Building heating diagnosis method and system based on digital twinning and knowledge enhancement in cold region

Abstract

The application provides a cold region building heating diagnosis method and system based on digital twinning and knowledge enhancement. The method obtains the supply and return water temperature, circulating flow, pressure difference, heat pump power, defrosting state, indoor and outdoor environment, alarm log and equipment account, constructs the equipment corresponding relationship and equipment connection relationship, identifies the current working condition mark based on the fault contrast digital twinning model and reconstructs the normal operation track, compares with the operation data to form an abnormal performance set containing the deviation position, deviation direction and linkage change, merges the easily confused fault groups based on the fault distinguishing knowledge graph and generates a fault distinguishing condition table, generates a candidate fault contrast track under the working condition constraint and matches, excludes the judgment combined with the condition table, outputs the target fault, associated equipment and positioning basis or manual checking items, and can improve the diagnosis consistency and positioning efficiency.

Description

Technical Field

[0001] This invention relates to the field of operational diagnosis of heating systems in cold-region buildings, and particularly to a diagnostic method for heating systems in cold-region buildings based on digital twins and knowledge enhancement. Background Technology

[0002] Heating systems for buildings in cold regions typically consist of a heat pump, its defrosting control system, circulating pump, heat exchanger, valves, piping, and terminal branches. During operation, they are significantly affected by changes in outdoor temperature and humidity, as well as fluctuations in building load. The supply and return water temperature difference, circulating flow rate, pressure difference, and heat pump power vary frequently depending on operating conditions. To ensure indoor temperatures meet standards and maintain stable system operation, continuous monitoring of operational data and alarm information is necessary to promptly identify anomalies and pinpoint specific equipment or link segments, providing a basis for maintenance and troubleshooting.

[0003] Existing technical solutions mostly employ monitoring platforms to collect data on supply and return water temperature, circulation flow rate, pressure difference, heat pump power, and alarm logs. These data are then combined with fixed thresholds or empirical rules to issue over-limit alarms and fault indications. Some solutions introduce simulation models or energy consumption models based on physical mechanisms to estimate and compare normal operating ranges under typical conditions. Other solutions utilize data-driven methods to detect and classify anomalies in multi-measuring-point time series data, or build a knowledge base based on equipment ledgers and topological relationships, using rule-based retrieval to provide candidate faults and handling suggestions to support the operational diagnosis of heating systems.

[0004] The above-mentioned solutions often suffer from scattered diagnostic criteria under complex cold-climate conditions. Fixed thresholds and static rules are difficult to adapt to changes in operating conditions, making it difficult to uniformly explain the performance differences of the same equipment under different outdoor environments and defrosting stages. When relying solely on single-point or local indicators for judgment, it is difficult to form consistent evidence regarding the linkage sequence between supply and return water temperature difference, flow rate, pressure difference, power, and defrosting, making it difficult to distinguish easily confused faults. The lack of a mechanism to uniformly compare abnormal performance with candidate fault comparison trajectories under the same operating condition baseline leads to unstable root cause localization and localization data output.

[0005] Therefore, a diagnostic method for heating systems in cold-region buildings that can overcome the shortcomings of the existing technology is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a diagnostic method for heating systems in cold-region buildings based on digital twins and knowledge enhancement. The core technical problem to be solved by this application is: in the scenario where the operating conditions of heating systems in cold-region buildings are constantly changing and the equipment topology is coupled, how to combine the equipment connection relationship and fault differentiation knowledge on the basis of the same current operating condition benchmark and the same normal operating trajectory to eliminate easily confused faults and provide traceable root cause location results.

[0007] The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement according to embodiments of the present invention includes:

[0008] S1. Obtain the supply and return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, indoor and outdoor environment, alarm logs and equipment ledger of the heating system for cold-region buildings, and form heating operation data and equipment connection relationships according to the equipment correspondence.

[0009] S2. Based on the heating operation data, call the fault comparison digital twin model to determine the current operating condition identifier and normal operating trajectory, and compare the heating operation data with the normal operating trajectory to obtain a set of abnormal manifestations including deviation position, deviation direction and linkage changes;

[0010] S3. Based on the abnormal behavior set and equipment connection relationship, call the fault differentiation knowledge graph. First, group multiple candidate faults into easily confused fault groups according to the common abnormal behaviors. Then, along the equipment connection relationship, heat and water flow transfer relationship and fault causal relationship, determine the corresponding temperature difference change location, flow rate and pressure difference correspondence, power change and defrost status correspondence for each candidate fault in the easily confused fault group. Write each correspondence into the fault differentiation condition table and output the easily confused fault group and the fault differentiation condition table.

[0011] S4. Based on the easily confused fault groups, fault differentiation condition table, current operating condition identifier, and normal operating trajectory, call the fault comparison digital twin model. Under the constraint of the current operating condition identifier, generate the fault comparison trajectory corresponding to each candidate fault. Then, according to the fault differentiation condition table, extract the order of supply and return water temperature difference changes, the direction of circulation flow and pressure difference linkage, and the correspondence between heat pump power changes and defrosting status from the fault comparison trajectory. Compare the extraction results with the abnormal performance set item by item to obtain the fault matching results of each candidate fault.

[0012] S5. Input the fault matching results into the fault differentiation knowledge graph, and make exclusion judgments according to the fault differentiation condition table. When one candidate fault is retained, a root cause judgment result is formed. When multiple candidate faults are retained or no candidate faults are retained, a fault group to be checked and missing judgment basis are formed.

[0013] S6. Based on the root cause determination results, output the target fault, associated equipment and location basis to form the fault location results. Based on the fault group to be checked and the missing judgment basis, output the manual verification results.

[0014] Optionally, S1 is as follows:

[0015] Acquire supply water temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature and indoor set temperature, and generate operational measurement point data according to the measurement point markings and acquisition equipment;

[0016] Extract alarm category, alarm duration and corresponding device from alarm logs, and extract device identifier, device location, pipeline node and device type from device ledger to form device attribute data;

[0017] Based on the operational measurement point data and equipment attribute data, determine the corresponding relationship of the equipment, and write the supply water temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, indoor and outdoor environment, alarm type, alarm duration and equipment location into the corresponding equipment according to the equipment correspondence. Then, calculate the supply and return water temperature difference based on the supply water temperature and return water temperature to form heating operation data.

[0018] Based on the equipment correspondence and the pipeline nodes, medium flow direction and connection order in the equipment attribute data, the equipment association between heat pumps, circulating pumps, heat exchangers, valves and terminal branches is determined, and the equipment connection relationship is formed.

[0019] Optionally, S2 is as follows:

[0020] The heating operation data is arranged into heat source side input segment, circulation side input segment and terminal side input segment according to the medium flow direction in the equipment connection relationship. Then, the heat source side input segment, circulation side input segment and terminal side input segment are written into the working condition coding layer of the fault comparison digital twin model. The supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, indoor set temperature, alarm category, alarm duration and equipment location are associated and mapped to form a segmented input sequence.

[0021] Using the segmented input sequence as input, the heat source operating status, hydraulic transmission status, terminal load status and alarm triggering status are encoded segment by segment, and the segmented encoding results are fused according to the heat transfer path and water flow transfer path in the equipment connection relationship to form the current operating condition vector;

[0022] Based on the current operating condition vector, the outdoor environment range, defrosting operation stage, heat pump output level, circulation flow status and terminal temperature difference demand are combined and judged to form the current operating condition identifier. Based on the current operating condition identifier, the current operating condition judgment thresholds corresponding to the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow, pressure difference, heat pump power and defrosting status are determined respectively.

[0023] Input the current operating condition vector into the normal operating trajectory layer. Under the constraints of the current operating condition identifier, reconstruct the corresponding normal operating trajectory according to the equipment location for the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power and defrosting status, forming a normal operating trajectory covering the heat source side, circulation side and terminal side.

[0024] The heating operation data is compared with the normal operation trajectory one by one according to the equipment location. For the difference that meets the current operating condition judgment threshold, the deviation variable, deviation amount, difference direction and the equipment that occurred are extracted to form the initial deviation result.

[0025] Based on the initial deviation results and equipment connection relationships, the deviation variables are mapped to specific equipment segments in the heating chain along the equipment connection relationships, and sorted according to the order of appearance of the deviation variables on the heat source side, circulation side and terminal side to form the deviation position and deviation direction;

[0026] Based on the correspondence between the deviation location, deviation direction, and supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrost status, the order of supply and return water temperature difference changes, the correspondence between circulation flow rate and pressure difference, and the correspondence between heat pump power changes and defrost status are determined to form a linkage change. The deviation location, deviation direction, and linkage change are combined into an abnormal behavior set. The current operating condition identifier, normal operating trajectory, and abnormal behavior set are provided for use in the fault differentiation knowledge graph and fault comparison digital twin model.

[0027] Optionally, S3 specifically refers to:

[0028] The deviation location, deviation direction, and linkage changes in the abnormal behavior set are written into the abnormal behavior node of the fault differentiation knowledge graph. The connection order of heat pump, circulating pump, heat exchanger, valve and terminal branch in the equipment connection relationship is written into the equipment node. The fault node and condition node corresponding to the heating operation data are retrieved to form the graph input sequence.

[0029] The graph input sequence is mapped to nodes, and abnormal behavior nodes, device nodes, fault nodes and condition nodes are converted into unified node representations. The nodes are then aligned according to the connection direction between abnormal behavior nodes and device nodes, device nodes and fault nodes, and fault nodes and condition nodes to form an initial relationship representation.

[0030] The initial relation representation is input into the first graph relation layer corresponding to the device connection relationship. The device segment propagation is performed on the deviation position and deviation direction to limit the location of the abnormality to the heat source side, circulation side or end side, thus forming the device positioning result.

[0031] The equipment positioning results are input into the second graph relationship layer corresponding to the heat and water flow transfer relationship. The changes in supply and return water temperature difference, circulation flow rate and pressure difference are propagated through the link, and the changes are sorted according to the direction of medium flow to form the link propagation results.

[0032] The link propagation results are input into the third graph relationship layer corresponding to the fault causal relationship. Causal correlation is performed on the heat pump power change and defrost status change. In the fault merging layer, fault nodes are screened according to the threshold of the number of common abnormal manifestations. Fault nodes that meet the combination of the same equipment connection relationship, heat and water flow transfer relationship and fault causal relationship are grouped into easily confused fault groups.

[0033] Based on the easily confused fault group, equipment location results, and link propagation results, the location of temperature difference change is determined for each candidate fault within the easily confused fault group. The changes in circulation flow rate and pressure difference are correlated to form a correspondence between flow rate and pressure difference. Then, the changes in heat pump power and defrost status are correlated to form a correspondence between power change and defrost status, thus obtaining the differentiation condition results.

[0034] The results of the differentiation conditions are written into the condition node, and the relationship between temperature difference change location, flow rate and pressure difference, and power change and defrosting status are summarized through the differentiation condition output layer to generate a fault differentiation condition table. The easily confused fault groups and the fault differentiation condition table are then provided for use by the fault comparison digital twin model.

[0035] Optionally, S4 specifically refers to:

[0036] Input the easily confused fault group and the fault differentiation condition table into the fault comparison layer of the fault comparison digital twin model, and use the current working condition identifier and normal operating trajectory as the constraint input of the fault comparison layer. Determine the action link of each candidate fault according to the corresponding equipment in the connection relationship between candidate faults and equipment within the easily confused fault group to form the action sequence of candidate faults.

[0037] Based on the current operating condition identifier, the current operating condition boundary is determined by the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power and defrosting status in the normal operation trajectory. Then, the current operating condition boundary is written into the candidate fault action sequence to form the operating condition constraint sequence.

[0038] Using the operating condition constraint sequence and normal operating trajectory as input, the fault impact calculation is performed on each candidate fault in the candidate fault action sequence. The impact of each candidate fault on the heat source side, circulation side and terminal side is expanded along the equipment connection relationship to supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power and defrosting status, forming the fault comparison trajectory corresponding to each candidate fault.

[0039] Based on the location of temperature difference changes in the fault differentiation criteria table, the order of supply and return water temperature difference changes on the heat source side, circulation side, and terminal side is extracted from the fault comparison trajectory of each candidate fault to form the order of supply and return water temperature difference changes.

[0040] Based on the correspondence between flow rate and differential pressure in the fault differentiation condition table, the direction of change of circulating flow rate and the direction of change of differential pressure are extracted from the fault comparison trajectory of each candidate fault, and the direction of change of circulating flow rate and the direction of change of differential pressure are matched to form the linkage direction of circulating flow rate and differential pressure.

[0041] Based on the correspondence between power change and defrost status in the fault differentiation condition table, the heat pump power change and defrost status switching are extracted from the fault comparison trajectory of each candidate fault, and the heat pump power change and defrost status switching are sequentially matched to form the correspondence between heat pump power change and defrost status.

[0042] The sequence of supply and return water temperature difference changes, the direction of linkage between circulation flow and pressure difference, and the correspondence between heat pump power changes and defrosting status are input into the matching layer. The deviation position, deviation direction, and linkage changes in the abnormal performance set are compared item by item to form the corresponding fault matching result for candidate faults that meet the fault differentiation condition table.

[0043] Optional, S5 specifically includes:

[0044] The fault matching results are input into the fault nodes of the fault differentiation knowledge graph, and the fault differentiation condition table is input into the condition nodes. According to the location of temperature difference change, the correspondence between flow rate and pressure difference, and the correspondence between power change and defrosting status, the corresponding faults are searched to form the candidate fault retention results.

[0045] The candidate fault retention results are excluded. When the fault matching result meets the temperature difference change location, flow rate and pressure difference correspondence and power change and defrost status correspondence in the fault differentiation condition table, the corresponding candidate fault is retained. When the fault matching result lacks the temperature difference change location, flow rate and pressure difference correspondence or power change and defrost status correspondence, the corresponding candidate fault is screened out to form the remaining candidate fault set.

[0046] The remaining candidate fault set is counted. When the remaining candidate fault set contains a candidate fault, the retained candidate fault is associated with the deviation position and equipment connection relationship in the abnormal behavior set to form the root cause determination result.

[0047] When the remaining candidate fault set contains multiple candidate faults or does not contain any candidate faults, the unmet fault differentiation condition entries are extracted to form the basis for missing fault judgment, and the remaining candidate fault set is identified as the fault group to be checked.

[0048] Optional, S6 specifically includes:

[0049] Align the root cause determination results with the device connection relationship and the deviation position in the abnormal behavior set to determine the target fault and associated device corresponding to the retained candidate fault;

[0050] Based on the target fault, abnormal manifestation set, fault differentiation condition table and fault matching results, the deviation location, deviation direction, supply and return water temperature difference change sequence, circulation flow and pressure difference linkage direction, heat pump power change and defrost status correspondence relationship are summarized to form the positioning basis;

[0051] The target fault, associated equipment, and location data are associated with the current operating condition indicator and normal operating trajectory to form the fault location result;

[0052] Match the fault groups to be checked with the missing judgment criteria, and determine the missing temperature difference change location, flow rate and pressure difference correspondence, or power change and defrosting status correspondence for each candidate fault in the fault group to be checked, so as to form the manual check results.

[0053] Optionally, the current operating condition identifier is used to constrain the corresponding reconstruction of the normal operating trajectory. Under the constraint of the current operating condition identifier, the heating operation data is compared with the normal operating trajectory to form an abnormal performance set. Under the constraint of the current operating condition identifier, the normal operating trajectory is used as the reference basis, and the fault comparison trajectory corresponding to each candidate fault is generated according to the easily confused fault group and the fault differentiation condition table, so that the abnormal performance set and the fault comparison trajectory are formed on the same current operating condition identifier and the same normal operating trajectory.

[0054] Optionally, the easily confused fault group is used to determine the candidate fault range when the fault comparison digital twin model generates the fault comparison trajectory. The fault comparison digital twin model generates the fault comparison trajectory corresponding to each candidate fault according to the corresponding equipment in the equipment connection relationship of each candidate fault in the easily confused fault group. Based on the fault differentiation condition table, the order of supply and return water temperature difference change, the direction of circulation flow and pressure difference linkage, and the correspondence between heat pump power change and defrost status are extracted from the fault comparison trajectory corresponding to each candidate fault. The results are then compared with the abnormal performance set item by item to form the fault matching result of each candidate fault.

[0055] Optionally, a cold-region building heating diagnostic system based on digital twins and knowledge enhancement includes:

[0056] The data acquisition and construction module is used to acquire water supply temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, indoor and outdoor environment, alarm logs and equipment ledgers, and form heating operation data and equipment connection relationships according to equipment correspondence.

[0057] The knowledge graph reasoning module configures a fault differentiation knowledge graph, which is used to output easily confused fault groups and fault differentiation condition tables based on the abnormal behavior set and device connection relationship.

[0058] The digital twin comparison module is configured with a fault comparison digital twin model, which is used to determine the current operating condition identifier and normal operating trajectory based on heating operation data, and compare the heating operation data with the normal operating trajectory to form a set of abnormal manifestations including deviation position, deviation direction and linkage changes; the digital twin comparison module is also used to generate the fault comparison trajectory corresponding to each candidate fault and form the fault matching result after receiving the easily confused fault group and the fault differentiation condition table.

[0059] The elimination judgment and location verification module is used to perform elimination judgment based on the fault matching results and the fault differentiation condition table. When one candidate fault is retained, the root cause judgment result is generated and the fault location result is output. When multiple candidate faults are retained or no candidate faults are retained, the fault group to be verified, the missing judgment basis, and the manual verification result are output.

[0060] The beneficial effects of this invention are:

[0061] (1) This invention proposes an improved method for diagnosing heating in cold-region buildings. Under the constraints of equipment connection relationships, the method segments and fuses the supply and return water temperature, circulation flow rate, pressure difference, heat pump power, defrosting status, indoor and outdoor environment, and alarm information to form a current operating condition vector. This vector is then combined to determine the current operating condition identifier. Under the constraints of this identifier, the normal operating trajectory covering the heat source side, circulation side, and end side is reconstructed. When comparing the operating data with the normal operating trajectory one by one according to equipment location, a judgment threshold that changes with the operating condition is introduced. Deviation variables are mapped to link segments, further extracting the deviation position, deviation direction, and the order of supply and return water temperature difference changes, the linkage direction of flow rate and pressure difference, and the correspondence between power changes and defrosting status. This establishes the abnormal performance set on a unified operating condition benchmark and a unified normal trajectory benchmark, avoiding the influence of operating condition drift caused by relying solely on fixed thresholds or local rules, and improving the comparability and consistency of abnormal evidence under different outdoor environments and defrosting stages.

[0062] (2) This invention proposes a novel knowledge-enhanced differentiation mechanism. It simultaneously writes the set of abnormal manifestations and the equipment connection relationships into a fault differentiation knowledge graph, propagating and locating faults along the equipment connection relationships and the heat and water flow transfer relationships. Furthermore, it associates power changes with defrosting status with fault causal relationships. This first groups candidate faults with many common abnormal manifestations into easily confused fault groups, and then forms a fault differentiation condition table for each candidate fault within the group, consisting of the correspondence between temperature difference changes, flow rate and pressure difference, and the correspondence between power changes and defrosting status. This condition table also serves as the basis for extracting relational quantities from the digital twin fault comparison trajectory and as the basis for knowledge graph exclusion judgment. This transforms the differentiation of candidate faults from single-point features to a constraint test of the sequence and linkage relationships of links, unlike schemes that only provide a candidate list or only perform classification and scoring. When a unique judgment cannot be made, it outputs the missing judgment criteria to point to the type of evidence that needs supplementary verification.

[0063] (3) This invention proposes a holistic heating diagnosis method that combines digital twins and knowledge graphs. Under the current operating condition constraints, the normal operating trajectory is used as the reference basis. For each candidate fault in the easily confused fault group, a corresponding fault reference trajectory is generated. The relationship quantity is extracted from the reference trajectory according to the fault differentiation condition table and compared item by item with the abnormal performance set to form a fault matching result. Then, the knowledge graph performs the elimination judgment according to the same condition table, and outputs the target fault-related equipment and the location basis, or outputs the fault group to be checked and the manual verification result. This closed-loop structure enables a consistent data interface and judgment caliber between the candidate fault range determination, fault reference trajectory generation, relationship quantity extraction and matching, and elimination judgment. The location basis can be mapped to the specific deviation position, deviation direction and linkage source, thereby providing traceable diagnostic output for the root cause location of cold region heating systems in the scenario of topological coupling and frequent operating condition changes. Attached Figure Description

[0064] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0065] Figure 1 This is a flowchart of a cold-region building heating diagnosis method based on digital twins and knowledge enhancement proposed in this invention. Detailed Implementation

[0066] In Example 1, reference Figure 1 A diagnostic method for heating systems in cold-region buildings based on digital twins and knowledge enhancement includes:

[0067] S1. Obtain the supply and return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, indoor and outdoor environment, alarm logs and equipment ledger of the heating system for cold-region buildings, and form heating operation data and equipment connection relationships according to the equipment correspondence.

[0068] S2. Based on the heating operation data, call the fault comparison digital twin model to determine the current operating condition identifier and normal operating trajectory, and compare the heating operation data with the normal operating trajectory to obtain a set of abnormal manifestations including deviation position, deviation direction and linkage changes;

[0069] S3. Based on the abnormal behavior set and equipment connection relationship, call the fault differentiation knowledge graph. First, group multiple candidate faults into easily confused fault groups according to the common abnormal behaviors. Then, along the equipment connection relationship, heat and water flow transfer relationship and fault causal relationship, determine the corresponding temperature difference change location, flow rate and pressure difference correspondence, power change and defrost status correspondence for each candidate fault in the easily confused fault group. Write each correspondence into the fault differentiation condition table and output the easily confused fault group and the fault differentiation condition table.

[0070] S4. Based on the easily confused fault groups, fault differentiation condition table, current operating condition identifier, and normal operating trajectory, call the fault comparison digital twin model. Under the constraint of the current operating condition identifier, generate the fault comparison trajectory corresponding to each candidate fault. Then, according to the fault differentiation condition table, extract the order of supply and return water temperature difference changes, the direction of circulation flow and pressure difference linkage, and the correspondence between heat pump power changes and defrosting status from the fault comparison trajectory. Compare the extraction results with the abnormal performance set item by item to obtain the fault matching results of each candidate fault.

[0071] S5. Input the fault matching results into the fault differentiation knowledge graph, and make exclusion judgments according to the fault differentiation condition table. When one candidate fault is retained, a root cause judgment result is formed. When multiple candidate faults are retained or no candidate faults are retained, a fault group to be checked and missing judgment basis are formed.

[0072] S6. Based on the root cause determination results, output the target fault, associated equipment and location basis to form the fault location results. Based on the fault group to be checked and the missing judgment basis, output the manual verification results.

[0073] In this embodiment, S1 specifically refers to:

[0074] The supply water temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, and indoor set temperature are respectively written as: ,in These correspond to the following parameters in sequence: supply water temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, and indoor set temperature. Representation field Sampling time, Representation field When the sampling value recorded by the acquisition device includes the frequency measurement point of the circulating pump, the frequency of the circulating pump is written as... When the data acquisition equipment records valve opening measurement points, the valve opening should be written as... According to a unified timeline Align all measurement point sequences. , Indicates a uniform sampling interval. This represents the maximum aligned window, satisfying... The most recent sampled value is written as The circulation pump frequency and valve opening are written according to the same rules as follows: and Any field in the index that represents the supply water temperature, return water temperature, circulation flow rate, differential pressure, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, and indoor set temperature is included in the index. When no sampled value satisfies the window constraint, the index Without entering the runtime feature construction, the index will be... Each measuring point at the location is recorded as follows: ,in Indicates the measurement point identification. Indicates the identification of the data acquisition device. Indicates the type of measurement point. Indicates the measured value;

[0075] Write each alarm log entry as ,in Indicates the time when the alarm occurred. Indicates the alarm category. Indicates the duration of the alarm. Indicate the device identifier corresponding to the alarm, and write each device log record as follows. ,in Indicates equipment identification. Indicates the location of the equipment. This represents the sequence of pipe nodes recorded according to the installation order. To indicate the device type, the unique values ​​of the alarm category are encoded in the order of their appearance to form an alarm category dictionary. The unique values ​​of equipment locations are encoded in the order of the equipment ledger to form an equipment location dictionary. , pipeline node sequence Based on the system medium flow direction from the supply branch to the return branch, the inlet pipe node is analyzed. and outlet pipeline nodes The connection sequence is generated according to the node arrangement order and equipment location in the equipment ledger. The device attribute record for each device is written as follows ;

[0076] According to the measurement point records and device attribute records Determine the corresponding equipment Data collection device identification Equipment identification If they match, a binding is established directly. If the data acquisition device identifier cannot be directly bound, the binding is established based on the pipeline node, installation location, and measurement point category parsed from the measurement point identifier, and the inlet pipeline node. Outlet pipeline node Equipment location and equipment type When multiple devices meet the matching criteria, the device with the exact same location is selected. If the device locations cannot be distinguished, the connection order is selected. For the smallest device, based on the device correspondence, the supply water temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, and defrost status are written to the corresponding device. The outdoor temperature, outdoor humidity, indoor average temperature, and indoor set temperature are copied and written to all devices, unifying the timeline. Falling into the alarm time and When constructing a time interval, the alarm category and alarm duration are written to the corresponding alarm device to unify the time axis timing. If the alarm time interval is not within the alarm time range, the alarm category code is written as: Write the alarm duration as The same device in the index When multiple alarm logs exist, select the alarm log with the longest duration and classify the alarms accordingly. Dictionary by Alarm Category Mapped to alarm category code , place the device By device location dictionary Mapped to device location code The temperature difference between the supply and return water is obtained by subtracting the supply water temperature from the return water temperature. When the circulating pump equipment is matched to the circulating pump frequency measurement point, the corresponding measurement point value is written as... When the valve equipment is matched to the valve opening measurement point, the corresponding measurement point value is written as... If the equipment type is inconsistent with the circulating pump or valve, or if the corresponding measuring point is not matched, and All are written as ;

[0077] Heating operation data is organized into operation feature vectors according to equipment granularity. , The 16 components correspond sequentially to the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, indoor set temperature, circulation pump frequency, valve opening, alarm category code, alarm duration, and equipment location code. Equipment connection relationships are written as follows: , Indicates the number of devices and the number of outlet pipeline nodes. With inlet pipe node Consistent and the connection order is satisfied At that time, Recorded as The remaining positions are denoted as In scenarios involving valve diversion and parallel connection of terminal branches, multiple connection points are respectively set... According to the connection order Arrange all running feature vectors to obtain , As input for heating operation data, As input for device connection relationships.

[0078] In this embodiment, S2 specifically refers to:

[0079] The fault-referenced digital twin model is denoted as... , Indicates the parameters of the working condition coding layer. Indicates the parameters of the normal operating trajectory layer. This represents the parameters of the fault control layer, and the number of devices is denoted as [missing information]. Take continuous time on a unified timeline A decision window is composed of time indices. Index at each time At this point, read the first... Operational feature vector of each device location and device connection matrix At the same time, read the first Connection order of each device location and device location code Run feature vectors Keep as dimension, Each component corresponds sequentially to the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, indoor set temperature, circulation pump frequency, valve opening degree, alarm category code, alarm duration, and equipment location code. The equipment location code set is pre-divided into a heat source side code set. Cyclic side encoding set and end-side encoding set The equipment location code belongs to Device index in connection order Ascending order composition The equipment location code belongs to Device index in connection order Ascending order composition The equipment location code belongs to Device index in connection order Ascending order composition ,according to The media flow direction is arranged in sequence. This forms a heat source-side input section, a circulation-side input section, and an end-side input section;

[0080] The operating condition coding layer receives the media flow direction after it is arranged. The working condition coding layer uses a two-layer fully connected structure with shared parameters to associate and map the location of each device. The first layer sets... One neuron, second layer setup The nth neuron outputs the nth neuron. Each device location 3D positional encoding vector The circulating pump frequency is only retained as a measured value at the location of the circulating pump equipment, and the valve opening is only retained as a measured value at the location of the valve equipment. For equipment locations without corresponding measuring points, the corresponding components will be written as... Alarm category code, alarm duration and device location code are all mapped to all device locations. As a result, supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, indoor set temperature, circulation pump frequency, valve opening, alarm category code, alarm duration and device location code are all entered into the operating condition coding layer. There are no fields that are only written but not involved in the calculation.

[0081] Will The average of all positional encoding vectors is obtained by averaging them in connection order. 3D heat source operating state vector ,Will The average of all positional encoding vectors is obtained by averaging them in connection order. Hydraulic transport state vector ,Will The average of all positional encoding vectors is obtained by averaging them in connection order. End-of-dimension load state vector The alarm category code is not... The device index is composed of connections. and to The mean of the position encoding vector is obtained Dimensional alarm trigger state vector , When empty, Written as Zero-dimensional vector, and The mean value of each element is used to form the water flow path code. The water flow path code is then combined with... The heat transfer path code is generated by averaging each element, and then the heat transfer path code is combined with... Calculate the mean of each element to form Current working condition vector Each time index within the decision window is encoded and fused independently. When the decision window moves forward one time index along a unified time axis, the model parameters are maintained. , , The same reasoning process is repeated without changing the existing logic.

[0082] Current operating condition vector Divide into five segments and connect to five decision heads respectively. To the Access to outdoor environment detection head, outdoor environment detection head settings The nth output neuron, the nth To the The defrost operation phase determination head is connected to the maintenance system; the defrost operation phase determination head is set. The nth output neuron, the nth To the Connect the heat pump output level judgment head, and set the heat pump output level judgment head. The nth output neuron, the nth To the Access loop traffic status determination header, loop traffic status determination header settings The nth output neuron, the nth To the Terminal temperature difference demand determination head, terminal temperature difference demand determination head settings Each output neuron and five decision heads select the category corresponding to the neuron with the highest output score to form the current operating condition identifier. The four categories of outdoor environment assessment heads correspond to the following in order: , , , The four categories of the defrosting operation phase correspond to stable heating, entering defrost, continuous defrosting, and defrosting recovery, respectively. The heat pump output level, circulation flow status, and terminal temperature difference demand correspond to low, medium, and high categories, respectively. After combining all operating conditions and coding them in a fixed order, the following results are obtained: The current operating condition is identified, and the threshold parameter is denoted as follows: Threshold parameter table Store a set of identifiers for each current operating condition. dimensional benchmark threshold vector and eight groups The operating condition adjustment coefficient vector has seven components arranged in the order of supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrost status. The defrost status is directly numerically encoded, and stable heating is written as... Enter defrost as Defrosting continues to be written as Defrosting recovery is written as The outdoor temperature, outdoor humidity, indoor average temperature, and indoor set temperature are read from the operating feature vector of any device location. These four components remain the same for all device locations, and are identified according to the current operating conditions. In the threshold parameter table Addressing in the middle, obtaining the first The location of each device corresponds to Current operating condition judgment threshold The construction method is as follows:

[0083] ;

[0084] In the formula, Indicates time index First The current operating condition determination threshold vector for each device location. Indicates the current operating condition. The corresponding baseline threshold vector, , , , , , , , Indicates the current operating condition. The corresponding operating condition adjustment coefficient vector, Indicates time index The outdoor temperature at that location Indicates time index The average indoor temperature at that location Indicates time index The indoor set temperature is Indicates time index Outdoor humidity at the location Indicates time index First The supply and return water temperature difference at each equipment location, Indicates time index First The circulating flow rate at each device location; for device locations without circulating flow rate measurement points, [the following will be observed]. Written as , Indicates time index First The differential pressure at each equipment location; for equipment locations without differential pressure measuring points, [the following will be considered]. Written as , Indicates time index First The heat pump power at each equipment location; for equipment locations without heat pump power measurement points, [the following will be included]. Written as , Indicates time index First The defrost status code is assigned to each device location. Device locations without defrost status measurement points will be... Written as , Indicates time index Current operating status indicator at the location Indicates the device location index. Indicates a unified time index. It represents absolute value operation. The seven components of the eight sets of working condition adjustment coefficient vectors correspond to seven types of judgment thresholds. The unit of each component is pre-calibrated according to the corresponding threshold unit to ensure that the dimensions of each summation item are consistent.

[0085] Current operating condition vector Input normal operation trajectory layer, the normal operation trajectory layer consists of The transition sublayer of 1 neuron and The reconstructed sublayer of each output neuron is composed of, first, Input transition sublayer to obtain 3D reconstruction of intermediate vectors Then With the Device location code for each device location They are jointly sent to the reconstruction sublayer and identified by the current working condition. exist Select one group from the group of working condition reconstruction parameters, and then... Reconstruction of equipment locations dimensional normal value vector , Each output neuron corresponds sequentially to the normal values ​​of supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrosting status, in the order of connection. Arrange all Get the time index Normal operating trajectory The normal operating trajectory covers all equipment locations on the heat source side, circulation side, and terminal side;

[0086] The first The operating feature vector of each device location The first seven components and Compare each device location individually, checking the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, and heat pump power. The absolute value of the difference between the actual and normal values ​​should not be less than [amount missing]. When corresponding components are involved, an initial deviation record is generated. , Indicates deviation variable, Indicates the deviation amount. Indicates the direction of the difference. This indicates that when the device identifier shows an actual value higher than the normal value, it will... Written as When the actual value is lower than the normal value, Written as The defrosting state uses a length of Index window When a discrepancy is found between the actual defrost status code and the normal defrost status code within the index window, an initial deviation record for the defrost status is generated. The number of discrepancies within the index window is written as the deviation amount. If the actual defrost status code first enters a higher status or remains in a higher status for a longer period, the direction code is written as... When the actual defrosting state code enters a higher state later or remains there for a shorter period, the direction code is written as... index the time All initial deviation records are arranged in the order of their generation to form the initial deviation record sequence. , Indicates time index The initial number of deviation records at the location;

[0087] Based on the initial deviation record sequence Device connection matrix and connection order Each initial deviation record is mapped to the corresponding device segment, and then reordered in ascending order according to the heat source side, circulation side, and terminal side, as well as the intra-side connection order. The device segment code corresponding to the first initial deviation record after reordering is written as the deviation position. The first initial deviation record after reordering also provides the deviation direction. The heat source side appears first and the direction is coded as follows: When written as The heat source side appears first and the direction is coded as When written as The loop side appears first and the direction is encoded as When written as The loop side appears first and the direction is encoded as When written as The end side appears first and the direction is encoded as When written as The end side appears first and the direction is encoded as When written as The deviation position is used to characterize the specific position of the variable in the heating link corresponding to the equipment connection relationship in which the heating operation data deviates from the normal operation trajectory. The deviation direction is used to characterize the direction of the difference between the heating operation data and the normal operation trajectory, as well as the order in which the difference appears on the heat source side, the circulation side, and the terminal side.

[0088] Extract the linked changes from the reordered initial deviation record sequence. When the deviation record of supply and return water temperature difference first appears on the heat source side, encode the order of supply and return water temperature difference changes. Written as When it first appears on the loop side, it is written as When it first appears on the distal side, it is written as When there is no record of supply and return water temperature difference deviation, it is written as The directional encoding of both the circulating flow deviation record and the differential pressure deviation record is simultaneously... At that time, the correspondence between circulating flow rate and pressure difference is encoded. Written as When the direction code is one positive and one negative, it is written as The direction encoding is also When written as If any record is missing, write as Heat pump power deviation records and defrost status deviation records are in the index window. When the same time index appears, the correspondence between heat pump power changes and defrost status is encoded. Written as When it appears in an adjacent time index, it is written as When no corresponding relationship is formed, it is written as ,Will , and As three components of the linked changes, the first deviation of each of the following is extracted according to the reordered initial deviation record sequence: supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrost status. If no corresponding deviation record exists, it is written as... ,form Dimensional anomaly vector , of The components are, in order: supply water temperature deviation, return water temperature deviation, supply and return water temperature difference deviation, circulation flow deviation, pressure difference deviation, heat pump power deviation, defrost status deviation, deviation location code, deviation direction code, supply and return water temperature difference change sequence code, circulation flow and pressure difference correspondence code, and heat pump power change and defrost status correspondence code.

[0089] Arrange the current operating condition identifiers corresponding to the five time indices within the determination window into a current operating condition identifier sequence in chronological order. The current operating condition identifier sequence is An integer encoding matrix arranges the normal operation trajectories corresponding to five time indices into a normal operation trajectory sequence in chronological order. The normal operating trajectory sequence is A real tensor will determine whether the condition within the decision window is satisfied. Time indexes are denoted in ascending order as , This represents the number of abnormal time indices, and the corresponding abnormal behavior vectors are arranged in chronological order to form an abnormal behavior set. The set of abnormal behaviors is Real number matrix, current operating condition identifier sequence and normal operating trajectory sequence A set of abnormal behaviors formed under the same time index and the same device location order. The fault differentiation knowledge graph is obtained by combining the deviation position, deviation direction, and linkage changes within the same time index range. When calling the knowledge graph, the abnormal behavior set is directly read. When the fault comparison layer of the fault comparison digital twin model is invoked, the current operating condition identifier sequence is directly read. and normal operating trajectory sequence .

[0090] In this embodiment, S3 specifically refers to:

[0091] The fault differentiation knowledge graph is denoted as , The network parameters of the fault differentiation knowledge graph are represented by the number of nodes exhibiting abnormal behavior for the same judgment window. The number of device nodes is denoted as The number of faulty nodes after screening is denoted as The number of candidate faults within an easily confused fault group is denoted as The actual input to the fault differentiation knowledge graph consists of four types of nodes and three types of connections. Each abnormal behavior node is written as , The components correspond sequentially to the following: supply water temperature deviation, return water temperature deviation, supply and return water temperature difference deviation, circulation flow rate deviation, pressure difference deviation, heat pump power deviation, defrost status deviation, deviation location code, deviation direction code, supply and return water temperature difference change sequence code, circulation flow rate and pressure difference correspondence code, and heat pump power change and defrost status correspondence code. Each device node is written as , Each component corresponds in sequence to whether it is a heat pump, a circulating pump, a heat exchanger, a valve, a terminal branch, whether it is located on the heat source side, whether it is located on the circulation side, whether it is located on the terminal side, and the connection order. In-degree, out-degree, direct upstream device type code, direct downstream device type code, number of hops along the device connection matrix to the heat pump, number of hops along the device connection matrix to the terminal branch, branch marker; in-degree is based on the device connection matrix. The The column takes values The number of elements is determined, and the out-degree is determined by the first element. The value in the row is The number of elements is determined, and the direct upstream equipment type code and the direct downstream equipment type code are respectively based on the number of elements. The device that is directly connected to the device with the closest connection order is determined; if no directly connected device exists, it is written as... The number of hops from the equipment connection matrix to the heat pump and the number of hops from the equipment connection matrix to the terminal branch are determined by the number of edges of the shortest connection path. If no connected path exists, it is written as... Branch markers are in-degree greater than Or out-degree is greater than When written as In other cases, it is written as , No. Each fault node is written as , Each component is individually written in the following order: heat exchanger frosting, insufficient circulation flow, valve malfunction, heat exchanger heat transfer degradation, circulation pump malfunction, heat pump performance degradation, bypass malfunction, pipeline blockage, sensor drift, and control logic malfunction. To ensure that the number of common malfunctions can be directly calculated, a reference coding vector is synchronously written to each fault node. The five components of the reference coding vector correspond sequentially to the deviation location coding, deviation direction coding, temperature difference change location coding, flow rate and pressure difference correspondence coding, and power change and defrost status correspondence coding. The reference coding vector is directly written from the fixed correspondence between fault category and equipment connection relationship, heat and water flow transfer relationship, and fault causality relationship, without being generated through network inference. Each condition node is written as , The condition nodes correspond sequentially to the following: temperature difference changes first on the heat source side, temperature difference changes first on the circulation side, temperature difference changes first on the terminal side, circulation flow and pressure difference decrease in the same direction, circulation flow and pressure difference change in opposite directions, heat pump power increase synchronized with defrosting status, heat pump power increase and defrosting status misaligned, and the abnormal landing point consistent with the equipment connection relationship. The fault node is not retrieved in full, but is first filtered according to the deviation position code in the abnormal manifestation node. When the deviation position code is on the heat source side, the following are retrieved: heat exchanger frosting, heat exchanger heat exchange attenuation, heat pump performance attenuation, sensor drift, and control logic abnormality. When the deviation position code is on the circulation side, the following are retrieved: insufficient circulation flow, circulation pump abnormality, pipeline blockage, valve abnormality, bypass abnormality, sensor drift, and control logic abnormality. When the deviation position code is on the terminal side, the following are retrieved: insufficient circulation flow, valve abnormality, bypass abnormality, pipeline blockage, sensor drift, and control logic abnormality.

[0092] The fixed matching relationship between equipment type and fault category is written as an equipment fault matching matrix. , No. The device type of the device node is the same as that of the first device node. The allowed device types for each fault node are consistent, and the first fault node... The device side of each device node and When the indicated device side is consistent, Written as If the aforementioned conditions are not met, Written as The input order of the fault differentiation knowledge graph in a single reasoning process is fixed as anomaly behavior node, device node, fault node, and condition node. The input sequence is written as follows: Node mapping layer settings One neuron, for dimension, dimension, peacekeeping The dimensional inputs are mapped to respectively The unified node representation, after completing node mapping, first establishes connections from abnormal behavior nodes to device nodes according to the consistency relationship between the deviation position code and the device side. When the deviation position code points to the heat source side, connect all heat source side device nodes; when the deviation position code points to the loop side, connect all loop side device nodes; when the deviation position code points to the end side, connect all end side device nodes. Then, according to the device fault adaptation matrix... Establish a connection from the device node to the faulty node, and then follow the reference coding vector. Establish a connection between the fault node and the condition node by establishing a consistent relationship with the condition node encoding, forming an initial relationship representation;

[0093] First graph relational layer settings There are 10 neurons, and the connection direction from the abnormal behavior node to the device node and the device connection matrix are as follows. The connection direction executes device segment propagation, and the first graph relation layer outputs on the same-side connected device nodes for each abnormal node. 3D localization response vector, The positioning score is obtained by summing the components. The device node with the largest positioning score is denoted as the nth node. The device that caused the abnormal behavior node is denoted by the device index. When location scores are the same, select the connection order. The smallest device node generates the device location result according to the above rules. The equipment positioning results will limit the deviation position and direction to the specific equipment segment on the heat source side, circulation side, or terminal side. The second map relationship layer is set. Each neuron performs link propagation on the device location results according to the relationship between heat and water flow transfer, and distributes all device location results according to the device side order and connection order. Arranged in ascending order, for nodes exhibiting abnormal behavior including deviations in supply and return water temperature difference, extract the equipment side where the deviation first appears to form the location of the temperature difference change. For nodes exhibiting abnormal behavior including deviations in circulating flow rate and pressure difference, extract the codes for two deviation directions. When both directions decrease simultaneously, it forms a same-direction decrease in the correspondence between flow rate and pressure difference; when one direction increases and the other decreases, it forms a different-direction change in the correspondence between flow rate and pressure difference. Then, retain the connection order position corresponding to each abnormal behavior node to form the link propagation result. The third graph relationship layer is set. The system uses a number of neurons and performs causal relationships on heat pump power changes and defrost status changes according to the fault causal relationship. When the heat pump power deviation is positive and the correspondence between heat pump power change and defrost status is encoded as synchronous, the corresponding fault node is connected to the node indicating the synchronization condition of heat pump power increase and defrost status. When the heat pump power deviation is positive and the correspondence between heat pump power change and defrost status is encoded as misaligned, the corresponding fault node is connected to the node indicating the misalignment condition of heat pump power increase and defrost status. The order of supply and return water temperature difference changes is encoded in three ways: first change on the heat source side, first change on the circulation side, and first change on the terminal side, corresponding one-to-one with the location of temperature difference changes. The correspondence between circulation flow rate and pressure difference is encoded one-to-one with the correspondence between flow rate and pressure difference. The correspondence between heat pump power change and defrost status is encoded one-to-one with the correspondence between power change and defrost status. For the first... The number of abnormal behaviors that occur together among the faulty nodes. Calculate using the following formula:

[0094] ;

[0095] in, Indicates the first The number of common abnormal behaviors of each faulty node. Indicates the index of the faulty node. Index of nodes exhibiting abnormal behavior. Indicates the number of nodes exhibiting abnormal behavior. This represents an indicator function, where the value is set to a value when the condition within the parentheses is met. The value is taken when the condition in parentheses is not met. , Represents the device fault adaptation matrix The Middle Line number Column elements, Indicates the first The corresponding device index in the device location results for each abnormal node. Indicates the first Offset position encoding in anomalous nodes Indicates the first Deviation direction encoding in anomalous behavior nodes. Indicates the first The sequence encoding of supply and return water temperature difference changes in each abnormal node. Indicates the first Encoding the correspondence between circulating flow and differential pressure in each abnormal node. Indicates the first The coding of the correspondence between heat pump power changes and defrosting status in each abnormal node. Indicates the first Offset position encoding in the reference encoding vector of each fault node Indicates the first Off-direction encoding in the reference encoding vector of each fault node Indicates the first The location encoding of temperature difference changes in the reference encoding vector of each fault node. Indicates the first The correspondence between flow rate and differential pressure is encoded in the reference coding vector of each fault node. Indicates the first Encoding the correspondence between power changes and defrosting status in the reference coding vector of each fault node;

[0096] Fix the quantity threshold to The number of abnormal manifestations occurring together is not less than The faulty node enters the fault merging layer, and the fault merging layer is configured. One output neuron, The output components correspond sequentially to the heat exchange attenuation group, hydraulic deficiency group, flow path regulation group, heat pump defrosting coupling group, measurement and control group, and terminal distribution group. The heat exchange attenuation group covers heat exchanger frosting and heat exchanger attenuation; the hydraulic deficiency group covers insufficient circulation flow, circulation pump malfunction, and pipe blockage; the flow path regulation group covers valve malfunction and bypass malfunction; the heat pump defrosting coupling group covers heat exchanger frosting, heat pump performance degradation, and control logic malfunction; the measurement and control group covers sensor drift and control logic malfunction; and the terminal distribution group covers valve malfunction, bypass malfunction, and insufficient circulation flow. The output score of each output neuron is not less than [value missing]. When, the corresponding component is written as The output score is less than When, the corresponding component is written as ,get Easily confused fault group vector The value is taken from the easily confused fault group vector. The components are expanded to form a candidate fault sequence. ;

[0097] Candidate Fault Sequence Each candidate fault in the process is processed sequentially using a distinguishing condition output, and the distinguishing condition output layer is configured. Each output neuron receives only one candidate fault's corresponding fault node representation, the device location result of the candidate fault's corresponding equipment, and the link propagation result of the candidate fault at a time, distinguishing the conditional output layer's preceding... The output neurons correspond to the temperature difference changing first on the heat source side, the temperature difference changing first on the circulation side, and the temperature difference changing first on the terminal side, respectively. The one with the largest score is recorded as the location of the temperature difference change. The output neuron and the first The output neurons correspond to the same decrease in circulating flow rate and pressure difference, and the opposite change in circulating flow rate and pressure difference, respectively. The one with the largest score is taken as the correspondence between flow rate and pressure difference. The output neuron and the first Each output neuron corresponds to a synchronous increase in heat pump power with defrosting status and a misaligned increase in heat pump power with defrosting status. The neuron with the largest score is used to represent the correspondence between power change and defrosting status. Each output neuron is used to verify the device positioning result and the device connection matrix. Whether they match, the output score is not less than When written as The output score is less than When written as Thus generating the first The candidate faults correspond to dimensional conditional vector , of The components correspond sequentially to the following: temperature difference changes first on the heat source side, temperature difference changes first on the circulation side, temperature difference changes first on the terminal side, circulation flow rate and pressure difference decrease in the same direction, circulation flow rate and pressure difference change in opposite directions, heat pump power increase and defrost status are synchronized, heat pump power increase and defrost status are misaligned, and the abnormal landing point is consistent with the equipment connection relationship. Each condition vector... Write back to the corresponding condition node, and in the order of candidate faults. The order of arrangement is written as a fault differentiation condition table. , for When reading the matrix and fault-referenced digital twin model, the candidate fault sequence is used. Read in one-to-one order Easily confused fault group vector Fault Differentiation Criteria Table All sectors.

[0098] In this embodiment, S4 specifically refers to:

[0099] Read the easily confused fault group vector formed in step S3 Candidate Fault Sequence Fault Differentiation Criteria Table ,in Indicates the number of candidate faults, and reads the current operating condition identifier sequence formed in step S2. Normal operating trajectory sequence Abnormal behavior set Device connection matrix Equipment Fault Adaptation Matrix and current working condition judgment threshold ,in Indicates the number of device locations. Indicates the number of faulty nodes. Indicates the number of vectors exhibiting abnormal behavior. of Each component corresponds sequentially to the current operating condition judgment thresholds for supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrosting status, and is used in a fault comparison digital twin model. Fault control layer index at each time Directly retrieve the code already generated by the working condition coding layer Current working condition vector , for the One candidate fault, the fault comparison layer will , and condition vector Cascaded in sequence dimensional core input vector, Each component is composed of Current working condition vector, Easily confused fault group vector sum The fault control layer is constructed using dimensional conditional vectors. The first fully connected layer of neurons, The second fully connected layer of neurons and The trajectory generation head of each output neuron performs one forward inference, and the trajectory generation head's... The output components correspond to the effects of supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrost status, respectively.

[0100] For the first Candidate faults First read the condition vector ,when When the action side is limited to the heat source side, when When, the action side is limited to the cyclic side, when At that time, the effective side is limited to the end side, and then the equipment fault adaptation matrix is ​​used. The middle screening meets the requirements Furthermore, the device location index whose device side is consistent with the aforementioned function side, and the device location index with the smallest connection order, is denoted as . ,by Starting from the device connection matrix Generate candidate fault action sequence ,in , Indicates the first The action sequence of candidate faults is determined by the length of the action link. For the heat source side, candidate faults are arranged in sequence according to the medium flow direction of the heat pump, heat exchanger, circulating pump, and terminal branch. For the circulating side, candidate faults are arranged in sequence according to the medium flow direction of the circulating pump or valve, adjacent pipe nodes, and terminal branch, with upstream equipment directly connected to the starting point added. For the terminal side, candidate faults are arranged in sequence according to the return water direction of the terminal branch, valve, circulating pump, and adjacent equipment on the heat source side. or At that time, retain all equipment locations capable of transmitting circulating flow and differential pressure in the candidate fault action sequence. or At that time, the heat pump position and the heat exchanger position directly connected to the heat pump are forcibly retained in the candidate fault action sequence. At that time, the candidate fault action sequence is truncated to contain only A single-position sequence;

[0101] For each device location in the candidate fault action sequence and each time index From the normal operating trajectory Read the corresponding dimensional normal value vector Then read the current operating condition threshold of the same device location. ,Will Subtract each component For the corresponding components, we obtain dimensional lower boundary vector ,Will Each component plus For the corresponding components, we obtain Upper boundary vector Supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, and heat pump power are retained according to real number boundaries. The components corresponding to the defrost state are restricted to the discrete coding interval after the upper and lower boundaries are calculated. ,Will and Cascaded in sequence Current working condition boundary vector , Each component corresponds sequentially to the lower boundary of supply water temperature, the upper boundary of supply water temperature, the lower boundary of return water temperature, the upper boundary of return water temperature, the lower boundary of supply and return water temperature difference, the upper boundary of supply and return water temperature difference, the lower boundary of circulation flow rate, the upper boundary of circulation flow rate, the lower boundary of differential pressure, the upper boundary of differential pressure, the lower boundary of heat pump power, the upper boundary of heat pump power, the lower boundary of defrost status, and the upper boundary of defrost status. All components are arranged according to time index and connection order. , forming the first The sequence of operating condition constraints corresponding to each candidate fault;

[0102] For the first Candidate Faults and Time Index ,Will The core input vector is fed into the fault control layer to obtain the starting point of action. of 3D Fault Effect Vector Perform bitwise recursion on the candidate fault action sequence, the first... The upstream propagation difference vector at each device location is denoted as... ,when At that time, Written as A zero-dimensional vector, when At that time, the fault reference value vector already generated at the previous equipment location is subtracted from the normal value vector at the previous equipment location by components to obtain... When the equipment is located as a heat pump or heat exchanger, retain the corresponding components of supply water temperature, return water temperature, supply and return water temperature difference, heat pump power, and defrost status. When the equipment is located as a circulating pump, valve, or main pipeline node, retain the corresponding components of supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, and pressure difference. When the equipment is located as a terminal branch, retain the corresponding components of return water temperature, supply and return water temperature difference, and circulation flow rate. Components not permitted to propagate are... and The Chinese text is written directly as , No. The candidate fault is in the first Device location and time index place Dimensional Fault Comparison Value Vector Generate using the following formula:

[0103] ;

[0104] In the formula, Indicates time index First The candidate fault is in the first The location of each functioning device is generated Dimensional fault reference value vector, Indicates time index First The candidate fault is in the first Location of each functional device Upper boundary vector, Indicates time index First The candidate fault is in the first Location of each functional device dimensional lower boundary vector, Indicates time index First The location of each functional device is within the normal operating trajectory dimensional normal value vector, Indicates the first The impact of each candidate fault on the link length. Indicates the position index in the candidate fault action sequence. Indicates time index First A candidate fault is generated at the point of action. 3D fault impact vector Indicates time index First The candidate fault is in the first Location of each functional device Upstream propagation difference vector, Indicates time index First The starting point of each candidate fault is within the normal operating trajectory dimensional normal value vector, Indicates the first The first candidate fault action sequence Device location index. Indicates the first The starting device location index in the candidate fault action sequence. Indicates the candidate fault index. Indicates a unified time index. and Both indicate truncation operations performed on a component basis;

[0105] After processing all affected device locations according to the above recursive rules, device locations not in the candidate fault action sequence will be directly retained in the corresponding normal operation trajectory. To maintain the normal value, the device location in the candidate fault action sequence is written as the corresponding and for the first The defrost state corresponding to each component is mapped using the nearest integer, limited to... , , or Arranged in order of device location to obtain the time index Fault comparison trajectory at the location Then arrange them in chronological order to get the first Fault reference trajectory sequence corresponding to each candidate fault Fault comparison trajectory The components are always arranged in the following order: supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrosting status.

[0106] For each candidate fault, from and Extract the relational quantities consistent with the fault differentiation condition table. Organize all equipment locations on the heat source side, circulation side, and terminal side into three side-specific index sets. Then, sequentially check whether the difference between the fault control value and the normal value of the supply and return water temperature difference for each side within the five time indices reaches the corresponding threshold. The supply and return water temperature difference components, the side that first meets the condition is written as the order of change of the supply and return water temperature difference, and the side that appears first is denoted as the heat source side. When the loop side appears first, it is denoted as When the terminal side appears first, it is recorded as When none of the sides reach the threshold, it is recorded as Regarding the circulating flow rate and pressure difference, firstly with... or Find the first time index that reaches the threshold in the corresponding device location, then compare the direction of the fault reference value relative to the normal value at the same time index. When the circulating flow rate decreases and the differential pressure decreases, record the linkage direction of the circulating flow rate and differential pressure as follows: When the circulating flow rate is opposite to the pressure difference, it is denoted as When the circulating flow rate increases and the differential pressure increases, it is recorded as... When no valid correspondence is formed, it is recorded as For heat pump power and defrost status, locate the first heat pump power increase time index and the first defrost status code change time index at the heat pump location. If the two time indices are the same, record the correspondence between heat pump power change and defrost status as follows: The two time indices differ Each sampling step size is used to record the relationship between heat pump power changes and defrost status. Other cases are recorded as Then , and Encoded as dimensional candidate relation vector , of The components correspond sequentially to the following: temperature difference changes first on the heat source side, temperature difference changes first on the circulation side, temperature difference changes first on the terminal side, circulation flow rate and pressure difference decrease in the same direction, circulation flow rate and pressure difference change in opposite directions, heat pump power increase and defrosting state are synchronized, and heat pump power increase and defrosting state are out of sync. or hour, The The component and the first Each component is written as ,when hour, The The component and the first Each component is written as ;

[0107] Matching layer uses Matching fully connected layers of neurons and The fault matching head of the output neuron, for the first The candidate fault and the first Anomalous behavior vectors ,Will , and Cascaded in sequence Dimensional matching input vector , among which the former Each component comes from the abnormal behavior vector, the middle one Each component comes from the candidate relation vector, and finally... Each component comes from the condition vector, and the matching layer pairs Perform one forward inference and generate Temporary fault matching vector , The components are arranged in the following order: heat exchanger frosting, insufficient circulation flow, valve malfunction, heat exchanger heat transfer degradation, circulation pump malfunction, heat pump performance degradation, bypass malfunction, pipeline blockage, sensor drift, and control logic malfunction. Matching is performed item by item: when The deviation position code is consistent with the side of the first device position in the candidate fault action sequence. The deviation direction code is consistent with the first valid change direction of the fault reference trajectory. The order coding of supply and return water temperature difference changes and Consistent The correspondence between circulating flow rate and pressure difference is encoded and Consistent The coding of the correspondence between heat pump power change and defrost status Consistent, and At that time, candidate faults The corresponding fault category component is written as If any condition is not met, the candidate fault will be... The corresponding components are written as The matching vectors of all temporary faults corresponding to the same candidate fault are obtained by taking the maximum value of each component. Fault matching results According to the candidate fault sequence Arrange all in order The fault matching result matrix is ​​obtained. .

[0108] In this embodiment, S5 specifically refers to:

[0109] Fault Differentiation Knowledge Graph In the exclusion phase, relation propagation training is no longer performed. Instead, a line-by-line search is executed according to the candidate fault order. The actual search input for a single candidate fault includes... 3D fault matching result vector , dimensional conditional vector Candidate Fault Identifier First active device location index First functional device location code First functional device connection sequence and abnormal behavior set Conditional vector of The components correspond sequentially to the following: temperature difference changes first on the heat source side; temperature difference changes first on the circulation side; temperature difference changes first on the terminal side; circulation flow rate and pressure difference decrease in the same direction; circulation flow rate and pressure difference change in opposite directions; heat pump power increase is synchronized with defrosting status; heat pump power increase and defrosting status are misaligned; and the abnormal landing point is consistent with the equipment connection relationship. The fault differentiation knowledge graph outputs for each candidate fault. Candidate fault retention record ;

[0110] For the first First, identify the candidate faults according to a fixed fault dictionary. Mapped to fault category index Then Write to the corresponding faulty node, Write the corresponding condition node, and then read the earliest anomaly vector from the anomaly behavior set. and take the first Each component is used as the earliest deviation position code, and then the extracted supply and return water temperature difference change sequence is read. Circulation flow rate and differential pressure linkage direction Correspondence between heat pump power changes and defrosting status Based on this, candidate fault retention records are generated. ,in This indicates that the location of the temperature difference change meets the marking. This indicates that the relationship between flow rate and pressure difference satisfies the flag. This indicates that the relationship between power change and defrosting status satisfies the flag. This indicates a deviation from the correct location and is consistent with the device connection. Indicates that candidate faults are reserved;

[0111] when The Each component is ,and correspond or correspond or correspond At that time, Written as ;

[0112] when correspond or correspond At that time, Written as ,when correspond or correspond At that time, Written as ;

[0113] When the earliest deviation from the position code is on the side The sides are consistent and At that time, Written as Unsatisfied flags are directly written as ;

[0114] when , and At that time, Written as There exists any relation marked as At that time, Written as ;

[0115] All are listed in order of candidate faults. ,form The candidate fault retention result sequence;

[0116] All will be satisfied The candidate fault identifiers form the remaining candidate fault set. For all conditions are met Candidate faults, generate Dimensional missing record ,when At that time, The former Each component was written as is. The former Each component, when At that time, The The component and the first Each component was written as is. The The component and the first Each component, when At that time, The The component and the first Each component was written as is. The The component and the first Each component, when At that time, The Each component is written The Each component, all positions that have been satisfied are written as... Arrange all missing records in the order of candidate faults to form Missing sequence judgment criteria ;

[0117] When the remaining candidate fault set When there is only one candidate fault, the index corresponding to the candidate fault will be retained and denoted as . Then, in the abnormal behavior set, search in chronological order for results related to... The index of the anomaly vector that first satisfies the condition and is consistent with the anomaly vector on the same side is denoted as . Then place the first active device position First functional device location code First functional device connection sequence and The offset position encoding association forms the root cause determination result. , of Each component corresponds to the following in sequence: candidate fault identifier, root cause device location index, root cause device location code, root cause device connection order, and abnormal behavior deviation location code.

[0118] When the remaining candidate fault set When there are multiple candidate faults, Write as the fault group to be checked Then, the condition vectors corresponding to all retained candidate faults are compared column by column. If there is a difference in value between a certain column and the retained candidate faults, then that column is written into an additional missing record. The additional missing record is used to mark the fault differentiation condition entries in the current abnormal behavior set that have not yet been differentiated. The position of temperature difference change corresponds to the first... To the The column shows the relationship between flow rate and differential pressure, corresponding to the [number]th [column]. The component and the first The relationship between power change and defrosting status corresponds to the first component. The component and the first Each component, the abnormal landing point is consistent with the device connection relationship, corresponding to the first... The column will append additional missing records to the missing records determination criteria sequence. middle;

[0119] When the remaining candidate fault set When no candidate faults are included, the missing fault criterion sequence is used for judgment. Candidate fault identifiers with at least one non-zero component are selected from the list, and these selected candidate fault identifiers are grouped into fault groups to be checked. If the judgment criteria sequence is missing If all candidate faults have non-zero components, then all candidate fault identifiers are written into the fault group to be checked. Therefore, the exclusion judgment is always performed around the temperature difference change location, the flow rate and pressure difference correspondence, and the power change and defrost status correspondence in the fault differentiation condition table. The process of retaining candidate faults, screening candidate faults, generating root cause judgment results, and generating fault groups to be checked and missing judgment basis sequences is completed one by one.

[0120] In this embodiment, S6 specifically refers to:

[0121] This step uses a deterministic alignment and summarization method to read the root cause determination results. Fault group pending verification Missing information is determined based on the sequence. Candidate Fault Sequence Fault Differentiation Criteria Table Fault matching result matrix Abnormal behavior set Candidate Fault Response Sequence Device connection matrix Equipment location coding sequence Connection order sequence Current operating condition identifier sequence and normal operating trajectory sequence Root cause determination results of Each component sequentially retains the candidate fault identifier, root cause device location index, root cause device location code, root cause device connection order, and abnormal behavior deviation location code. Each missing record... of Each component corresponds sequentially to the following: temperature difference changes first on the heat source side; temperature difference changes first on the circulation side; temperature difference changes first on the terminal side; circulation flow rate and pressure difference decrease in the same direction; circulation flow rate and pressure difference change in opposite directions; heat pump power increase and defrost status are synchronized; heat pump power increase and defrost status are misaligned; and the abnormal landing point is consistent with the equipment connection relationship. Each fault matching result vector of The components are arranged in the following order: heat exchanger frosting, insufficient circulation flow, valve malfunction, heat exchanger heat transfer degradation, circulation pump malfunction, heat pump performance degradation, bypass malfunction, pipeline blockage, sensor drift, and control logic malfunction. The first component of each malfunction manifestation vector is... The components up to the first Each component corresponds sequentially to the deviation position code, deviation direction code, supply and return water temperature difference change sequence code, circulation flow rate and pressure difference linkage direction code, and heat pump power change and defrost status correspondence code. Current operating condition identifier sequence for Integer matrix, normal operation trajectory sequence for Real tensors, denoted as target faults The associated device location index is denoted as The positioning reference vector is denoted as , of Each component corresponds sequentially to the deviation position code, deviation direction code, supply and return water temperature difference change sequence code, circulation flow rate and pressure difference linkage direction code, and heat pump power change and defrost status correspondence code.

[0122] when When not empty, in the candidate fault sequence Search and No. The candidate fault identifiers with consistent components are denoted as the retrieved row index. And the same candidate fault identifier is written as the target fault. ,Will No. The device location index corresponding to each component is denoted as . ,Will No. The deviation position code corresponding to each component is denoted as Search the abnormal behavior set in chronological order for the code that deviates from the position. The first abnormal behavior vector retrieved is used as the alignment abnormal behavior vector, if the device location index... If the device side and the offset position encoding of the alignment anomaly manifestation vector correspond to the same side, then... Write it directly as If the device location index If the devices involved are not on the same side, then in the candidate fault action sequence Search Device Location Index ,satisfy or And device location index The device side and alignment anomaly vector are consistent, and the device position index with the smallest connection order is written as... If no directly connected device location index is found, The first device location index with consistent side is written as This aligns the root cause analysis results, device connectivity relationships, and abnormal behavior sets with any discrepancies, and identifies the target fault. and associated device location index ;

[0123] In determining and Then, read the first... Row condition vector and the Row fault matching result vector ,when Mid-target failure Corresponding components are At that time, retrieve the device location index from the abnormal behavior set in chronological order. The anomaly vector with consistent characteristics across different devices is used to encode the first non-zero offset position. The Each component, the first non-zero deviation direction code retrieved is written into... The Each component, and forward The sequence encoding of the supply and return water temperature difference changes at the same activation location of each component is written into the code. The Each component, and No. The component or the first The activation positions of each component are consistent with the direction encoding of the circulating flow and differential pressure linkage. The Each component, and No. The component or the first The correspondence between heat pump power changes and defrost status at each component activation location is encoded and written. The If no corresponding code is found for a component, the corresponding component is written as... Then from the normal operating trajectory sequence Extract associated device locations exist to The normal value vectors at five time indices form Each The components of the normal value vector correspond, in order, to the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrost status. , , , , , and Combined into fault location results ,when When empty, no fault location result is generated. ;

[0124] When the fault group is pending verification When not empty, proceed in the candidate fault sequence. The order of arrangement retains the fault identifiers to be checked, forming a sequence of faults to be checked. ,in Indicates the number of faults to be verified, for the first fault... One fault icon pending verification In the candidate fault sequence Searching for the same fault identifier in the middle, the retrieved row index is denoted as Then read the missing records. Write the manual verification record as ,when The The component, the first The component or the first Each component is At that time, respectively Write the code corresponding to the temperature difference changing first on the heat source side, the temperature difference changing first on the circulation side, or the temperature difference changing first on the terminal side, where all three components are... At that time, Written as ,when The Each component is At that time, Written as the code corresponding to the circulating flow rate and pressure difference decreasing in the same direction, when The Each component is At that time, Write it as the code corresponding to the opposite direction of the change in circulating flow rate and pressure difference, where both components are... At that time, Written as ,when The Each component is At that time, Write the code corresponding to the synchronous increase in heat pump power and defrosting status, when The Each component is At that time, This is written as the code corresponding to the increase in heat pump power and the misalignment of defrosting status, with both components being... At that time, Written as ,when The Each component is At that time, , and These are written as candidate fault action sequences. The first device location index, the corresponding device location code, and the corresponding device connection order, when The Each component is At that time, , and Simultaneously written as Arrange all manually checked records in the order of the faults to be checked to form the manual check results. .

[0125] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0126] This invention directly addresses the core issues in cold-region building heating diagnosis scenarios through a closed-loop technical path of normal trajectory comparison under operating condition identification constraints and knowledge-enhanced exclusion judgment. When outdoor temperature and humidity fluctuations and defrosting stage switching cause operating characteristics to drift with operating conditions and equipment topology coupling causes multiple types of faults to present similar appearances, relying solely on single-point thresholds or static rules can easily lead to difficulty in distinguishing candidate faults and cause business consequences of unstable positioning basis. Under the constraint of only available data such as supply and return water temperature, circulation flow, pressure difference, heat pump power, defrosting status, indoor and outdoor environment alarm logs, and equipment ledgers, this invention first constructs equipment connection relationships based on equipment correspondence, and encodes and fuses the segmented inputs of the heat source side, circulation side, and terminal side within the judgment window to form the current operating condition vector and the current operating condition identifier. Under the constraint of this operating condition identifier, the normal operating trajectory is reconstructed and compared with the operating data to obtain an abnormal performance set including the deviation position, deviation direction, temperature difference change order, flow rate, pressure difference linkage, power, and defrosting correspondence, making abnormal evidence comparable on the same operating condition benchmark.

[0127] Based on this, the present invention inputs the set of abnormal manifestations and the equipment connection relationships into the fault differentiation knowledge graph. It propagates and reasons along the equipment connection relationships, heat and water flow transfer relationships, and fault causal relationships, categorizes candidate faults into easily confused fault groups, and generates a fault differentiation condition table. The key intermediate quantities on which fault differentiation depends are explicitly carried by the condition table and can be reliably invoked. Subsequently, based on the same current operating condition identifier and the same normal operating trajectory, the fault comparison digital twin model generates fault comparison trajectories for each candidate fault only within the scope of the easily confused fault groups. It extracts the comparison relationship quantities according to the fault differentiation condition table and matches them item by item with the set of abnormal manifestations. Finally, the knowledge graph performs elimination judgment based on the same condition table to form the root cause judgment result or outputs the fault group to be checked and the missing judgment basis. This makes the diagnosis process from anomaly discovery to candidate convergence to localization basis output a closed-loop and deliverable process.

Claims

1. A diagnostic method for heating systems in cold-region buildings based on digital twins and knowledge enhancement, characterized in that, include: S1. Obtain the supply and return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, indoor and outdoor environment, alarm logs and equipment ledger of the heating system for cold-region buildings, and form heating operation data and equipment connection relationships according to the equipment correspondence. S2. Based on the heating operation data, call the fault comparison digital twin model to determine the current operating condition identifier and normal operating trajectory, and compare the heating operation data with the normal operating trajectory to obtain a set of abnormal manifestations including deviation position, deviation direction and linkage changes; S3. Based on the abnormal behavior set and equipment connection relationship, call the fault differentiation knowledge graph. First, group multiple candidate faults into easily confused fault groups according to the common abnormal behaviors. Then, along the equipment connection relationship, heat and water flow transfer relationship and fault causal relationship, determine the corresponding temperature difference change location, flow rate and pressure difference correspondence, power change and defrost status correspondence for each candidate fault in the easily confused fault group. Write each correspondence into the fault differentiation condition table and output the easily confused fault group and the fault differentiation condition table. S4. Based on the easily confused fault groups, fault differentiation condition table, current operating condition identifier, and normal operating trajectory, call the fault comparison digital twin model. Under the constraint of the current operating condition identifier, generate the fault comparison trajectory corresponding to each candidate fault. Then, according to the fault differentiation condition table, extract the order of supply and return water temperature difference changes, the direction of circulation flow and pressure difference linkage, and the correspondence between heat pump power changes and defrosting status from the fault comparison trajectory. Compare the extraction results with the abnormal performance set item by item to obtain the fault matching results of each candidate fault. S5. Input the fault matching results into the fault differentiation knowledge graph, and make exclusion judgments according to the fault differentiation condition table. When one candidate fault is retained, a root cause judgment result is formed. When multiple candidate faults are retained or no candidate faults are retained, a fault group to be checked and missing judgment basis are formed. S6. Based on the root cause determination results, output the target fault, associated equipment and location basis to form the fault location results. Based on the fault group to be checked and the missing judgment basis, output the manual verification results.

2. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement as described in claim 1, characterized in that, S1 specifically refers to: Acquire supply water temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature and indoor set temperature, and generate operational measurement point data according to the measurement point markings and acquisition equipment; Extract alarm category, alarm duration and corresponding device from alarm logs, and extract device identifier, device location, pipeline node and device type from device ledger to form device attribute data; Based on the operational measurement point data and equipment attribute data, determine the corresponding relationship of the equipment, and write the supply water temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, indoor and outdoor environment, alarm type, alarm duration and equipment location into the corresponding equipment according to the equipment correspondence. Then, calculate the supply and return water temperature difference based on the supply water temperature and return water temperature to form heating operation data. Based on the equipment correspondence and the pipeline nodes, medium flow direction and connection order in the equipment attribute data, the equipment association between heat pumps, circulating pumps, heat exchangers, valves and terminal branches is determined, and the equipment connection relationship is formed.

3. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement as described in claim 1, characterized in that, S2 specifically refers to: The heating operation data is arranged into heat source side input segment, circulation side input segment and terminal side input segment according to the medium flow direction in the equipment connection relationship. Then, the heat source side input segment, circulation side input segment and terminal side input segment are written into the working condition coding layer of the fault comparison digital twin model. The supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, defrost status, outdoor temperature, outdoor humidity, indoor average temperature, indoor set temperature, alarm category, alarm duration and equipment location are associated and mapped to form a segmented input sequence. Using the segmented input sequence as input, the heat source operating status, hydraulic transmission status, terminal load status and alarm triggering status are encoded segment by segment, and the segmented encoding results are fused according to the heat transfer path and water flow transfer path in the equipment connection relationship to form the current operating condition vector; Based on the current operating condition vector, the outdoor environment range, defrosting operation stage, heat pump output level, circulation flow status and terminal temperature difference demand are combined and judged to form the current operating condition identifier. Based on the current operating condition identifier, the current operating condition judgment thresholds corresponding to the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow, pressure difference, heat pump power and defrosting status are determined respectively. Input the current operating condition vector into the normal operating trajectory layer. Under the constraints of the current operating condition identifier, reconstruct the corresponding normal operating trajectory according to the equipment location for the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power and defrosting status, forming a normal operating trajectory covering the heat source side, circulation side and terminal side. The heating operation data is compared with the normal operation trajectory one by one according to the equipment location. For the difference that meets the current operating condition judgment threshold, the deviation variable, deviation amount, difference direction and the equipment that occurred are extracted to form the initial deviation result. Based on the initial deviation results and equipment connection relationships, the deviation variables are mapped to specific equipment segments in the heating chain along the equipment connection relationships, and sorted according to the order of appearance of the deviation variables on the heat source side, circulation side and terminal side to form the deviation position and deviation direction; Based on the correspondence between the deviation location, deviation direction, and supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power, and defrost status, the order of supply and return water temperature difference changes, the correspondence between circulation flow rate and pressure difference, and the correspondence between heat pump power changes and defrost status are determined to form a linkage change. The deviation location, deviation direction, and linkage change are combined into an abnormal behavior set. The current operating condition identifier, normal operating trajectory, and abnormal behavior set are provided for use in the fault differentiation knowledge graph and fault comparison digital twin model.

4. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement as described in claim 1, characterized in that, S3 specifically refers to: The deviation location, deviation direction, and linkage changes in the abnormal behavior set are written into the abnormal behavior node of the fault differentiation knowledge graph. The connection order of heat pump, circulating pump, heat exchanger, valve and terminal branch in the equipment connection relationship is written into the equipment node. The fault node and condition node corresponding to the heating operation data are retrieved to form the graph input sequence. The graph input sequence is mapped to nodes, and abnormal behavior nodes, device nodes, fault nodes and condition nodes are converted into unified node representations. The nodes are then aligned according to the connection direction between abnormal behavior nodes and device nodes, device nodes and fault nodes, and fault nodes and condition nodes to form an initial relationship representation. The initial relation representation is input into the first graph relation layer corresponding to the device connection relationship. The device segment propagation is performed on the deviation position and deviation direction to limit the location of the abnormality to the heat source side, circulation side or end side, thus forming the device positioning result. The equipment positioning results are input into the second graph relationship layer corresponding to the heat and water flow transfer relationship. The changes in supply and return water temperature difference, circulation flow rate and pressure difference are propagated through the link, and the changes are sorted according to the direction of medium flow to form the link propagation results. The link propagation results are input into the third graph relationship layer corresponding to the fault causal relationship. Causal correlation is performed on the heat pump power change and defrost status change. In the fault merging layer, fault nodes are screened according to the threshold of the number of common abnormal manifestations. Fault nodes that meet the combination of the same equipment connection relationship, heat and water flow transfer relationship and fault causal relationship are grouped into easily confused fault groups. Based on the easily confused fault group, equipment location results, and link propagation results, the location of temperature difference change is determined for each candidate fault within the easily confused fault group. The changes in circulation flow rate and pressure difference are correlated to form a correspondence between flow rate and pressure difference. Then, the changes in heat pump power and defrost status are correlated to form a correspondence between power change and defrost status, thus obtaining the differentiation condition results. The results of the differentiation conditions are written into the condition node, and the relationship between temperature difference change location, flow rate and pressure difference, and power change and defrosting status are summarized through the differentiation condition output layer to generate a fault differentiation condition table. The easily confused fault groups and the fault differentiation condition table are then provided for use by the fault comparison digital twin model.

5. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement according to claim 1, characterized in that, S4 specifically refers to: Input the easily confused fault group and the fault differentiation condition table into the fault comparison layer of the fault comparison digital twin model, and use the current working condition identifier and normal operating trajectory as the constraint input of the fault comparison layer. Determine the action link of each candidate fault according to the corresponding equipment in the connection relationship between candidate faults and equipment within the easily confused fault group to form the action sequence of candidate faults. Based on the current operating condition identifier, the current operating condition boundary is determined by the supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power and defrosting status in the normal operation trajectory. Then, the current operating condition boundary is written into the candidate fault action sequence to form the operating condition constraint sequence. Using the operating condition constraint sequence and normal operating trajectory as input, the fault impact calculation is performed on each candidate fault in the candidate fault action sequence. The impact of each candidate fault on the heat source side, circulation side and terminal side is expanded along the equipment connection relationship to supply water temperature, return water temperature, supply and return water temperature difference, circulation flow rate, pressure difference, heat pump power and defrosting status, forming the fault comparison trajectory corresponding to each candidate fault. Based on the location of temperature difference changes in the fault differentiation criteria table, the order of supply and return water temperature difference changes on the heat source side, circulation side, and terminal side is extracted from the fault comparison trajectory of each candidate fault to form the order of supply and return water temperature difference changes. Based on the correspondence between flow rate and differential pressure in the fault differentiation condition table, the direction of change of circulating flow rate and the direction of change of differential pressure are extracted from the fault comparison trajectory of each candidate fault, and the direction of change of circulating flow rate and the direction of change of differential pressure are matched to form the linkage direction of circulating flow rate and differential pressure. Based on the correspondence between power change and defrost status in the fault differentiation condition table, the heat pump power change and defrost status switching are extracted from the fault comparison trajectory of each candidate fault, and the heat pump power change and defrost status switching are sequentially matched to form the correspondence between heat pump power change and defrost status. The sequence of supply and return water temperature difference changes, the direction of linkage between circulation flow and pressure difference, and the correspondence between heat pump power changes and defrosting status are input into the matching layer. The deviation position, deviation direction, and linkage changes in the abnormal performance set are compared item by item to form the corresponding fault matching result for candidate faults that meet the fault differentiation condition table.

6. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement according to claim 1, characterized in that, S5 specifically refers to: The fault matching results are input into the fault nodes of the fault differentiation knowledge graph, and the fault differentiation condition table is input into the condition nodes. According to the location of temperature difference change, the correspondence between flow rate and pressure difference, and the correspondence between power change and defrosting status, the corresponding faults are searched to form the candidate fault retention results. The candidate fault retention results are excluded. When the fault matching result meets the temperature difference change location, flow rate and pressure difference correspondence, and power change and defrost status correspondence in the fault differentiation condition table, the corresponding candidate fault is retained. When the fault matching result lacks the temperature difference change location, flow rate and pressure difference correspondence, or power change and defrost status correspondence, the corresponding candidate fault is screened out to form the remaining candidate fault set. The remaining candidate fault set is counted. When the remaining candidate fault set contains a candidate fault, the retained candidate fault is associated with the deviation position and equipment connection relationship in the abnormal behavior set to form the root cause determination result. When the remaining candidate fault set contains multiple candidate faults or does not contain any candidate faults, the unmet fault differentiation condition entries are extracted to form the basis for missing fault judgment, and the remaining candidate fault set is identified as the fault group to be checked.

7. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement according to claim 1, characterized in that, S6 specifically refers to: Align the root cause determination results with the device connection relationship and the deviation position in the abnormal behavior set to determine the target fault and associated device corresponding to the retained candidate fault; Based on the target fault, abnormal manifestation set, fault differentiation condition table and fault matching results, the deviation location, deviation direction, supply and return water temperature difference change sequence, circulation flow and pressure difference linkage direction, heat pump power change and defrost status correspondence relationship are summarized to form the positioning basis; The target fault, associated equipment, and location data are associated with the current operating condition indicator and normal operating trajectory to form the fault location result; Match the fault groups to be checked with the missing judgment criteria, and determine the missing temperature difference change location, flow rate and pressure difference correspondence, or power change and defrosting status correspondence for each candidate fault in the fault group to be checked, so as to form the manual check results.

8. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement according to claim 3, characterized in that, The current operating condition identifier is used to constrain the corresponding reconstruction of the normal operating trajectory. Under the constraint of the current operating condition identifier, the heating operation data is compared with the normal operating trajectory to form an abnormal performance set. Under the constraint of the current operating condition identifier, the normal operating trajectory is used as the reference basis, and the fault comparison trajectory corresponding to each candidate fault is generated according to the easily confused fault group and the fault differentiation condition table. This makes the abnormal performance set and the fault comparison trajectory formed on the same current operating condition identifier and the same normal operating trajectory.

9. The method for diagnosing heating systems in cold-region buildings based on digital twins and knowledge enhancement according to claim 4, characterized in that, The easily confused fault group is used to determine the candidate fault range when the fault comparison digital twin model generates the fault comparison trajectory. The fault comparison digital twin model generates the fault comparison trajectory corresponding to each candidate fault according to the corresponding equipment in the equipment connection relationship of each candidate fault in the easily confused fault group. Based on the fault differentiation condition table, the order of supply and return water temperature difference change, the direction of circulation flow and pressure difference linkage, and the correspondence between heat pump power change and defrost status are extracted from the fault comparison trajectory corresponding to each candidate fault. The results are compared with the abnormal performance set item by item to form the fault matching result of each candidate fault.

10. A cold-region building heating diagnosis system based on digital twins and knowledge enhancement, used to execute the cold-region building heating diagnosis method based on digital twins and knowledge enhancement according to any one of claims 1-9, characterized in that, include: The data acquisition and construction module is used to acquire water supply temperature, return water temperature, circulation flow rate, pressure difference, heat pump power, defrost status, indoor and outdoor environment, alarm logs and equipment ledgers, and form heating operation data and equipment connection relationships according to equipment correspondence. The knowledge graph reasoning module configures a fault differentiation knowledge graph, which is used to output easily confused fault groups and fault differentiation condition tables based on the abnormal behavior set and device connection relationship. The digital twin comparison module is configured with a fault comparison digital twin model, which is used to determine the current operating condition identifier and normal operating trajectory based on heating operation data, and compare the heating operation data with the normal operating trajectory to form a set of abnormal manifestations including deviation position, deviation direction and linkage changes; The digital twin comparison module is also used to generate fault comparison trajectories corresponding to each candidate fault and form fault matching results after receiving easily confused fault groups and fault differentiation condition tables; The elimination judgment and location verification module is used to perform elimination judgment based on the fault matching results and the fault differentiation condition table. When one candidate fault is retained, the root cause judgment result is generated and the fault location result is output. When multiple candidate faults are retained or no candidate faults are retained, the fault group to be verified, the missing judgment basis, and the manual verification result are output.

Images