A cooling pump optimization method for central air-conditioning chillers based on knowledge graph

Abstract

The present invention relates to the field of machine learning technology, specifically a cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph, comprising the following steps: obtaining the alignment relationship between the pump start-stop signal and the current response time, extracting the action segments and the path node sequence to construct a logical sequence, identifying the control signal alternation and direction deviation to extract the path restriction number, and screening the response delay and load fluctuation segment corresponding adjustment identification area. In the present invention, by constructing a behavioral mapping between the pump action and the signal timing, extracting the alternating signal and direction deviation fragments in the control chain, generating a path restriction identification for constraining control interference, combining the load response with the matching regulation trigger area of ​​the equipment action segment, screening the path that can be accessed to replace the restricted segment, reorganizing the path structure and sorting the instruction chain, enhancing the matching ability of the path scheduling and the response synergy of the linkage node, and promoting the control process to maintain continuity and stability under differentiated states.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph. Background Art

[0002] The field of machine learning technology involves building models based on data and adjusting model parameters through training to achieve prediction and discrimination of unknown data. Core issues include feature extraction, model training, algorithm optimization, and inference control. It is widely used in scenarios such as intelligent control, relying on statistical analysis and iterative solutions to build systems with decision-making capabilities. Among them, traditional central air-conditioning chiller cooling pump optimization methods refer to control methods that adjust the start and stop and operating status of cooling pumps by setting thresholds and logical judgments based on data such as ambient temperature and humidity, load, and water temperature. They are usually implemented using parameter settings based on empirical rules and fixed-time control instructions. They lack dynamic learning and data association analysis capabilities, and are difficult to adapt to the optimization control needs under changing system states.

[0003] Existing technologies mainly use fixed control logic in the equipment control process, and rely on static data settings that cannot reflect the mapping relationship between equipment action timing and signals. Path identification lacks a basis for judging direction deviations and changes in action sequences, and control decisions often lag in a state. In scenarios where there is alternating interference or sequence dislocation between node responses, it is impossible to clearly identify effective control paths, resulting in repeated instructions, redundant responses, and path conflicts during the system execution process. The control chain lacks flexible adjustment capabilities, and in scenarios where the cooling load frequently fluctuates or the equipment load coupling is high, it is easy to cause response imbalance and path interference, affecting the continuous achievement of the system regulation goals. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and propose a cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph, comprising the following steps:

[0006] S1: Obtain the start and stop control time points of the cooling pump, call the current startup monitoring data, align the signal time with the pump number, locate the section where each response action is located according to the time sequence, write the pump number and action time into the map node set, and obtain the pump response behavior mapping set;

[0007] S2: Based on the pump response behavior mapping set, retrieve the action time interval in the pump response behavior mapping set, read the cold water outlet temperature and cooling load feedback node, locate the node response time sequence position, compare the action sequence with the path number relationship, and obtain the control path node logic sequence;

[0008] S3: Based on the control path node logic sequence, extract adjacent control signals in the control path, read the start / stop status data, determine whether the time position and direction are consistent, identify signal alternation and opposite direction paths, extract the path number and add it to the graph index to obtain the conflict path restriction identification set;

[0009] S4: Based on the conflict path restriction identification set, the conflict path number is input, the cooling load response data is retrieved, the response delay path segments are screened, the equipment action time period is identified, the overlapping response segments and the load fluctuation segments are matched, and the control node trigger adjustment identification area is obtained.

[0010] As a further solution of the present invention, the pump response behavior mapping set includes pump number, action time, and response segment; the control path node logic sequence includes temperature node position, load node timing, and control path number; the conflict path restriction identification set includes signal direction conflict number, path conflict sequence, and control index label; the control node trigger adjustment identification area includes response delay segment, action coverage range, and load change interval.

[0011] As a further solution of the present invention, the cooling load feedback node refers to a node that returns monitoring results to the target control node of the control path during the operation of the central air-conditioning chiller. The node collects real-time changes in the cold water outlet temperature, flow rate, and equipment power, and converts the data into feedback information that can be used for control logic analysis.

[0012] The response timing position refers to analyzing the response behavior of the control node and mapping the time point of each response action to the time axis.

[0013] As a further solution of the present invention, the graph index refers to locating the control path and node relationship through the index number in the knowledge graph structure, and corresponding the path segment number and node number through the index content;

[0014] The cooling load response refers to calling the cold water outlet temperature and pump load fluctuation response data during the control behavior triggering process of the central air-conditioning chiller, and combining the time axis of the control signal with the control signal time and load behavior.

[0015] As a further solution of the present invention, the specific steps of S1 are:

[0016] S101: Obtain the start and stop control signal time points of the cooling pump in the central air conditioning chiller operation log, filter the pump number and issuance time corresponding to the signal field, match the signal time with the pump number in chronological order, and associate the starting control signal with the response signal to obtain a signal time alignment sequence;

[0017] S102: Based on the signal time alignment sequence, extract the current startup monitoring data corresponding to the pump number, locate the time period of the current rising boundary, compare the relationship between the control signal time fields, identify the time segments of the control and response actions, screen the time period and number matching results of the response action, and obtain a pump number response time matching list;

[0018] S103: Based on the pump number response time pairing list, filter the response time corresponding to each number, arrange according to the time field, write each group of pump numbers and action time into the map device node field in sequence, and connect them according to the corresponding timing of the response action to obtain the pump response behavior mapping set.

[0019] As a further solution of the present invention, the specific steps of S2 are:

[0020] S201: Based on the action time interval in the pump response behavior mapping set, collect response time data of the cold water outlet temperature sensor within the time period, read the time field and corresponding device number of each data, write the number and time information into the node queue in sequence, and obtain a temperature response node sequence;

[0021] S202: Based on the temperature response node sequence, collecting response time data of cooling load feedback nodes in the same period, counting the number of times each group of feedback nodes and temperature nodes appear in the same time segment, calculating the frequency ratio of each group of feedback nodes to the temperature nodes in a specified period, and obtaining a feedback temperature correlation result;

[0022] S203: Based on the feedback temperature association result, obtain the node number data in the control path, read the response time corresponding to each group of numbers, compare the number time with the feedback action time, and put the node numbers into the path node sequence in chronological order to obtain the control path node logical sequence.

[0023] As a further solution of the present invention, the specific steps of S3 are:

[0024] S301: Based on the device control signals between adjacent nodes in the control path node logic sequence, call the start and stop status data of the adjacent nodes in the control cycle, identify the start and end symbols of each group of control directions, calculate the direction switching frequency of each pair of adjacent control actions, and obtain a direction switching frequency detailed table;

[0025] S302: Based on the direction switching frequency detailed table, extract the node pair numbers of the direction switching, remove duplicates in the numbers, call the time sequence data in the node sequence, and classify the nodes with direction changes into number sets according to the time sequence to obtain a node set that controls direction changes;

[0026] S303: Based on the control direction change node set, extract the path segment number associated with the number in the graph structure, retrieve the control path number to which the path segment belongs, write the control path number into the path list, and obtain the conflict path restriction identifier set.

[0027] As a further solution of the present invention, the specific steps of S4 are:

[0028] S401: Based on the conflict path restriction identifier set, the path number is input, the time series data of the control signal and the cooling load change is retrieved, the signal triggering time is compared with the start time of the load change period, the numbers of the signal response sequence lag are screened, and a list of delayed response segment numbers is obtained;

[0029] S402: Based on the delayed response fragment number list, filter the control devices and action coverage time periods corresponding to the numbers, extract cooling load fluctuation change data within the same time period, and obtain the response overlap section index table based on the start and end range of the corresponding data;

[0030] S403: Based on the response overlap section index table, extract the associated control node number, locate the action time and signal position of the number in the original path, filter the node number in the cross response section, and obtain the control node trigger adjustment identification area.

[0031] As a further embodiment of the present invention, the present invention comprises:

[0032] S5: Based on the control node trigger adjustment identification area, identify the operation path segment number, retrieve the corresponding path control node time period, extract the unrestricted path segment to replace the original path segment, update the graph connection relationship, write the control instruction sequence, and obtain the cooling pump optimization control instruction chain;

[0033] The cooling pump optimization control instruction chain includes optimizing path segments, adjusting path structures, and updating control instruction sequences.

[0034] As a further solution of the present invention, the specific steps of S5 are:

[0035] S501: Based on the control node trigger adjustment identification area, identify the path segment number in the current cooling pump operation process, track the time period control node pointed to by each segment number in the graph structure, and associate the time period of each segment with the number information of the control node to obtain a path control time period node mapping table;

[0036] S502: Based on the path control period node mapping table, exclude path segments with restricted numbers, identify operable segment numbers, locate the control nodes connected to each segment, and obtain a replaceable path segment number sequence;

[0037] S503: Based on the replaceable path segment number sequence, the restricted area segment in the original path is compared with the corresponding control node trigger order, the replacement path number is updated, and the trigger information is written according to the node response order to obtain the cooling pump optimization control instruction chain.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are:

[0039] In the present invention, by constructing a behavioral mapping between pump action and signal timing, extracting alternating signals and direction deviation fragments in the control chain, generating path restriction identifiers for constraining control interference, combining the load response with the matching regulation trigger area of ​​the equipment action section, screening the accessible path to replace the restricted section, reorganizing the path structure and sorting the instruction chain, enhancing the matching ability of path scheduling and the response coordination of linkage nodes, and promoting the control process to remain continuous and stable under differentiated states. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the steps of the present invention;

[0041] Figure 2 This is a schematic diagram of the refinement of S1 of the present invention;

[0042] Figure 3 This is a schematic diagram of the refinement of S2 of the present invention;

[0043] Figure 4 This is a schematic diagram of the refinement of S3 of the present invention;

[0044] Figure 5 This is a schematic diagram of the refinement of S4 of the present invention;

[0045] Figure 6 This is a detailed schematic diagram of S5 of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0047] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0048] See also Figure 1 The present invention provides a technical solution: a cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph, comprising the following steps:

[0049] S1: Obtain the start and stop control signal time points of the cooling pump in the operation record of the central air-conditioning chiller unit, call the current startup monitoring data of the corresponding pump number, align the control signal trigger time and the current response time according to the device number, locate the section of each response action according to the chronological order, write the pump number and the corresponding action time in the map device node list, and obtain the pump response behavior mapping set;

[0050] S2: Based on the action time interval in the pump response behavior mapping set, the cold water outlet temperature sensor node and the cooling load feedback node are correspondingly called, the response position of the node in the cooling process is located in chronological order, the control sequence relationship between the nodes is associated with the path node number sequence, and the control path node logic sequence is obtained;

[0051] S3: Based on the device control signals between adjacent nodes in the control path node logic sequence, the start / stop status data of the control nodes in the corresponding time period is called, and the time position and direction change of the control action are compared. The nodes where the signals alternate and the directions between actions are inconsistent are identified, and the direction conflict numbers are added to the graph control index table and the paths where the numbers are located are written into the list to obtain the conflict path restriction identification set;

[0052] S4: Based on the conflict path restriction identification set, the path sequence number is input to retrieve the corresponding cooling load response change data between the control signals, the segment numbers of the signal response sequence delay are screened, the action range of the corresponding control device is located, and the interval relationship between the response overlapping time segments and the load change period is matched to obtain the control node trigger adjustment identification area;

[0053] S5: Based on the control node trigger adjustment identification area, identify the path segment number during the current operation of the cooling pump, retrieve the time period control node connected to the corresponding path in the graph structure, select the accessible segment in the unrestricted path, replace the original path content in the restricted area, and write the corresponding instruction trigger sequence to obtain the cooling pump optimization control instruction chain.

[0054] The pump response behavior mapping set includes pump number, action time, and response segment; the control path node logic sequence includes temperature node position, load node timing, and control path number; the conflict path restriction identification set includes signal direction conflict number, path conflict sequence, and control index label; the control node trigger adjustment identification area includes response delay segment, action coverage range, and load change range; the cooling pump optimization control instruction chain includes optimized path segment, adjusted path structure, and updated control instruction sequence.

[0055] See also Figure 2 , the specific steps of S1 are:

[0056] S101: Obtain the start and stop control signal time points of the cooling pump in the central air conditioning chiller operation log, filter the pump number and issuance time corresponding to the signal field, match the signal time with the pump number in chronological order, and associate the starting control signal with the response signal to obtain a signal time alignment sequence;

[0057] First, read the signal field content line by line for each record in the log, separate the cooling pump number that appears in the record and store it in an array, record the corresponding issuance time of each record in an independent data column, compare multiple records corresponding to the same number in chronological order, mark the earlier signal time as the start, and mark the subsequent signal time as the response. During the execution process, the time data of each record needs to be numerically parsed. For example, when the record number P1 in the log has a pump start signal at 10.3 seconds and a pump stop signal at 18.6 seconds, the time between 10.3 and 18.6 is established during the comparison. Corresponding relationship, when executing, all numbers need to be traversed, and a time array is established for each number respectively. The time sequence relationship is calculated in sequence, and the starting signal time of different numbers is paired with the subsequent response signal time. During the execution process, the data is filtered, and irrelevant records that do not belong to the cooling pump number are deleted. The duplicate numbers in the remaining data are merged and processed. By comparing the record list of each number item by item, the first control signal time of each number is obtained and compared with the subsequent signal time. If it is found that the first time is less than the subsequent time, the two are recorded as a group. For example, during the execution process, number P2 is recorded in the log for the first time The first time P3 was recorded in the log was 12.5 seconds, and the second time was recorded at 20.0 seconds. The time sequence was determined by reading and comparing the time values. The first time P3 was recorded in the log was 8.2 seconds, and the second time was recorded at 9.7 seconds. The corresponding relationship was also established by comparing the values. After all numbers were compared, their time was aligned. During the execution process, the filtering action was further refined. First, all fields in the log were read and the field contents were checked one by one. Only when the cooling pump number was identified in the field, the data was retained. The time values ​​of all retained data were converted to a unified unit for subsequent comparison. If an abnormal value, such as a negative value, appeared in the time field, it was removed. Deduplication was performed on the number array. The operation is then performed, and multiple records of each number are compared with each other in a loop traversal manner to ensure that the time series of each number is arranged in order from the minimum time to the maximum time. During the comparison process, if the difference between two times is less than 0.1 seconds, it is considered a continuous signal, and these two records are marked in the same time period. For the filtering action involved in the execution process, the corresponding example is that the number P4 appears three times in the log, with the time respectively at 5.5 seconds, 5.6 seconds, and 14.2 seconds. When performing the comparison, 5.5 and 5.6 are first associated with the same segment, and then 5.6 and 14.2 are created into a new segment. After executing all the numbering processing, the results are integrated to obtain the signal time alignment sequence.

[0058] S102: Based on the signal time alignment sequence, extract the current startup monitoring data corresponding to the pump number, locate the time period of the current rising boundary, compare the relationship between the control signal time fields, identify the time segments of the control and response actions, screen the time period and number matching results of the response action, and obtain a pump number response time matching list;

[0059] First, read each record in the alignment sequence and obtain the pump number therein. The current monitoring data field corresponding to the pump number is retrieved from the operation record in sequence. Each current monitoring data is scanned in chronological order. The current values ​​in the monitoring data curve are compared point by point. The boundary segment where the current value rises continuously and the span exceeds 1.5 amperes is found. The starting time and end time of the boundary segment are recorded. The starting time is compared with the control signal time in the alignment sequence one by one. If the control signal time is earlier than the current boundary starting point, it is recorded as the corresponding segment started. If the control signal time is later than the current boundary starting point, The record is a hysteresis segment. During the execution process, an example is used to illustrate that the current value of number P1 in the record is 2.0 amps at 8 seconds, rises to 3.7 amps at 8.4 seconds, and rises to 4.1 amps at 8.9 seconds. By comparing point by point, it can be confirmed that 8 seconds to 8.9 seconds is the current rising segment. The signal time of number P1 in the alignment sequence is 7.8 seconds. Since 7.8 is less than 8.0, it is recorded as the starting segment. Taking number P2 as an example, the current is 1.9 amps at 12.3 seconds, 3.2 amps at 12.8 seconds, and 3.6 amps at 13.1 seconds. The same comparison is made. The result shows that 12.3 to 13.1 seconds is the rising section, which is then compared with the signal time of 11.9 seconds for number P2 in the alignment sequence. Since 11.9 is less than 12.3, it is recorded as the starting section. The monitoring data of number P3 is scanned. The current is 1.7 amps at 5.1 seconds, 1.8 amps at 5.4 seconds, and 3.5 amps at 5.9 seconds. It is determined that 5.1 to 5.9 seconds is the rising section. The control signal time of number P3 in the alignment sequence is 5.6 seconds, which is recorded as the lag section because 5.6 is greater than 5.1. During the execution process, such time and value are performed on each number respectively. Corresponding operations are performed, and the response action time periods are filtered one by one according to the number. The starting time and end time of the same number are output in pairs. The numbers that do not meet the rising boundary conditions are deleted and not processed. The current data of number P4 are 2.1 amps, 2.0 amps, and 1.9 amps. After comparison, no rising segment is found and it is not recorded. The current data of number P5 are 1.8 amps, 2.5 amps, and 2.9 amps. After comparison, the rising segment is confirmed to be the corresponding time period and classified according to the signal time. After comparing one by one, the time periods that meet the conditions are paired and sorted according to the number to obtain a pump number response time pairing list.

[0060] S103: Based on the pump number response time pairing list, filter the response time corresponding to each number, arrange them according to the time field, write each set of pump numbers and action times into the map device node field in sequence, and connect them according to the corresponding time sequence of the response action to obtain the pump response behavior mapping set;

[0061] First, read the number field of each record in the pairing list one by one, and extract the time period value corresponding to each number in the record separately. Perform item-by-item inspection on the extracted time periods, eliminate time periods less than 0.1 seconds, and compare the values ​​of the remaining time periods in ascending order. If it is found that time period T1 is greater than time period T2, the two are swapped, and traverse again until all time periods are arranged in ascending order. Perform a paired association action on each number and corresponding time period after arrangement, write the number into the number field of a device node record table, and write the time period into the time field of the record table. When writing, check the existing content in the record table one by one. If the same number is found to be repeated, add a serial number suffix to the subsequent records to ensure distinction. For example, the pairing list record time periods of number P1 are 3.2, 1.1, and 4.8 respectively. After comparison, they are adjusted to 1.1, 3.2, 4.8, and write them into the corresponding columns of the row number P1 in the record table in sequence. The time periods of number P2 are 2.5, 2.3, and 2.9. After comparison, they are adjusted to 2.3, 2.5, and 2.9 and written into the corresponding columns of the row number P2 in the record table. There is only one time period 3.6 for number P3, which is written directly without adjustment. During the execution process, the homing connection between the action time and the number is processed one by one, and a one-to-one mapping action is performed on the number array and the time period array. Check whether there is a time period misalignment in each pair of mappings. If the time period data corresponding to number P4 is 1.7, 1.9, and 2.1 and the record table sequence is 2.1, 1.7, and 1.9, then the exchange is performed to adjust the time period sequence to 1.7, 1.9, and 2.1. Then, number P4 and the adjusted time period sequence are written into the record table in sequence. After completing the sorting of all numbers and action times, the pump response behavior mapping set is obtained.

[0062] See also Figure 3 , the specific steps of S2 are:

[0063] S201: Based on the action time interval in the pump response behavior mapping set, collect the response time data of the cold water outlet temperature sensor within the time period, read the time field and corresponding device number of each data, and write the number and time information into the node queue in sequence to obtain the temperature response node sequence;

[0064] First, read the action start time and action end time for each record in the mapping set one by one, split these time periods into separate time parameters T-start and T-end one by one, retrieve the historical records of the cold water outlet temperature sensor, perform a traversal action on each temperature sensor data, compare the time field t with T-start and T-end one by one, if t is greater than or equal to T-start and t is less than or equal to T-end, then the data is recorded as valid data, if t is less than T-start or t is greater than T-end, then it is skipped. During the execution process, the time period T-start corresponding to P1 is 5.0 and T-end is 7.5. For example, when the temperature sensor records the time at 5.1, 5.3, 6.8, and 7.6, only 5.1, 5.3, and 6.8 meet the time period conditions, and 7.6 is skipped. Continuing with the example of P2 time period T-start is 10.2 and T-end is 12.8, when the temperature sensor records the time at 10.0, 10.5, 11.1, and 12.7, the Among them, 10.5, 11.1, and 12.7 meet the conditions, and 10.0 is skipped. The same operation is performed for the time period P3 with T-start of 3.3 and T-end of 4.1. When the temperature sensor records time at 3.4, 3.6, 3.9, and 4.3, 3.4, 3.6, and 3.9 meet the conditions, and 4.3 is skipped. When executing the action, the time field t of each data that meets the conditions is paired with the corresponding device number field, and the number and t are written into a continuous data queue in order. When writing, it is necessary to check each item to see if there are repeated numbers at the same time. If duplication is found, the subsequent time will be recorded in the time array of the same number. If the time field is found to be empty or has an abnormal value, the data will not be entered. For example, number P1 has the time 5.1, 5.3, and 6.8, number P2 has the time 10.5, 11.1, and 12.7, and number P3 has the time 3.4, 3.6, and 3.9. These data are written in the order of number and time. After all numbers are processed, the temperature response node sequence is obtained.

[0065] S202: Based on the temperature response node sequence, the response time data of the cooling load feedback nodes in the same period are collected, the number of times each group of feedback nodes and the temperature node appear in the same time segment is counted, and the frequency ratio of each group of feedback nodes to the temperature node in the specified period is calculated to obtain the feedback temperature correlation result;

[0066] The formula for calculating the frequency ratio of each feedback node to the temperature node within a specified period is:

[0067] ;

[0068] in, Represents the frequency ratio of the jth feedback node to the ith temperature node in a specified period, represents the number of co-occurrences of the i-th temperature node and the j-th feedback node in the t-th time period, represents the average number of co-occurrences of the ith temperature node with all feedback nodes in the tth time period, represents the response delay time of the jth feedback node in the tth time period, represents the average response delay time of the jth feedback node in T time periods, represents the response time deviation equalization factor of the j-th feedback node, Represents the total number of time periods divided by the specified period;

[0069] Dimensional limitation:

[0070] Represents the frequency ratio of the j-th feedback node to the i-th temperature node within a specified period. After normalization, it is a statistical ratio and dimensionless.

[0071] represents the number of co-occurrences of the i-th temperature node and the j-th feedback node in the t-th time period. It is a dimensionless number.

[0072] Represents the average number of co-occurrences of the i-th temperature node with all feedback nodes in the t-th time period, the number of times counted, dimensionless;

[0073] Represents the response delay time of the jth feedback node in the tth time period, unit: second (s);

[0074] represents the average response delay time of the jth feedback node in T time periods, in seconds;

[0075] Represents the response time deviation equalization factor of the j-th feedback node, which is usually a normalized compensation coefficient and dimensionless;

[0076] Represents the total number of time periods divided by the specified period, count, dimensionless;

[0077] Overall dimensional analysis:

[0078] Represents the molecular part is the difference in degree, dimensionless;

[0079] Represents the denominator It is the accumulated time difference, unit: second (s);

[0080] Therefore, the dimension of the whole fraction is 1 / time, but in patents, it is usually normalized or set as a frequency ratio after a time reference, so It is a dimensionless index;

[0081] During the implementation process, through the linkage recording mechanism of temperature response and feedback control nodes, synchronous data collection was carried out for the selected feedback node group (numbered j) and temperature node (numbered i) in four consecutive time periods, and the following original observation values ​​were obtained:

[0082] The first period: ;

[0083] The second period: ;

[0084] The third period: ;

[0085] The 4th period: .

[0086] The average response delay time is calculated by:

[0087] ;

[0088] The first term in the numerator (the sum of absolute deviations):

[0089] ;

[0090] The second term in the numerator (square difference and square root):

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] Numerator total:

[0096] ;

[0097] Denominator:

[0098] ;

[0099] ;

[0100] ;

[0101] Frequency ratio calculation:

[0102] ;

[0103] Interpretation of results and numerical significance: The actual response intensity of the temperature node in the jth group of feedback nodes within the specified period is 15.624 times / second; this is compared with the preset benchmark value range (the typical control feedback rate setting range is 5–20 times / second);

[0104] Formula innovation description:

[0105] The benefit of the formula is that by aggregating the amplitude deviation and variance of the co-occurrence frequency and introducing the discrete degree of the response delay of the feedback node as a normalization factor, a coupling model is constructed that takes into account both response strength and delay stability. This avoids misjudging situations where occasional high-frequency fluctuations of the feedback node are misjudged as high-efficiency responses, thereby optimizing the priority screening effect of the feedback node.

[0106] S203: Based on the feedback temperature correlation result, obtain the node number data in the control path, read the response time corresponding to each group of numbers, compare the number time with the feedback action time, and put the node numbers into the path node sequence in chronological order to obtain the control path node logical sequence;

[0107] First, read the node number parameter N for each record in the association result one by one, use N as the index to search for all response times T of the corresponding number in the control path data, take out T in turn and write T into a temporary array, perform item-by-item comparison on the array, if it is found that Ti is greater than Tj and i is less than j for two adjacent time values ​​in the array, swap the two positions, and continue the loop until the time values ​​are sorted from small to large. For example, the response time corresponding to number N1 is 7.5, 6.2, and 8.1. After item-by-item comparison, it is adjusted to 6.2, 7.5, and 8.1. The response time corresponding to number N2 is 3.2, 3.6, and 2.9. After item-by-item comparison, it is adjusted to 2.9, 3.2, and 3.6. Number N3 has only 4.7 and does not need to be adjusted. After sorting is completed, each number is associated with the sorted time one by one. During the execution process, each group of numbers is checked. If a number appears multiple times in the feedback temperature association result, its duplicate items are processed separately in chronological order when writing. Number N4 is in the feedback temperature association result. The time of N5 in the feedback temperature correlation result is 9.2, 9.2, and 9.5. After comparison, the first entry 9.2 is retained, the second entry 9.2 is discarded, and 9.5 is retained. During the execution process, all numbers are put into the path node sequence in time sequence. The time corresponding to number N6 is 1.1, 1.3, and 1.8, which are written into the path node sequence in order. The time corresponding to number N7 is 2.5, 2.7, and 3.0, which are written into the path node sequence in order. The time corresponding to number N8 is 4.2 and 4.4, which are written into the path node sequence in order. Finally, all numbers and response times are written in sequence to obtain the control path node logic sequence.

[0108] See also Figure 4 , the specific steps of S3 are:

[0109] S301: Based on the device control signals between adjacent nodes in the control path node logic sequence, the start and stop status data of the adjacent nodes in the control cycle are called, the start and end symbols of each group of control directions are identified, and the direction switching frequency of each pair of adjacent control actions is calculated to obtain a detailed direction switching frequency table;

[0110] The calculation formula for the direction switching frequency of each pair of adjacent control actions is:

[0111] ;

[0112] in, Represents the direction switching frequency value between the xth node and the yth node in the control path node logic sequence, Represents the numerical identifier of the start / stop status data recorded by the xth node during the pth control action. Represents the numerical identifier of the start / stop status data recorded by the yth node during the pth control action. Represents the weight factor coefficient set between the x-th node and the y-th node, represents the total number of control actions counted between the xth node and the yth node, and p represents the serial index under the sum operation used to traverse each action data serial number;

[0113] Dimensional limitation:

[0114] Represents the frequency of direction switching between the xth node and the yth node in the control path node logic sequence. It is a dimensionless statistical indicator that has been processed by data.

[0115] The numerical identifier representing the start / stop status data recorded by the x-th node during the p-th control action, for example, start = 1, stop = 0. It is a state flag and dimensionless.

[0116] Represents the numerical identifier of the start and stop status data recorded by the y-th node during the p-th control action. It is also a state mark and dimensionless.

[0117] Represents the weight factor coefficient set between the xth node and the yth node, which is used to adjust the influence weight during calculation. It is dimensionless.

[0118] Represents the total number of control actions counted between the xth node and the yth node. It is a count value and dimensionless.

[0119] Represents the index number in the summation and operation process, is only used for traversal, has no physical dimension, and is dimensionless;

[0120] parameter Indicates the start and stop status of node x in the pth control action, which is derived from the actual cooling pump operation record. The "start" state is quantized to 1 and the "stop" state is quantized to 0. Non-numeric data are quantified according to the following standard:

[0121] Active = 1, Inactive = 0;

[0122] Taking monitoring as an example, the state sequence of node x in the 1st to 5th actions is: 1, 1, 0, 1, 0;

[0123] parameter Represents the state of node y at the same time sequence. After the same quantization processing, the state sequence is:

[0124] 0, 1, 1, 0, 0;

[0125] parameter Normalized by the average linkage success rate, in the past 30 control records, after node x sent a control signal, node y responded within 3 seconds 27 times;

[0126] but: , guaranteed to be a dimensionless real number, and its reasonable value range is (0,1);

[0127] parameter is the number of action samples between two nodes. In this example, 5 actions are counted. ;

[0128] Calculate the formula based on the above parameter values:

[0129] Step 1: Calculate the sum and average of the differences multiplied by the weights:

[0130] First action: ;

[0131] Second action: ;

[0132] Action 3: ;

[0133] Action 4: ;

[0134] Action 5: ;

[0135] The sum is ;

[0136] Average , it is definitely worth taking: ;

[0137] Step 2: Calculate the mean of the state difference and then take the square root:

[0138] First action: ;

[0139] Second action: ;

[0140] Action 3: ;

[0141] Action 4: ;

[0142] Action 5: ;

[0143] The sum is , the average value is , after square root is ;

[0144] Result calculation:

[0145] ;

[0146] Result interpretation and numerical significance: The switching frequency baseline value is set to 0.75. Based on the average level of normal fluctuation frequency under low interference conditions, a frequency result higher than 0.9546 indicates control direction fluctuation. The current calculation result is 0.9546.

[0147] Formula innovation description:

[0148] The benefit of the formula is that by introducing the success rate of interaction between nodes By embedding the control direction switching statistics as a weight factor, the proportion of more active or critical nodes in the control path in the frequency index is increased, thereby improving the control sensitivity and path determination accuracy. This breaks through the limitations of traditional processing direction differences by state average and enhances the method's ability to distinguish differences in multi-path control schemes.

[0149] S302: Based on the direction switching frequency detailed table, extract the node pair numbers of the direction switching, remove duplicates in the numbers, call the time sequence data in the node sequence, and classify the nodes with direction changes into number sets according to the time sequence to obtain a node set that controls direction changes;

[0150] First, read the node pair number A and number B for each record in the detailed table one by one, write A and B into a temporary number array, perform a duplicate check on the temporary array, if the same number is detected in the array, remove it from the subsequent processing sequence, and only retain the unique number. For example, the number pairs are P1-P2, P2-P3, and P3-P1. After the duplicate check, the number sets are P1, P2, and P3. Number P2 appears twice but only one copy is retained. Number P1 appears twice and only one copy is retained. After performing the deduplication action, call the time sequence data in the node sequence, search for the corresponding time T in the node sequence for each number in turn, and compare T in ascending order. If it is found that T1 is greater than T2 and T1 is written before T2, perform the exchange to adjust the order so that T2 is before T1. Repeat this process until the time sequence is completely arranged. For example The times when number P1 appears in the node sequence are 3.5, 5.2, and 2.8. After item-by-item comparison, the order is adjusted to 2.8, 3.5, and 5.2. The times when number P2 appears in the node sequence are 1.1, 1.4, and 1.3. After item-by-item comparison, the order is adjusted to 1.1, 1.3, and 1.4. The times when number P3 appears in the node sequence are 6.7, 6.1, and 6.4. After item-by-item comparison, the order is adjusted to 6.1, 6.4, and 6.7. The numbers and their corresponding time values ​​are written into the number set in the adjusted order. When writing, check whether there is a number that does not match the time. If so, remove the number and no longer write it. If no corresponding time is found in the time sequence for number P4, remove it from the number set. Finally, all the numbers that have completed the time sequence are arranged in sequence and integrated into the final set to obtain the control direction change node set.

[0151] S303: Based on the control direction change node set, extract the path segment number associated with the number in the graph structure, retrieve the control path number to which the path segment belongs, write the control path number into the path list, and obtain the conflict path restriction identifier set;

[0152] First, read each number in the node set in sequence, use the number as an index to search the associated path segment number in the graph structure one by one, and record the search results one by one in the order of the numbers. For example, when executing, if the path segment numbers associated with number X1 in the graph structure are A12, A15, and A18, then A12, A15, and A18 are written into the segment array in order. If the path segment numbers associated with number X2 are B05 and B07, then B05 and B07 are written into the segment array in order. The associated path rank numbers are C03, C04, and C09. After checking one by one, they are written into the rank array. After completing the collection of the path rank numbers, the rank numbers are compared. The rank number array is traversed in turn, and the control path number corresponding to each rank number is retrieved. The rank numbers are compared with the control path numbers one by one. If it is found that the rank number A12 is subordinate to the control path number K1, K1 is recorded. If the rank number A15 is subordinate to the control path number K2, K2 is recorded. If the rank number A18 is subordinate to the control path number K2, K2 is recorded. If the control path number is K1, record K1. If the rank number B05 is subordinate to the control path number K3, record K3. If the rank number B07 is subordinate to the control path number K3, record K3. If the rank number C03 is subordinate to the control path number K4, record K4. If the rank number C04 is subordinate to the control path number K4, record K4. If the rank number C09 is subordinate to the control path number K5, record K5. During the execution process, the control path numbers are checked for duplication. If repeated control path numbers are detected, the first record is retained and subsequent duplicates are discarded. For example, if the control path number K1 appears twice in the rank numbers A12 and A18, only K1 is retained once. If the control path number K3 appears twice in the rank numbers B05 and B07, only K3 is retained once. The retained control path numbers are then written into the path list in sequence. K1, K2, K3, K4, and K5 are written in sequence according to the record order. After all numbers and rank numbers are processed, a conflict path restriction identifier set is obtained.

[0153] See also Figure 5 , the specific steps of S4 are:

[0154] S401: Based on the conflict path restriction identification set, the path number is input, the time series data of the control signal and the cooling load change are retrieved, the signal trigger time is compared with the start time of the load change period, the numbers of the signal response sequence are filtered out, and a list of delayed response segment numbers is obtained;

[0155] First, read the input path number one by one, use each path number as an index to retrieve the corresponding control signal time series data and cooling load change time series data in the record library, check the retrieved time data one by one, record the control signal triggering time in array S in sequence, and record the cooling load change period start time in array L in sequence, compare S and L item by item, if it is found that Si is less than or equal to Li, it is considered that the triggering sequence of this number is in line with the normal timing, if it is found that Si is greater than Li, it is considered that the triggering sequence of this number is lagging, and the number is recorded in the hysteresis sequence array D. For example, the control signal time of number R1 is 3.1, 6.5, and 9.2, and the corresponding cooling load change start time is 3.0, 6.0, and 9.0, respectively. By comparison, it is found that 3.1>3.0 is determined to be lagging, 6.5>6.0 is determined to be lagging, and 9.2>9.0 is determined to be lagging. Number R1 is written into the hysteresis array, and the control signal time of number R2 is 4.0, 5.2, and 7.0, respectively, corresponding to the cooling load The starting time of the load change is 4.0, 5.5, and 7.2 respectively. By comparison, it is found that 4.0=4.0 is not delayed, 5.2<5.5 is not delayed, 7.0<7.2 is not delayed, and number R2 is not written into the delay array. The control signal time of number R3 is 2.5, 3.3, and 4.1 respectively. The starting time of the corresponding cooling load change is 2.0, 3.0, and 3.8 respectively. By comparison, it is found that 2.5>2.0 is delayed, 3.3>3.0 is delayed, 4.1>3.8 is delayed, and number R 3. Write the lag array. When performing the screening, traverse the arrays S and L one by one and judge according to the same index. If there are multiple lags, record the number each time. If the same number meets the lag conditions at different positions at the same time, the number is recorded multiple times in D. Then, D is checked repeatedly, and the repeated numbers are removed and retained. For example, number R1 has three lags but is only retained once in D, and number R3 has three lags but is only retained once in D. After completing all number processing, a list of delayed response fragment numbers is obtained.

[0156] S402: Based on the delayed response segment number list, the control devices and action coverage time periods corresponding to the numbers are screened, cooling load fluctuation change data within the same time period is extracted, and the start and end ranges of the corresponding data are obtained to obtain a response overlapping section index table;

[0157] Read each number in the number list one by one and search the corresponding control device in the device record, retrieve the start time T-start and end time T-end of the control device action, record T-start and T-end in the action time period array, check each numbered time period, if T-end is less than or equal to T-start, discard the number, if T-end is greater than T-start, retain and proceed to the next step. For example, the action start time corresponding to number M1 is 3.2 seconds and the end time is 5.8 seconds, which meets the conditions and is retained. , the action start time corresponding to number M2 is 7.5 seconds, and the end time is 7.0 seconds. Since the end time is less than the start time, it is discarded. The action start time corresponding to number M3 is 10.0 seconds, and the end time is 12.4 seconds. It meets the conditions and is retained. For each retained number, the cooling load fluctuation record is retrieved in turn. For each data in the record, the fluctuation time t is read one by one and it is determined whether t is between T-start and T-end. If T-start≤t≤T-end is satisfied, the data is marked as valid fluctuation data. If not, it is ignored. For example, when number M1 is The time interval of number M3 is from 10.0 to 12.4 seconds, and the cooling load fluctuation recording time is 9.8 seconds, 10.5 seconds, and 12.3 seconds. By judging that 10.5 and 12.3 meet the conditions but 9.8 does not, the time range interval of each fluctuation data that meets the conditions is recorded, and the start and end time of each numbered fluctuation data are compared one by one. If it is found that the interval between consecutive fluctuation data is less than 0.2 seconds, it is qualified. And it is a segment. If the interval is greater than 0.2 seconds, it will be recorded separately. For example, the effective fluctuation time of No. M1 is 3.3 seconds and 3.5 seconds with an interval of 0.2 seconds, which are merged and recorded as 3.3 to 3.5 seconds. Another fluctuation time of No. M1 is 4.0 seconds and 4.3 seconds with an interval of 0.3 seconds, which are recorded separately as 4.0 to 4.0 seconds and 4.3 to 4.3 seconds. The effective fluctuation time of No. M3 is 10.5 seconds and 10.8 seconds with an interval of 0.3 seconds, which are recorded separately, and 12.3 seconds are recorded separately. The fluctuation time periods and numbers corresponding to all numbers are matched and written into the index table one by one to obtain the corresponding overlapping segment index table.

[0158] S403: Based on the response overlap section index table, extract the associated control node number, locate the action time and signal position of the number in the original path, filter the node numbers in the cross response section, and obtain the control node trigger adjustment identification area;

[0159] First, read the associated control node number for each record in the index table one by one, write the read number into the number array, perform a duplicate check on the array, if it is found that the number appears repeatedly, only retain the number that appears for the first time, retrieve the action time and signal position of the retained number one by one in the original path data, record the action time T and signal position P corresponding to each number in a temporary mapping table, perform a comparison action on the mapping table from small to large according to time T, if it is found that Ti is greater than Tj and i is less than j, swap the positions to ensure the time sequence, for example, number K1 action time 4.2 seconds signal position 15, number K2 action time 3.8 seconds signal position 9, after adjusting the order by comparison, K2 is in front and K1 is in the back, number K3 action time 5.0 seconds signal position 20, the order remains unchanged, after completing the time sequence adjustment, scan the mapping table one by one, check each Whether the action time of each number is within the time range of the cross-response segment, if T is between the start point and the end point of the segment, the number is recorded as the number within the cross-response, if T is earlier than the start point of the segment or later than the end point of the segment, it is skipped and not recorded. For example, the segment time start point is 3.5 seconds and the end point is 4.5 seconds. The time of number K2 is 3.8 seconds and it is retained within the range, the time of number K1 is 4.2 seconds and it is retained within the range, and the time of number K3 is 5.0 seconds and it is discarded because it is not within the range. After completing the time range screening for all numbers, the filtered numbers are written into the result array in sequence, and the numbers in the result array are repeatedly checked to see if there are multiple records of the numbers. If there are duplicates, the subsequent entries with the same number are deleted. For example, if number K2 is recorded twice, only the first record is retained, and number K1 is only recorded once and does not need to be processed. Finally, the processed number array is integrated and output to obtain the control node trigger adjustment identification area.

[0160] See also Figure 6 , the specific steps of S5 are:

[0161] S501: Based on the control node trigger adjustment identification area, identify the path segment number during the current cooling pump operation process, track the time period control node pointed to by each segment number in the graph structure, and associate the time period of each segment with the number information of the control node to obtain a path control time period node mapping table;

[0162] Read each control node number in the identification area one by one, search the read number in the equipment operation record, obtain the actual corresponding path segment number in the current cooling pump operation process, write these path segment numbers into the segment array in sequence, perform verification on the segment array one by one, check whether each segment number exists in the pre-stored map structure, if so, retrieve its corresponding time period control node number, establish a one-to-one correspondence between the path segment number and the retrieved control node number, and for each combination of segment number and control node number, further read the starting point T-start and end point T-end of the time period where the segment number is located. For example, the starting point of the time period corresponding to segment number L1 is 5.0 seconds and the end point is 7.5 seconds, the starting point of the time period corresponding to segment number L2 is 8.1 seconds and the end point is 9.4 seconds, and the starting point of the time period corresponding to segment number L3 is 10.2 seconds and the end point is 12.0 seconds. Associate the time period data with the control node number, and associate L1 with control node C11, L2 It is associated with control node C15, and L3 is associated with control node C18. During the execution process, multiple control node numbers are compared in sequence. If the same segment number is found to appear multiple times in different time periods, each time period is recorded one by one. For example, segment number L4 appears twice in the running record. The first time period starts at 3.2 seconds and ends at 3.9 seconds, and the second time period starts at 4.1 seconds and ends at 4.8 seconds. The corresponding control node numbers are C21 and C22 respectively, and the two mappings are recorded respectively. If it is found during the comparison that the segment number has no corresponding control node number, it is deleted and not entered. For example, when segment number L5 cannot match the control node number, it is directly discarded and the result is not written. During the execution process, the mapping formed by all segment numbers and time period data is checked one by one for duplicate records. If duplicate entries are detected, the records are merged to retain the first entry. After all traversals are completed, the segment number, time period start and end point, and corresponding control node number are written into the result mapping table in sequence to obtain the path control time period node mapping table.

[0163] S502: Based on the path control period node mapping table, exclude the path segments with restricted numbers, identify the operable segment numbers, locate the control nodes connected to each segment, and obtain a sequence of replaceable path segment numbers;

[0164] First, read each path segment number in the mapping table one by one, and compare the read segment numbers with the pre-recorded restricted number list one by one. When performing the comparison, the segment number and each number in the list are judged to be equal one by one. If they are equal, the segment number is determined to be restricted and eliminated. If they are not equal, the segment number is marked as an operable segment number. For example, segment numbers D1, D2, and D3 are compared with the restricted list L2, L4, and L6. D1 is not retained in the list, D2 is eliminated in the list, and D3 is not retained in the list. After the screening is completed, the retained segment numbers are processed one by one, and the segment number is used as an index to retrieve the corresponding control node number in the mapping table. The segment number and the retrieved control node number are paired and recorded. For example, segment number D1 corresponds to control node number C7 in the mapping table, so the association between D1 and C7 is recorded. Segment number D3 corresponds to control node number C12 in the mapping table, so the association between D3 and C12 is recorded. , the rank number D5 corresponds to the control node number C15 in the mapping table, then the association between D5 and C15 is recorded. During the execution process, each record is repeatedly checked. If it is found that the same rank number has been recorded once, the subsequent identical records are discarded. If it is found that the same control node number corresponds to multiple rank numbers, all of them are retained. For example, rank numbers D7 and D8 correspond to the same control node number C20, then D7 and D8 are retained at the same time. Finally, all retained rank numbers are arranged in chronological order. When performing the arrangement, the time period starting value T-start of the rank number in the mapping table is compared one by one. If T-start-i is greater than T-start-j and i is less than j, the positions of the two records are swapped. For example, the starting point of rank number D1 is 2.1 seconds, the starting point of D3 is 3.5 seconds, and the starting point of D5 is 1.8 seconds. After comparison, the order is adjusted to D5, D1, and D3. The result sequence is written one by one in the adjusted order to obtain the replaceable path rank number sequence.

[0165] S503: Based on the sequence of the alternate path segment numbers, the restricted area segments in the original path are compared with the corresponding control node triggering order, the alternate path numbers are updated, and trigger information is written according to the node response order to obtain a cooling pump optimization control instruction chain;

[0166] First, read each rank number in the sequence one by one, and compare the rank number with the control node trigger sequence list one by one. When comparing, start from the first rank number and search the position index of the same number in the trigger sequence list. Record the order of appearance of the number according to the index sequence, and write the number and sequence together into a temporary table. For example, rank number X1 is indexed at position 2 in the trigger sequence list, rank number X2 is indexed at position 4, and rank number X3 is indexed at position 6. After sorting according to the index, the order is X1, X2, and X3. After completing the sequence confirmation, perform the action of comparing the restricted area rank in the original path. For each identified replaceable rank number, check whether the same rank number exists in the restricted area list. If so, perform the replacement action and replace the restricted rank number with the rank number in the current sequence. If not, keep it unchanged. For example, the restricted area ranks are Y1, Y2, and Y3, and the replacement sequence is X1, X2, and X3. After checking, it is found that Y1 matches X1 and is replaced, and Y2 matches X2 Perform replacement, match Y3 with X3 for replacement, and write the replacement result into the updated path segment array. After completing the replacement action, traverse the updated segment array in sequence, and search the control node number corresponding to each segment number in the trigger sequence list again to obtain the response sequence number R of each control node, and record the segment number and R in pairs. For example, segment number X1 corresponds to control node C5, and C5 has a sequence number of 1 in the trigger sequence list, then record X1-C5-1, segment number X2 corresponds to control node C7, and C7 has a sequence number of 3 in the trigger sequence list, then record X2-C7-3, segment number X3 corresponds to control node C9, and C9 has a sequence number of 5 in the trigger sequence list, then record X3-C9-5. During execution, check one by one whether the same segment number is written repeatedly. If there is a duplicate, delete the subsequent duplicate entries until all segment numbers are written according to the trigger sequence. Combine the sorted segment numbers and the corresponding control node sequences into a record table to obtain the cooling pump optimization control instruction chain.

[0167] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph, characterized in that: The following steps are involved: S1: Obtain the start and stop control time points of the cooling pump, call the current startup monitoring data, align the signal time with the pump number, locate the section where each response action is located according to the time sequence, write the pump number and action time into the map node set, and obtain the pump response behavior mapping set; S2: Based on the pump response behavior mapping set, retrieve the action time interval in the pump response behavior mapping set, read the cold water outlet temperature and cooling load feedback node, locate the node response time sequence position, compare the action sequence with the path number relationship, and obtain the control path node logic sequence; S3: Based on the control path node logic sequence, extract adjacent control signals in the control path, read the start / stop status data, determine whether the time position and direction are consistent, identify signal alternation and opposite direction paths, extract the path number and add it to the graph index to obtain the conflict path restriction identification set; S4: Based on the conflict path restriction identification set, the conflict path number is input, the cooling load response data is retrieved, the response delay path segments are screened, the equipment action time period is identified, the overlapping response segments are matched with the load fluctuation segments, and the control node trigger adjustment identification area is obtained; S5: Based on the control node trigger adjustment identification area, identify the running path segment number, retrieve the corresponding path control node time period, extract the unrestricted path segment to replace the original path segment, update the graph connection relationship, write the control instruction timing, and obtain the cooling pump optimization control instruction chain.

2. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1 is characterized in that: The pump response behavior mapping set includes the pump number, action time, and response segment; the control path node logic sequence includes the temperature node position, load node timing, and control path number; the conflict path restriction identification set includes the signal direction conflict number, path conflict sequence, and control index label; the control node trigger adjustment identification area includes the response delay segment, action coverage range, and load change interval.

3. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1 is characterized in that: The cooling load feedback node refers to the node that returns monitoring results to the target control node of the control path during the operation of the central air-conditioning chiller. The node collects real-time changes in the cold water outlet temperature, flow rate, and equipment power, and converts the data into feedback information that can be used for control logic analysis. The response timing position refers to analyzing the response behavior of the control node and mapping the time point of each response action to the time axis.

4. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1 is characterized in that: The graph index refers to locating the control path and node relationship through the index number in the knowledge graph structure, and corresponding the path segment number and node number through the index content; The cooling load response refers to calling the cold water outlet temperature and pump load fluctuation response data during the control behavior triggering process of the central air-conditioning chiller, and combining the time axis of the control signal with the control signal time and load behavior.

5. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1 is characterized in that: The specific steps of S1 are: S101: Obtain the start and stop control signal time points of the cooling pump in the central air conditioning chiller operation log, filter the pump number and issuance time corresponding to the signal field, match the signal time with the pump number in chronological order, and associate the starting control signal with the response signal to obtain a signal time alignment sequence; S102: Based on the signal time alignment sequence, extract the current startup monitoring data corresponding to the pump number, locate the time period of the current rising boundary, compare the relationship between the control signal time fields, identify the time segments of the control and response actions, screen the time period and number matching results of the response action, and obtain a pump number response time matching list; S103: Based on the pump number response time pairing list, filter the response time corresponding to each number, arrange according to the time field, write each group of pump numbers and action time into the map device node field in sequence, and connect them according to the corresponding timing of the response action to obtain the pump response behavior mapping set.

6. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1 is characterized in that: The specific steps of S2 are: S201: Based on the action time interval in the pump response behavior mapping set, collect response time data of the cold water outlet temperature sensor within the time period, read the time field and corresponding device number of each data, write the number and time information into the node queue in sequence, and obtain a temperature response node sequence; S202: Based on the temperature response node sequence, collecting response time data of cooling load feedback nodes in the same period, counting the number of times each group of feedback nodes and temperature nodes appear in the same time segment, calculating the frequency ratio of each group of feedback nodes to the temperature nodes in a specified period, and obtaining a feedback temperature correlation result; S203: Based on the feedback temperature association result, obtain the node number data in the control path, read the response time corresponding to each group of numbers, compare the number time with the feedback action time, and put the node numbers into the path node sequence in chronological order to obtain the control path node logical sequence.

7. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1 is characterized in that: The specific steps of S3 are: S301: Based on the device control signals between adjacent nodes in the control path node logic sequence, call the start and stop status data of the adjacent nodes in the control cycle, identify the start and end symbols of each group of control directions, calculate the direction switching frequency of each pair of adjacent control actions, and obtain a direction switching frequency detailed table; S302: Based on the direction switching frequency detailed table, extract the node pair numbers of the direction switching, remove duplicates in the numbers, call the time sequence data in the node sequence, and classify the nodes with direction changes into number sets according to the time sequence to obtain a node set that controls direction changes; S303: Based on the control direction change node set, extract the path segment number associated with the number in the graph structure, retrieve the control path number to which the path segment belongs, write the control path number into the path list, and obtain the conflict path restriction identifier set.

8. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1 is characterized in that: The specific steps of S4 are: S401: Based on the conflict path restriction identifier set, the path number is input, the time series data of the control signal and the cooling load change is retrieved, the signal triggering time is compared with the start time of the load change period, the numbers of the signal response sequence lag are screened, and a list of delayed response segment numbers is obtained; S402: Based on the delayed response fragment number list, filter the control devices and action coverage time periods corresponding to the numbers, extract cooling load fluctuation change data within the same time period, and obtain the response overlap section index table based on the start and end range of the corresponding data; S403: Based on the response overlap section index table, extract the associated control node number, locate the action time and signal position of the number in the original path, filter the node number in the cross response section, and obtain the control node trigger adjustment identification area.

9. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1, characterized in that: The cooling pump optimization control instruction chain includes optimizing path segments, adjusting path structures, and updating control instruction sequences.

10. The cooling pump optimization method for a central air-conditioning chiller based on a knowledge graph according to claim 1, characterized in that: The specific steps of S5 are: S501: Based on the control node trigger adjustment identification area, identify the path segment number in the current cooling pump operation process, track the time period control node pointed to by each segment number in the graph structure, and associate the time period of each segment with the number information of the control node to obtain a path control time period node mapping table; S502: Based on the path control period node mapping table, exclude path segments with restricted numbers, identify operable segment numbers, locate the control nodes connected to each segment, and obtain a replaceable path segment number sequence; S503: Based on the replaceable path segment number sequence, the restricted area segment in the original path is compared with the corresponding control node trigger order, the replacement path number is updated, and the trigger information is written according to the node response order to obtain the cooling pump optimization control instruction chain.

Images