certain ratio. The training sample set is used for model parameter training, while the validation sample set is used for performance verification during model training. The partitioning ratio can be set according to the number of samples and model training requirements to ensure that the training sample set can provide sufficient learning data for the model, and the validation sample set can effectively evaluate the model's generalization ability.

[0109] Step S1474: Train the situation recognition model using the training sample set. Calculate the loss value between the membership value output by the model and the actual situation type label through forward propagation. Adjust the weight parameters of each layer of the model based on the loss value using the backpropagation algorithm until the situation recognition model's recognition accuracy on the validation sample set reaches a preset threshold. Save the parameters of the trained situation recognition model to form a preset situation recognition model that can be used for inference.

[0110] The historical situation description matrix from the training sample set is input into the situation recognition model. Through forward propagation calculations in the input layer, feature mapping layer, and association analysis layer, the membership values ​​for each preset situation type are obtained from the output layer. The cross-entropy loss function is used to calculate the loss between the output membership values ​​and the actual situation type labels. This loss value is then propagated back from the output layer to the input layer using a backpropagation algorithm. The weight parameters of each layer are adjusted based on the loss value, including the connection weights of the input layer, the convolution kernel parameters of the feature mapping layer, and the attention weights of the association analysis layer. During training, the model is periodically validated using a validation sample set. The recognition accuracy of the model on the validation sample set is calculated. When the accuracy reaches a preset threshold, training stops, and the model parameters at this point are saved, forming the preset situation recognition model.

[0111] Step S1475: Input the current situation description matrix into the trained situation recognition model. Through the processing of the input layer, feature mapping layer, and correlation analysis layer, the output layer outputs membership values ​​of various preset situation types. The membership values ​​represent the probability that the current situation belongs to each preset situation type.

[0112] The current situation description matrix constructed in step S146 is input into the preset situation recognition model. After being converted into vector form by the input layer, the current situation description matrix is ​​passed to the feature mapping layer, where multiple feature extractors extract key features of each dimension and generate high-dimensional feature vectors. The high-dimensional feature vectors are then passed to the association analysis layer, where important feature weights are strengthened through a graph attention mechanism. The processed feature vectors are then passed to the output layer, where they are calculated by the softmax activation function to output the membership value of each preset situation type. The membership value is between 0 and 1, and the sum of the membership values ​​of all preset situation types is 1. The larger the membership value, the higher the probability that the current situation belongs to that type.

[0113] Step S148: Generate situation awareness results that include the current building operation situation type, situation description matrix and membership value.

[0114] The preset situation type with the largest membership value among the outputs is selected as the current building operation situation type. This type is then integrated with the constructed situation description matrix and the membership values ​​of all preset situation types to form the situation perception result. This situation perception result comprehensively reflects the overall operation situation of the current building.

[0115] Step S150: Generate a dynamic control scheme based on the current building operation status type and target association cognitive network. The dynamic control scheme includes equipment control priority and parameter adjustment sequence.

[0116] Based on situational awareness results and target association awareness networks, dynamic control schemes are formulated from aspects such as equipment selection, strategy matching, and priority ranking. Specifically, this is achieved through the following sub-steps.

[0117] Step S151: Extract the set of equipment entity nodes and the set of environmental entity nodes related to the current building operation status from the target association cognitive network.

[0118] Based on the characteristics of the current building operation status, query the equipment entity nodes and environmental entity nodes related to this status in the target association cognitive network. For example, if the current status is "air conditioning failure causing abnormal temperature in the area", then query the air conditioning equipment entity nodes (such as air conditioning indoor units, compressors, regulating valves, etc.) and the corresponding environmental entity nodes (such as environmental nodes of this area and adjacent areas) related to the area with abnormal temperature, and integrate the above nodes into a set of equipment entity nodes and a set of environmental entity nodes respectively.

[0119] Step S152: Based on the current building operation status type, query the preset status-control strategy library to obtain a preliminary control strategy framework. The preliminary control strategy framework includes the suggested control equipment type and environmental parameter adjustment direction.

[0120] The pre-defined situation-control strategy library stores control strategies corresponding to different situation types, with each situation type associated with a preliminary control strategy framework. Based on the identifier of the current building operation situation type, a matching query is performed in the situation-control strategy library to obtain the corresponding preliminary control strategy framework. This preliminary control strategy framework clarifies the types of equipment to be controlled for this situation, such as air conditioning equipment and fresh air equipment, as well as the direction of adjustment for the environmental parameters that need to be adjusted, such as increasing temperature, decreasing humidity, or increasing fresh air volume.

[0121] Step S153: Match the relevant set of equipment entity nodes with the equipment types in the preliminary control strategy framework to determine the specific target control equipment list.

[0122] The process involves analyzing the suggested control equipment types within the initial control strategy framework, traversing each node in the set of equipment entity nodes, and determining whether its equipment type matches the suggested control equipment type. Nodes with matching types are then selected to form a specific target control equipment list. This target control equipment list includes the specific equipment identifiers and corresponding basic attribute information for which control operations are required.

[0123] Step S154: Analyze the operational coordination relationship of each equipment entity node in the target control equipment list, determine the equipment control sequence constraints, and generate a preliminary control sequence based on the equipment control sequence constraints and the operational status parameters of the equipment entity nodes.

[0124] In the target association cognitive network, the collaborative operation relationships between the entity nodes of each device in the target control equipment list are queried, and the control sequence constraints are determined according to the type of collaborative relationship. For example, for devices with interlocking control relationships, control should be performed according to the triggering order, first controlling the triggering device and then controlling the triggered device; for devices with parallel operation relationships, they can be set to be controlled simultaneously or controlled in a certain priority order. Combining the current operating status parameters of each device, such as operating power and load rate, the devices are sorted under the premise of satisfying the sequence constraints to generate a preliminary control sequence.

[0125] Step S155: Based on the association strength value in the target association cognitive network, calculate the influence weight of each target control device on the environmental entity node, use the influence weight as an important reference factor for control priority, and sort the target control devices by priority in combination with the preliminary control order to generate a device control priority sequence.

[0126] For each device in the target control equipment list, the association strength value of its spatial association with each node in the environmental entity node set is queried. The average association strength value of the device to all relevant environmental entity nodes is calculated, and this average value is used as the initial influence weight of the device on the environmental entity nodes. If a device is associated with multiple environmental entity nodes, the association strength values ​​of each device are summed and divided by the number of environmental entity nodes to obtain the average association strength value. If it is associated with only a single environmental entity node, the association strength value is directly used as the initial influence weight. Subsequently, the initial influence weight is adjusted based on the current operating load rate of the device. The higher the operating load rate, the smaller the adjustment coefficient. The final influence weight is obtained by multiplying the initial influence weight by the adjustment coefficient. The final influence weight is combined with the preliminary control order to reorder the target control equipment: under the premise of satisfying the control order constraints, devices with higher final influence weights are prioritized; if the final influence weights are the same, they are arranged according to the preliminary control order. After the above sorting, a device control priority sequence containing device identifiers and corresponding priority levels is formed.

[0127] Step S156: For each target control device, determine the parameter adjustment range based on the environmental parameter adjustment direction in the preliminary control strategy framework and the current environmental data. Refer to the parameter adjustment effect in the historical operating data of the target control device, determine the specific parameter adjustment step size and number of adjustments within the parameter adjustment range, and form a parameter adjustment sequence.

[0128] The determination of the parameter adjustment sequence needs to be accurately calculated by combining historical data on the adjustment effect, which is achieved through the following sub-steps.

[0129] Step S1561: Extract historical adjustment records similar to the current situation type from the historical operating data of the target control equipment. The historical adjustment records include parameter values ​​before adjustment, adjustment step size, number of adjustments, parameter values ​​after adjustment, and corresponding environmental parameter changes.

[0130] Based on the current scenario type identifier, historical adjustment records with the same or similar scenario types are filtered from the historical operation data of the target control equipment. Each historical adjustment record must include the operating parameter values ​​of the equipment before adjustment, the step size of each adjustment, the total number of adjustments, the operating parameter values ​​after adjustment, and the environmental parameter changes of the corresponding environmental entity nodes during the adjustment process, ensuring that the extracted historical records can comprehensively reflect the adjustment effect of the equipment under similar scenarios.

[0131] Step S1562: Filter the historical adjustment records and retain the effective adjustment records in which the changes in environmental parameters after adjustment meet the expected control targets.

[0132] Set the range of environmental parameter changes corresponding to the expected control target, and compare the environmental parameter change values ​​in each historical control record with this range. If the environmental parameter change value is within the expected range, it indicates that the control operation has achieved the expected effect, and the historical control record is retained as a valid control record; if the environmental parameter change value exceeds or does not reach the expected range, the historical control record is removed, and only valid control records are retained for subsequent analysis.

[0133] Step S1563: Extract the relationship between the adjustment step size and the rate of change of environmental parameters from the effective adjustment records, wherein the rate of change of environmental parameters is the ratio of the change value of environmental parameters to the adjustment duration.

[0134] For each valid adjustment record, the rate of change of environmental parameters is calculated, which is the change value of environmental parameters divided by the total duration of that adjustment operation. A correlation analysis is performed between the adjustment step size and the corresponding rate of change of environmental parameters in the valid adjustment records. A model of the correspondence between the two is established using a fitting method to clarify the trend of the rate of change of environmental parameters as the adjustment step size increases.

[0135] Step S1564: Based on the upper and lower limits of the parameter adjustment range and the current parameter value, determine the total adjustment amount, which is the difference between the current parameter value and the target parameter value.

[0136] Based on the environmental parameter adjustment direction in the preliminary control strategy framework, and combined with current environmental data, the target environmental parameter values ​​are determined. Then, based on these target environmental parameter values, the target parameter values ​​that the control equipment needs to achieve are deduced. The upper and lower limits of the parameter adjustment range are determined by the safe operating parameter range of the equipment, ensuring that the equipment parameters do not exceed the safety boundaries during adjustment. The total adjustment amount is obtained by subtracting the target parameter value from the current equipment operating parameter value. If the current parameter value is greater than the target parameter value, the total adjustment amount is negative, indicating that the parameter needs to be adjusted downwards; if the current parameter value is less than the target parameter value, the total adjustment amount is positive, indicating that the parameter needs to be adjusted upwards.

[0137] Step S1565: Based on the relationship between the adjustment step size and the rate of change of environmental parameters in the historical effective adjustment records, preset the initial adjustment step size.

[0138] Referring to the relationship model between the adjustment step size and the rate of change of environmental parameters established in step S1563, and combining the current required rate of change of environmental parameters (obtained by dividing the difference between the target environmental parameter value and the current environmental parameter value by the expected adjustment duration), the corresponding adjustment step size is selected as the initial adjustment step size within the allowable adjustment step size range of the equipment. If the required rate of change of environmental parameters is fast, a larger initial adjustment step size is selected; if the required rate of change is slow, a smaller initial adjustment step size is selected.

[0139] Step S1566: Based on the initial adjustment step size and the total adjustment amount, calculate the required number of adjustments, where the number of adjustments is the ratio of the total adjustment amount to the initial adjustment step size.

[0140] Divide the absolute value of the total adjustment by the absolute value of the initial adjustment step size to obtain the initial number of adjustments. If the result is an integer, use that integer as the initial number of adjustments; if the result is a decimal, round it up to obtain the initial number of adjustments, ensuring that the target parameter value can be reached or approached through this number of adjustments.

[0141] Step S1567: Refer to the historical adjustment records for the number of adjustments and the stability of the adjustment effect of the same equipment under similar total adjustment values, and correct the initial number of adjustments.

[0142] From the effective adjustment records, historical records with total adjustment values ​​similar to the current total adjustment value are selected. The number of adjustments in these records and the stability of the environmental parameters after adjustment are analyzed (measured by the fluctuation range of environmental parameters within a preset time period after adjustment). If a stable adjustment effect can be achieved with fewer adjustments for similar total adjustment values, the number of initial adjustments is appropriately reduced; if fewer adjustments lead to larger fluctuations in environmental parameters, the number of initial adjustments is appropriately increased to improve the stability of the adjustment effect.

[0143] Step S1568: Based on the corrected number of adjustments and the total adjustment amount, redetermine the final adjustment step size so that the total adjustment amount is equal to the product of the final adjustment step size and the number of adjustments.

[0144] Divide the total adjustment amount by the corrected number of adjustments to obtain the final adjustment step size. If the calculated final adjustment step size exceeds the allowable adjustment step size range of the equipment, then use the maximum or minimum allowable adjustment step size of the equipment as the final adjustment step size, and adjust the number of adjustments again to ensure that the total adjustment amount is consistent with the product of the final adjustment step size and the number of adjustments, while meeting the requirements for safe operation of the equipment.

[0145] Step S1569: Arrange the parameter value changes and corresponding adjustment time intervals for each adjustment in the adjustment order to form a parameter adjustment sequence that includes adjustment step size, number of adjustments and adjustment time interval.

[0146] Based on the final adjustment step size and the corrected number of adjustments, determine the parameter value change for each adjustment (i.e., the final adjustment step size), and combine this with the adjustment time intervals of similar adjustments in historical effective adjustment records to set the time interval between each adjustment. Arrange the parameter value changes and adjustment time intervals for each adjustment in chronological order to form the parameter adjustment sequence for the target control device. The sequence clearly indicates the specific parameter changes and execution time intervals for each adjustment.

[0147] Step S157: Integrate the equipment control priority sequence with the corresponding parameter adjustment sequence to generate a dynamic control scheme that includes equipment identification, control priority, parameter adjustment sequence and adjustment order constraints.

[0148] Each device identifier in the device control priority sequence is associated and matched with its corresponding parameter control sequence. The control priority level and control sequence constraints for that device are also added, and this information is integrated in a unified format. The dynamic control scheme is stored in a structured data format for easy conversion into control commands. Each device entry includes the device identifier, control priority, step size for each adjustment, number of adjustments, adjustment time interval, and control sequence constraints to be followed, ensuring the completeness and executability of the control scheme.

[0149] Step S160: The dynamic control scheme is converted into control commands for the building automation system and sent to the corresponding equipment to perform control operations. The control operation data and environmental data of the equipment after control are collected to update the target association cognitive network.

[0150] For example, the execution of dynamic control schemes and the updating of the cognitive network associated with goals are achieved through the following sub-steps.

[0151] Step S161: Convert the dynamic control scheme into control commands for the building automation system and send them to the corresponding equipment to perform the control operation.

[0152] Each device entry in the dynamic control scheme is analyzed, and the parameter adjustment sequence is converted into a control command format recognizable by the device according to the device type and communication protocol. The control command includes information such as device identification, adjustment step size, adjustment time, and execution sequence to ensure that the device can accurately identify and execute the control operation. The control command is sent to the corresponding target control device via the building automation system's communication bus. During the sending process, a data verification mechanism is used to verify the completeness and accuracy of the command. If the verification fails, the command is resent until it is successfully sent. After receiving the control command, the target control device executes the parameter adjustment operation step by step according to the command requirements.

[0153] Step S162: During the execution of the dynamic control scheme, real-time operating data of each target control device and real-time environmental data of the corresponding environmental entity nodes are collected at preset sampling intervals.

[0154] A preset sampling interval is set, which is determined based on the duration of the control operation and the frequency of parameter changes to ensure comprehensive capture of parameter changes during the control process. Through sensors deployed on the equipment and environmental monitoring terminals within the area, the operating status parameters of each target control device (such as adjusted operating frequency and speed) and the environmental parameters of corresponding environmental entities (such as temperature and humidity) are continuously collected at the sampling interval. The collected data is transmitted to the data processing center in real time, maintaining data timestamp synchronization during transmission to facilitate subsequent analysis of the control effect.

[0155] Step S163: Compare the real-time operating data after adjustment with the operating data before adjustment, and calculate the changes in each operating state parameter.

[0156] The data processing center extracts the operating status parameters of each target control device before the start of the control operation (i.e., pre-control data) and during and after the control process (i.e., post-control data). For each operating status parameter, the post-control data is subtracted from the pre-control data to obtain the change in that parameter. A positive change indicates an increase in the parameter; a negative change indicates a decrease; and zero indicates no change. All changes in operating status parameters are categorized and stored according to device identifiers, forming an operating parameter change dataset.

[0157] Step S164: Compare the real-time environmental data after regulation with the environmental data before regulation, calculate the changes in each environmental parameter, and extract the correlation relationships and correlation strength values ​​related to the target regulation equipment and corresponding environmental entity nodes from the target correlation cognitive network.

[0158] Using a method similar to step S163, the environmental parameters before and after regulation for each environmental entity node are extracted. The change in each environmental parameter is calculated, i.e., the environmental parameter after regulation minus the environmental parameter before regulation, forming an environmental parameter change dataset. Simultaneously, in the target association cognitive network, based on the target regulation device identifier and the corresponding environmental entity node identifier, the association relationship type (such as spatial association relationship) and the current association strength value between the two are queried, and the association relationship information and association strength value are stored accordingly.

[0159] Step S165: Calculate the normalized actual impact effect index based on the changes in operating status parameters and environmental parameters; determine the expected impact effect index based on the correlation strength value; when the actual impact effect index is consistent with the expected impact effect index, maintain the original correlation strength value unchanged; when the actual impact effect index is greater than the expected impact effect index, increase the correlation strength value of the corresponding correlation relationship according to a preset ratio; when the actual impact effect index is less than the expected impact effect index, decrease the correlation strength value of the corresponding correlation relationship according to a preset ratio.

[0160] The changes in operational status parameters and environmental parameters are normalized to a range of 0 to 1. A weighted average is then calculated based on their respective weights (set according to parameter importance) to serve as the actual impact indicator. The expected impact indicator is determined through a preset mapping relationship based on the correlation strength value; a higher correlation strength value corresponds to a higher expected impact indicator. The actual impact indicator is compared with the expected impact indicator: if the difference is within a preset error range, they are considered consistent, and the original correlation strength value remains unchanged; if the actual impact indicator is greater than the expected impact indicator and exceeds the error range, the original correlation strength value is multiplied by a preset ratio greater than 1 to increase the correlation strength value; if the actual impact indicator is less than the expected impact indicator and exceeds the error range, the original correlation strength value is multiplied by a preset ratio less than 1 to decrease the correlation strength value.

[0161] Step S166: Update the association strength value of the corresponding association in the target association cognitive network.

[0162] Based on the calculation results of step S165, the corresponding association is located in the target association cognitive network, and the original association strength value is replaced with the adjusted association strength value. During the update process, the adjustment time, reason for adjustment (such as the actual impact being higher than expected), and values ​​before and after adjustment are recorded simultaneously to form an association strength update log for easy subsequent traceability and analysis. At the same time, it is ensured that the updated association strength value remains within the range of 0 to 1. If the adjusted value exceeds this range, 0 or 1 is used as the final update value.

[0163] Step S167: Add the dynamic control scheme, parameter changes before and after control, and impact assessment results of this control as new sample data to the historical case library of the target association cognitive network.

[0164] The full text of the dynamic control plan, the dataset of changes in operating parameters, the dataset of changes in environmental parameters, and the comparison results of actual and expected impact indicators (i.e., impact assessment results) during this control process were collected and integrated into a complete sample data set. A timestamp, scenario type identifier, and current building operational status type identifier were added to this sample data set, and then it was stored in the historical case database of the target-related cognitive network. The historical case database uses a categorized storage method, classifying the sample data according to scenario type or status type for easy subsequent retrieval and reuse.

[0165] Step S168: Periodically perform statistical analysis on the sample data in the historical case library to identify new potential relationships. When the support and confidence of a potential relationship reach a preset threshold, add a new relationship and an initial relationship strength value to the target relationship cognitive network.

[0166] A regular analysis cycle is set, and statistical analysis is performed on the sample data in the historical case library according to this cycle. Using association rule mining algorithms, unrecorded potential associations between different device entity nodes and between device entity nodes and environmental entity nodes are analyzed. The support (i.e., the proportion of sample data containing this association to the total sample data) and confidence (i.e., the probability that the parameter of one entity node changes when the parameter of another entity node changes) of the potential association are calculated. The support and confidence are compared with preset thresholds. If both reach or exceed the preset thresholds, the potential association is determined to be a valid association, and a corresponding relationship type label (such as spatial association label, collaborative operation label, etc.) is assigned to it. The initial association strength value is calculated according to the method in step S127. The new association and the initial association strength value are added to the target association cognitive network, realizing the dynamic expansion and optimization of the target association cognitive network.

[0167] Figure 2 The illustration shows exemplary hardware and software components of an AI-based intelligent control system 100 for building automation systems, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the AI-based intelligent control system 100 for building automation systems and to perform the functions described in this application.

[0168] The AI-based intelligent control system 100 for building automation can be a general-purpose server or a special-purpose server; both can be used to implement the AI-based intelligent control method for building automation of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0169] For example, the AI-based building automation intelligent control system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the AI-based building automation intelligent control system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The AI-based building automation intelligent control system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0170] For ease of explanation, only one processor is described in the AI-based building automation intelligent control system 100. However, it should be noted that the AI-based building automation intelligent control system 100 of this application may also include multiple processors. Therefore, the steps executed by one processor described in this application may also be executed jointly or individually by multiple processors. For example, if the processor of the AI-based building automation intelligent control system 100 executes steps A and B, it should be understood that steps A and B may also be executed jointly by two different processors or individually by one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0171] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned intelligent control method for building automation system based on artificial intelligence is realized.

[0172] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. An intelligent control method for a building automation system based on artificial intelligence, characterized in that, The method includes: Collect real-time operating data of each device in the building automation system and real-time environmental data of each functional area; A target association cognitive network is constructed based on the real-time operation data and real-time environment data. The target association cognitive network includes device entity nodes, environment entity nodes, and the relationship between entities. The target association cognitive network is used to perform context recognition processing on the real-time operation data and real-time environment data to generate a target context feature set. Based on the target context feature set and target association cognitive network, situational awareness processing is performed to identify the current building operation status type; A dynamic control scheme is generated based on the current building operation status type and the target association cognitive network. The dynamic control scheme includes equipment control priority and parameter adjustment sequence. The dynamic control scheme is converted into control commands for the building automation system and sent to the corresponding equipment to perform control operations. The control operation data and environmental data of the equipment after control are collected to update the target association cognitive network. The construction of the target association cognitive network based on the real-time operational data and real-time environmental data includes: The device identifier, operating status parameters, and operating time information are extracted from the real-time operating data. Each device identifier is treated as a device entity node, and the operating status parameters and operating time information are treated as attributes of the device entity node. Extract area identifiers, environmental parameters, and collection time information from the real-time environmental data. Treat each area identifier as an environmental entity node and the environmental parameters and collection time information as attributes of the environmental entity node. Analyze the spatial relationships between device entity nodes and environmental entity nodes, determine the environmental entity node corresponding to each device entity node, and the device entity node associated with each environmental entity node; Analyze the operational coordination relationships between equipment entity nodes, and identify pairs of equipment entity nodes that operate within the same functional area or have interlocking control relationships; Analyze the environmental impact relationships between environmental entity nodes and identify the mutual influence of environmental parameters between environmental entity nodes in adjacent areas; Set relationship type labels for spatial association, operational collaboration, and environmental impact, respectively. The relationship type labels include spatial association labels, collaborative operation labels, and environmental impact labels. The association strength value between different entity nodes is calculated based on historical operational data and historical environmental data. The association strength value is used to represent the tightness of the association relationship. By integrating device entity nodes, environment entity nodes, relationship type labels, and association strength values, a preliminary association graph structure is constructed. A dynamic adjustment mechanism is set for the preliminary correlation map structure to update the correlation strength value according to changes in real-time operating data and real-time environmental data; The integrity of the preliminary association graph structure is verified by the graph verification rules, and the preliminary association graph structure that passes the verification is determined as the initial version of the target association cognitive network.

2. The intelligent control method for building automation systems based on artificial intelligence according to claim 1, characterized in that, The calculation of the association strength value between different entity nodes based on historical operational data and historical environmental data includes: Collect historical operating data and historical environmental data within a preset time period. The historical operating data includes a sequence of historical operating status parameters of the device entity nodes, and the historical environmental data includes a sequence of historical environmental parameters of the environmental entity nodes. Time alignment processing is performed on the historical operating status parameter sequence and the historical environment parameter sequence to ensure that parameter data at the same point in time correspond and are associated. For spatial correlation, device entity nodes and environmental entity nodes within the same spatial region are selected, the correlation coefficient between the device operating status parameter sequence and the environmental parameter sequence is calculated, and the absolute value of the correlation coefficient is used as the initial correlation strength value of the spatial correlation. For operational coordination relationships, a pair of device entity nodes with coordinated operation characteristics is selected, the synchronization rate of change of the sequence of operating state parameters of the two device entity nodes is calculated, and the synchronization rate of change is used as the initial association strength value of the operational coordination relationship. For environmental impact relationships, environmental entity nodes in adjacent spatial regions are selected, and the lag correlation coefficient between environmental parameter sequences is calculated. The absolute value of the lag correlation coefficient is used as the initial correlation strength value of the environmental impact relationship. The lag correlation coefficient takes into account the time delay effect of environmental parameter changes. Based on a preset correlation strength attenuation factor, the initial correlation strength value of the correlation relationship that has not undergone state change within a preset attenuation monitoring time window is attenuated. Based on a preset correlation strength enhancement factor, the initial correlation strength value of the correlation relationship that has undergone synergistic changes more than a preset frequency threshold within a preset enhanced monitoring time window is enhanced. The processed association strength values ​​are normalized, and the normalized association strength values ​​are assigned to the corresponding association relationships between entity nodes as attribute values ​​of the association relationships.

3. The intelligent control method for building automation systems based on artificial intelligence according to claim 1, characterized in that, The step of using the target association cognitive network to perform context recognition processing on the real-time operating data and real-time environmental data to generate a target context feature set includes: The historical variation patterns of environmental parameters of environmental entity nodes are extracted from the target-related cognitive network to determine the normal fluctuation range and focused change patterns of each environmental parameter. Extract the historical change patterns of the operating status parameters of the device entity nodes from the target association cognitive network, and determine the normal operating range and focused operating mode of each operating status parameter; The environmental parameters in the real-time environmental data are compared with the normal fluctuation range of the corresponding environmental entity nodes to identify abnormal environmental parameters and the degree of abnormality. The operating status parameters in the real-time operating data are compared with the normal operating range of the corresponding device entity nodes to identify abnormal operating parameters and the degree of abnormality. Based on the spatial association relationship in the target association cognitive network, the environmental entity nodes corresponding to abnormal environmental parameters are associated with the associated equipment entity nodes to determine the set of equipment entity nodes that affect the environmental entity node. Based on the operational collaboration relationship in the target association cognitive network, the device entity node corresponding to the abnormal operation parameter is associated with other associated device entity nodes to determine the set of collaborative device entity nodes affected by the device entity node. Based on abnormal environmental parameters, abnormal operating parameters and their associated set of device entity nodes and set of cooperating device entity nodes, a preliminary scenario description vector is constructed. The preliminary scenario description vector is matched with a preset scenario template library, which contains feature vectors and scenario identifiers for various focused scenarios. Calculate the similarity value between the initial context description vector and the feature vectors of each context template, and select the context identifier corresponding to the context template with the highest similarity value as the current context type; Based on the current scenario type and the associated entity node attributes, generate a target scenario feature set that includes scenario identifier, a list of related entity nodes, a list of abnormal parameters, and the degree of abnormality.

4. The intelligent control method for building automation systems based on artificial intelligence according to claim 3, characterized in that, The step of matching the preliminary scenario description vector with a preset scenario template library includes: Construct a pre-defined scenario template library, which contains a variety of focused scenarios. Each focused scenario corresponds to a scenario template, and each scenario template contains a scenario identifier, a feature vector dimension description, and a standard feature vector. Extract feature values ​​of each dimension from the initial context description vector to determine the number of dimensions and the meaning of each dimension of the initial context description vector; Select a subset of situation templates from the situation template library that are consistent with the number of dimensions and the meaning of the dimensions of the initial situation description vector. For each situation template in the situation template subset, obtain its standard feature vector and calculate the cosine similarity value between the initial situation description vector and the standard feature vector. The cosine similarity value is used to represent the degree of closeness between the directions of the two vectors. Record the cosine similarity value of each scenario template and sort them in descending order of similarity value. Select the scenario template with the highest similarity value as the best matching template. When the cosine similarity value of the best matching template is greater than the preset similarity threshold, the context identifier of the best matching template is taken as the current context type; when the cosine similarity value of the best matching template is less than or equal to the preset similarity threshold, the current context type is marked as an unknown context, and the preliminary context description vector is added to the pending update list of the context template library.

5. The intelligent control method for building automation systems based on artificial intelligence according to claim 1, characterized in that, The situational awareness processing based on the target context feature set and target association cognitive network to identify the current building operation situation type includes: Extract each scenario identifier and its corresponding list of related entity nodes from the target scenario feature set to determine the distribution location and influence range of each scenario in the building space; Based on the environmental influence relationships in the target-related cognitive network, the mutual influence paths and influence intensity between different scenarios are analyzed. Based on the scope of influence of each scenario and the path of mutual influence, a scenario association network is constructed. The nodes of the scenario association network are scenario identifiers, and the edges are the influence relationships and influence strengths between scenarios. Temporal analysis is performed on the aforementioned context association network to extract context sequence change features within consecutive time periods. These sequence change features include the order of context occurrence, duration, and transition frequency. Based on the attributes of device entity nodes in the target association cognitive network, analyze the device operating load status corresponding to each scenario, and calculate the overall device load rate and load distribution balance. By combining context association network, sequence variation features, overall device load rate and load distribution balance, a situation description matrix is ​​constructed. The rows of the situation description matrix represent the situation type, the columns represent the situation assessment dimensions, and the element values ​​represent the assessment values ​​of the situation type in the corresponding dimension. The situation description matrix is ​​input into a preset situation recognition model. The situation recognition model learns from historical situation samples to output membership values ​​of multiple preset situation types and selects the preset situation type with the highest membership value as the current building operation situation type. Generate situation awareness results that include the current building operation status type, situation description matrix, and membership values.

6. The intelligent control method for building automation systems based on artificial intelligence according to claim 5, characterized in that, The situation description matrix is ​​input into a preset situation recognition model. This model, through learning from historical situation samples, outputs membership values ​​for various preset situation types, including: A pre-defined situation recognition model is constructed, comprising an input layer, a feature mapping layer, a correlation analysis layer, and an output layer. The input layer receives the situation description matrix and converts it into a matrix-vector form that the model can process. The feature mapping layer contains multiple parallel feature extractors, each extracting key features from different dimensions of the situation description matrix to generate a high-dimensional feature vector. The correlation analysis layer includes a graph attention-based correlation processor for modeling inter-entity relationships within the high-dimensional feature vector, strengthening the weights of important features. The output layer contains multiple neurons, each corresponding to a pre-defined situation type, and outputs the membership value of that situation type. Collect historical situation sample data, which includes a historical situation description matrix and corresponding actual situation type labels; The historical situational sample data is divided into a training sample set and a validation sample set; The situation recognition model is trained using a training sample set. The loss value between the membership value output by the model and the actual situation type label is calculated through forward propagation. Based on the loss value, the weight parameters of each layer of the model are adjusted through the backpropagation algorithm until the recognition accuracy of the situation recognition model on the validation sample set reaches a preset threshold. The parameters of the trained situation recognition model are saved to form a preset situation recognition model that can be used for inference. The current situation description matrix is ​​input into the trained situation recognition model. Through the processing of the input layer, feature mapping layer, and association analysis layer, the output layer outputs membership values ​​of various preset situation types. The membership values ​​represent the probability that the current situation belongs to each preset situation type.

7. The intelligent control method for building automation systems based on artificial intelligence according to claim 1, characterized in that, The dynamic control scheme generated based on the current building operation status type and the target association cognitive network includes: Extract the set of equipment entity nodes and the set of environmental entity nodes that are related to the current building operation status type from the target association cognitive network; Based on the current building operation status type, query the preset status-control strategy library to obtain a preliminary control strategy framework. The preliminary control strategy framework includes the suggested control equipment type and the direction of environmental parameter adjustment. Match the relevant set of equipment entity nodes with the equipment types in the preliminary control strategy framework to determine the specific target control equipment list; Analyze the operational coordination relationships of each equipment entity node in the target control equipment list, determine the equipment control sequence constraints, and generate a preliminary control sequence based on the equipment control sequence constraints and the operational status parameters of the equipment entity nodes; Based on the association strength value in the target association cognitive network, the influence weight of each target control device on the environmental entity node is calculated. The influence weight is used as an important reference factor for control priority. Combined with the preliminary control sequence, the target control devices are prioritized to generate a device control priority sequence. For each target control device, based on the environmental parameter adjustment direction in the preliminary control strategy framework and the current environmental data, the parameter adjustment range is determined. By referring to the parameter adjustment effect in the historical operating data of the target control device, the specific parameter adjustment step size and number of adjustments are determined within the parameter adjustment range to form a parameter adjustment sequence. The equipment control priority sequence is integrated with the corresponding parameter adjustment sequence to generate a dynamic control scheme that includes equipment identification, control priority, parameter adjustment sequence and adjustment order constraints.

8. The intelligent control method for building automation systems based on artificial intelligence according to claim 7, characterized in that, The parameter adjustment effect in the historical operating data of the reference target control device is used to determine the specific parameter adjustment step size and number of adjustments within the parameter adjustment range, forming a parameter adjustment sequence, including: Extract historical adjustment records that are the same as the current situation type from the historical operating data of the target control equipment. The historical adjustment records include parameter values ​​before adjustment, adjustment step size, number of adjustments, parameter values ​​after adjustment, and corresponding environmental parameter changes. The historical adjustment records are filtered to retain valid adjustment records in which the changes in environmental parameters after adjustment meet the expected control targets. Extract the relationship between the adjustment step size and the rate of change of environmental parameters from the effective adjustment records, wherein the rate of change of environmental parameters is the ratio of the change value of environmental parameters to the adjustment duration; Based on the upper and lower limits of the parameter adjustment range and the current parameter value, the total adjustment amount is determined, which is the difference between the current parameter value and the target parameter value. Based on the relationship between the adjustment step size and the rate of change of environmental parameters in historical effective adjustment records, the initial adjustment step size is preset. Based on the initial adjustment step size and the total adjustment amount, the required number of adjustments is initially calculated, where the number of adjustments is the ratio of the total adjustment amount to the initial adjustment step size. The initial number of adjustments was revised based on the number of adjustments and the stability of the adjustment effect of the same equipment under the same total adjustment amount in the historical adjustment records. Based on the corrected number of adjustments and the total adjustment amount, the final adjustment step size is re-determined so that the total adjustment amount is equal to the product of the final adjustment step size and the number of adjustments. Arrange the parameter value changes and corresponding adjustment time intervals for each adjustment in the adjustment order to form a parameter adjustment sequence that includes adjustment step size, number of adjustments, and adjustment time interval.

9. An intelligent control system for building automation based on artificial intelligence, characterized in that, The intelligent control system of the building automation system based on artificial intelligence includes a processor and a memory. The memory and the processor are connected. The memory is used to store programs, instructions or code. The processor is used to execute the programs, instructions or code in the memory to realize the intelligent control method of the building automation system based on artificial intelligence as described in any one of claims 1-8.

Images