Water-cooling air conditioning system control method suitable for overtime work

Abstract

The invention relates to the technical field of intelligent refrigeration control, in particular to a water-cooling air conditioning system control method suitable for overtime, which comprises the following steps: collecting a static pressure value through a differential pressure sensor to calculate differential pressure offset, performing moving average noise reduction and fitting a trend curve, performing multi-window analysis to extract a differential pressure change rate, and generating a flow guide instruction when temperature difference deviates; the valve is adjusted to the target opening degree, a completion signal is generated, a failure time window is generated by grey prediction temperature change, and a compensation signal is generated by detecting the opening degree of the flow matching guide valve. According to the method, the trend is extracted through static pressure and differential pressure analysis, moving average and time sequence fitting are combined, the hidden danger of flow unbalance is dynamically recognized, relay valve step length flow guiding is adjusted, non-target branch expansion is restrained, residual cold failure is predicted through a gray model, a backflow path and response efficiency are optimized, and accurate adjustment and temperature compensation of the guide valve are achieved; a cooling capacity regulation and control closed loop is constructed, and the energy-saving performance and stability of the air conditioning system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent refrigeration control, and particularly to a control method for a water-cooled air-conditioning system applicable to overtime work. Background Art

[0002] The technical field of intelligent refrigeration control includes the dynamic regulation and optimization control of the operating state of the refrigeration system, energy consumption management, environmental perception regulation, system linkage management, etc. The core of this technical field lies in realizing the efficient operation of refrigeration equipment in different usage scenarios through means such as sensor data acquisition, control strategy setting, and actuator response mechanism. This field covers the coordinated control of hardware and software, the formulation of temperature and humidity regulation strategies, energy consumption management during time periods, load prediction and equipment operation cycle matching, etc., and is commonly used for the intelligent operation scheduling of air-conditioning systems in office buildings, industrial factories, and public places and other areas.

[0003] Among them, the control method for a water-cooled air-conditioning system applicable to overtime work refers to a control strategy for the operation of a water-cooled air-conditioning system proposed for the problem of fluctuating refrigeration demand in office buildings during non-standard working hours. The patent theme mainly involves collecting regional personnel activity data and indoor temperature change data according to the actual usage situation of the office area, identifying the air-conditioning load demand during overtime periods, and judging the starting range of refrigeration equipment based on time period settings and personnel distribution. It controls the output capacity of the water-cooled air-conditioning system in an energy-saving operation mode by dynamically adjusting the running time of the chilled water pump, controlling the rotational speed of the cooling tower fan by frequency conversion, and preferentially adjusting the operating parameters of the high-efficiency refrigeration unit, so as to achieve targeted scheduling control for different usage areas in the overtime state.

[0004] The existing control modes in the overtime scenario mostly rely on preset time periods and fixed area settings, and it is difficult to identify the actual usage state and cold load fluctuations in real time, resulting in a lag in control response. When sudden changes occur in the load at the end of the branch, the system often cannot quickly identify the temperature change trend and flow offset characteristics, causing overcooling or uneven heating and cooling in some areas, affecting the comfort experience. Since most existing systems use a single temperature and humidity parameter as the judgment basis and do not fully utilize multi-dimensional data such as pressure, flow, and personnel distribution, the control decisions lack pertinence and foresight. At the same time, valve regulation is often executed in a static setting manner, ignoring the requirements of the opening adjustment rhythm for dynamic load adjustment, resulting in a slow response during the adjustment process and a large execution error. Especially in the overtime scenario of office buildings, problems such as cold energy waste in non-target areas and insufficient cold energy supply in target areas are likely to occur. For example, when employees work overtime on some floors, the system still supplies cooling in the full-floor mode, which not only increases the energy burden but also affects the local load balance of the system, restricting the refined execution effect of the refrigeration control strategy. Summary of the Invention

[0005] In order to solve the technical problems that in the prior art, the regulation mode in the overtime scenario mostly relies on preset time periods and fixed area settings, it is difficult to identify the actual usage status and cold load fluctuations in real time, resulting in a lag in regulation response. When there is a sudden change in the load at the end of a branch, the system often cannot quickly identify the temperature change trend and flow offset characteristics, causing overcooling or uneven heating and cooling in some areas, affecting the comfort experience. Since most existing systems use a single temperature and humidity parameter as the judgment basis and do not fully utilize multi-dimensional data such as pressure, flow rate, and personnel distribution, the control decision lacks pertinence and foresight. At the same time, valve regulation is often executed in a static setting manner, ignoring the requirements of the opening adjustment rhythm for dynamic load adjustment, resulting in a slow response and large execution errors during the adjustment process. Especially in the overtime scenario of office buildings, problems such as cold energy waste in non-target areas and insufficient cold energy supply in target areas are likely to occur. For example, when employees work overtime on some floors, the system still supplies cooling in the full-floor mode, which not only increases the energy burden but also affects the local load balance of the system, restricting the refined execution effect of the refrigeration control strategy. An embodiment of the present invention provides a control method for a water-cooled air-conditioning system applicable to overtime. The technical solution is as follows: On the one hand, a control method for a water-cooled air-conditioning system applicable to overtime is provided, and the method includes: S1: Collect the static pressure value at the end of the branch through a differential pressure sensor, calculate the static pressure offset between adjacent branches, use a moving average filtering algorithm to denoise the offset, perform trend fitting according to the time series, and generate a differential pressure trend curve; S2: Perform multi-window dynamic gradient analysis on the differential pressure trend curve, extract the normalized differential pressure change rate, and generate a diversion start instruction when the steady-state pressure decays in consecutive windows and the temperature of the target branch deviates from the thermal equilibrium reference value; S3: According to the diversion start instruction, identify the opening value of the relay valve of the target branch. If it does not reach the fully open threshold, increase it step by step to the fully open state. If the opening of the non-target branch exceeds the locking threshold, decrease it step by step to the locked state, and generate a valve adjustment completion signal; S4: Based on the valve adjustment completion signal, monitor the temperature at the end of the branch, collect the temperature change amount and input it into a gray prediction model, output the time point when the temperature reaches the high temperature threshold, and generate a remaining cold failure time window; S5: Based on the remaining cold failure time window, detect the branch flow rate, determine the area in combination with the personnel concentration strategy, select the branches that meet the bearing conditions to construct a priority return path, calculate the opening of the guide valve through the flow offset, and generate a remaining cold compensation execution instruction.

[0006] As a further solution of the present invention, the multi-window dynamic gradient analysis calculates the thermal response characteristics of the branch by analyzing the heat capacity, flow rate, and temperature parameters of the branch, and sets a threshold for the dynamic gradient; The fully open threshold is the upper limit of the controllable adjustment range of the valve, and the locking threshold is the lower limit of the controllable adjustment range of the valve. It is set according to the adjustable flow rate of the valve, and the step size is set based on the degree of opening change within the adjustment period. The high-temperature threshold is dynamically adjusted based on air-conditioning load parameter types such as instantaneous load, original load trend, regional cooling and heating demand ratio, external environmental parameters, etc. and their weight coefficients. The grey prediction model uses a single-variable grey prediction model. The sampling period is set according to the thermal response characteristics of the monitored object, and the data length covers a complete thermal dynamic stage. The bearing condition is that the current flow rate is lower than the designed bearing capacity. The differential pressure trend curve includes the differential pressure change amplitude, change rate characteristics, and trend stability index. The diversion start command includes the diversion trigger condition type, pressure decay judgment logic, and temperature difference response level. The valve adjustment completion signal includes the relay valve response status, valve action period, and adjustment status identifier. The remaining cooling failure time window includes the temperature control response time limit, prediction confidence interval, and heat capacity correction coefficient. The remaining cooling compensation execution signal includes the return path priority, flow distribution ratio, and guiding strategy parameters.

[0007] As a further solution of the present invention, the specific steps of S1 include: S101: Collect the static pressure value of the branch end through a differential pressure sensor, calculate the absolute difference between the static pressure values of adjacent branches, sort the difference sequence according to the time stamp and store it to generate a branch static pressure offset. S102: Invoke the branch static pressure offset, apply the moving average filtering algorithm, and perform a convolution operation on the sequence data based on the adaptive time window of the branch pressure fluctuation characteristics to generate a filtered offset sequence. The moving average filtering algorithm is an adaptive sliding filtering algorithm combined with the branch topology structure, and the window length is set according to the static pressure sampling frequency and noise characteristics. S103: Based on the filtered offset sequence, establish a cubic polynomial function relationship, calculate the trend component value corresponding to the time node through the least square method, and connect the component coordinate points in the continuous time interval to generate a differential pressure trend curve.

[0008] As a further solution of the present invention, the specific steps of S2 include: S201: Based on the differential pressure trend curve, construct multiple groups of time sliding windows according to the preset time scale, calculate the differential pressure difference between the start and end times of the window, calculate the change rate based on the difference and time span, record the corresponding change rate and change direction interval, and generate a normalized differential pressure change rate sequence. S202: Call the normalized differential pressure change rate sequence, identify the continuously decreasing section, extract the temperature data of the section, compare each point of the temperature sequence with the thermal balance reference value, screen out the temperature points whose offset exceeds the reference interval, set a dynamic temperature offset threshold in combination with the branch heat capacity parameter, screen out the offset abnormal points, and generate a steady-state offset interval identification value; The thermal balance reference value is set according to the interval formed by expanding the standard deviation by k times above and below the average value; S203: Based on the steady-state offset interval identification value, track the corresponding window index, screen out the overlapping index set that meets the conditions of continuous differential pressure decrease and temperature offset, summarize it in the order of the index to form an instruction list, and output the diversion start instruction; The condition for the continuous decrease of the differential pressure is that the least squares method is used to linearly fit the differential pressure value within the sliding window length, and the slope of the fitted straight line is negative.

[0009] As a further solution of the present invention, the differential pressure change rate is calculated using the formula: ; Wherein, represents the differential pressure change rate within the th time sliding window, represents the differential pressure difference between the differential pressure value at the th moment and the differential pressure value at the start moment of the window within the th time sliding window, with the unit of Pa, represents the arithmetic mean of the differential pressure differences within the th time sliding window, with the unit of Pa, represents the weighting coefficient corresponding to the th moment within the th time sliding window, represents the time span length of the th time sliding window, with the unit of s, represents the time interval between the th moment and the start moment of the window within the th time sliding window, with the unit of s, represents the number of time points included in the th time sliding window.

[0010] As a further solution of the present invention, the specific steps of S3 include: S301: Based on the diversion start instruction, obtain the current opening value of the relay valve of the target branch, compare the current opening value with the set full-open threshold, if the current opening value is less than the full-open threshold, then call the current opening value and the set step size for cumulative operation, update the real-time opening value of the relay valve of the target branch, and generate the opening change rate of the target branch; S302: Detect the current opening value of the relay valve of the non-target branch according to the opening change rate of the target branch, compare the detected opening value with the locking threshold. If the opening value is higher than the locking threshold, perform a decreasing calculation on the current opening value and the set step size, update the opening state of the relay valve of the non-target branch, and generate the opening change rate of the non-target branch. S303: Call the opening change rate of the non-target branch and the opening change rate of the target branch, determine whether the opening value of the target branch has reached the fully open threshold, and determine whether the opening value of the non-target branch is lower than the locking threshold. If both conditions are met, generate a valve adjustment completion signal. The step size is calculated as the maximum valve opening × response coefficient / response delay time. The response coefficient is an adjustment sensitivity parameter determined by experience, with a range of 0.1 to 0.5, and the response delay time is in seconds.

[0011] As a further solution of the present invention, the opening change rate of the non-target branch is calculated using the formula: ; Where represents the opening change rate of the th non-target branch in the th adjustment, in % / step, represents the current opening value of the relay valve of the th non-target branch in the th adjustment, in percentage, represents the set step size of the relay valve of the th non-target branch, in percentage, represents the adjustment weight coefficient of the th associated valve in the th non-target branch on the adjustment of this branch, obtained by fitting fluid simulation and experimental data based on the physical coupling relationship of the valves, represents the real-time opening value of the th associated valve in the th non-target branch, in percentage, represents the locking threshold of the th non-target branch, in percentage, represents the average opening value of the associated valves of the th non-target branch, in percentage, represents the number of associated valves in the non-target branch, represents the current adjustment step number, in step, represents the number of the non-target branch, represents the branch corresponding valve response delay correction factor, in 1 / s, represents the branch The valve response delay time, in seconds.

[0012] As a further solution of the present invention, the specific steps of S4 include: S401: Monitor the temperature at the end of the signal monitoring branch based on the valve adjustment completion signal, extract the continuous monitoring time and the corresponding temperature data, calculate the temperature difference between adjacent moments to construct a change amount sequence, and obtain the temperature change amount time series value; S402: Call the temperature change amount time series value and input it into the grey prediction model, generate an accumulated sequence and perform mean smoothing processing, identify the time node when the predicted sequence first exceeds the high temperature threshold, and obtain the predicted over-temperature time point value; S403: Calculate the time offset according to the predicted over-temperature time point value and the monitoring timestamp, set the interval extension width and then combine to form a complete time period integration value, and obtain the remaining cold failure time window.

[0013] As a further solution of the present invention, the specific steps of S5 include: S501: Detect the branch flow based on the remaining cold failure time window, extract the branch flow change value per unit time, determine the branch carrying capacity and screen the branches with fluctuations within the controllable range, introduce the working position distribution of overtime workers and the expected information of regional concentration, and combine with the branch flow change characteristics to identify the branches with long-term stability and adjacent to the concentration area, and establish a stable interval of flow change; S502: According to the branch numbers in the stable interval of flow change, call the branch position data and the flow load label, judge the path connectivity and screen the branches that meet the carrying range, combine with the overtime worker concentration strategy, calculate the geographical coupling degree between the available branches and the preset concentration area, and screen the optimal path group with the cold quantity response timeliness as the weight, and generate the available path carrying ratio distribution value; S503: Based on the available path carrying ratio distribution value and the matching flow offset percentage, calculate the required opening of the pilot valve and adjust the adjustment coefficient, consider the real-time cold load demand of the concentration area and the dynamic change trend during overtime, and real-time correct the opening of the pilot valve and the adjustment priority, and generate the remaining cold compensation execution instruction.

[0014] As a further solution of the present invention, when calculating the path carrying ratio distribution value, the formula is used: ; Where represents the available path carrying ratio distribution value of branch and path , which is used as a quantitative index for determining the branch return priority, and adjusts the opening of the pilot valve for dynamic flow guidance. represents branch The average flow rate change value per unit time within the current cycle, with the unit of m 3 / s, which is processed into a dimensionless parameter through Z-score normalization, representing the path the current cycle average flow load adjustment coefficient corresponding to the branch group, representing the branch and the path the connectivity judgment parameter between them. If connected, it takes 1; if not connected, it takes 0, representing the branch and the path the original average load flow value, with the unit of m 3 / s, which is processed into a dimensionless parameter through Z-score normalization, representing the total flow from the branch to the path within the current cycle, with the unit of m 3 , which is processed into a dimensionless parameter through Z-score normalization, representing the branch corresponding to the path the configured maximum number of branches, representing the path the standard deviation of the load flow of the branch under it within the current cycle, with the unit of m 3 / s.

[0015] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include: Through the real-time acquisition of the static pressure at the end of the branch and the dynamic analysis of the pressure difference between adjacent branches, based on the introduction of moving average filtering and time series trend fitting, it is possible to accurately master the trend of pressure changes. By extracting the pressure difference change rate through the gradient dynamic analysis method of multiple windows and combining the heat balance reference offset situation, it is determined whether there is a hidden danger of flow imbalance in the branch, making the diversion action have a clear trigger condition. Further, by adjusting the opening degree of the relay valve, while realizing the diversion of the target branch by step control, the unnecessary opening expansion of non-target branches is suppressed, and the spatial focusing of heat load adjustment is completed. In the detection of dynamic temperature changes, combined with the grey prediction model, the time point of residual cold failure is predicted in advance, which can provide a pre-judgment window for cold quantity scheduling to ensure the timeliness of cold quantity compensation. Through the dynamic detection of branch flow and the optimization of the bearing path, the return path has matching degree and response efficiency, and the precise closed-loop adjustment of the guiding valve control is realized under the flow deviation. The above process realizes a full-process closed loop in aspects such as pressure difference trend extraction, diversion timing judgment, valve linkage control, temperature prediction compensation, and return path optimization, strengthening the timeliness, spatial targeting, and execution accuracy of cold quantity allocation, and improving the energy-saving control ability and operation stability of the air-conditioning system under different usage scenarios. Description of the Drawings

[0016] Figure 1 This is a schematic diagram of the working process of the present invention. Specific embodiments

[0017] The following will describe the technical solutions in the present invention with reference to the accompanying drawings.

[0018] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two can be selected.

[0019] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0020] In the embodiments of the present invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0021] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0022] Please refer to Figure 1 , the embodiments of the present invention provide a control method for a water-cooled air-conditioning system applicable to overtime work. The processing flow of this method can include the following steps: S1: Collect the static pressure value at the end of the branch through a differential pressure sensor, calculate the static pressure offset between adjacent branches, use a moving average filtering algorithm to denoise the offset, perform trend fitting according to the time series, and generate a differential pressure trend curve; S2: Perform multi-window dynamic gradient analysis on the differential pressure trend curve, extract the normalized differential pressure change rate, and generate a diversion start instruction when the steady-state pressure decays in consecutive windows and the temperature of the target branch deviates from the thermal equilibrium reference value; S3: According to the diversion start instruction, identify the opening value of the relay valve in the target branch. If it does not reach the fully open threshold, increase it step by step to full open. If the opening of the non-target branch exceeds the locking threshold, decrease it step by step to locking, and generate a valve adjustment completion signal; S4: Based on the valve adjustment completion signal, monitor the temperature at the end of the branch, collect the temperature change amount and input it into the grey prediction model, output the time point when the temperature reaches the high temperature threshold, and generate the cold residual failure time window; S5: Detect the branch flow based on the cold residual failure time window, determine the area in combination with the personnel concentration strategy, select the branch that meets the bearing conditions to construct the priority return path, calculate the guiding valve opening through the flow offset, and generate the cold residual compensation execution instruction.

[0023] Specifically, the steps of S1 are as follows: S101: Collect the static pressure value of the branch at the end of the branch through the differential pressure sensor, calculate the absolute difference between the static pressure values of adjacent branches, sort and store the difference sequence according to the timestamp, and generate the branch static pressure offset; The process of collecting the static pressure value at the end of a branch through a differential pressure sensor needs to be configured in combination with the branch distribution in a specific building water supply system. First, select the position of the end monitoring point in the water supply branch network, and install digital differential pressure sensors at multiple positions. For example, arrange pressure monitoring points at the ends of the water supply branches on six floors of a building, numbered P1 to P6 respectively. The sampling frequency is set to 1 time per minute. At each time point, obtain the static pressure values P1(t) to P6(t) at the six branch ends respectively, and take them as a set of static pressure data to complete a complete data collection at time t. Subsequently, for two adjacent branches, such as P1 and P2, P2 and P3, and so on, perform the calculation operation of the static pressure difference between adjacent branches, that is, take the absolute value of the difference between the two. For example, taking the absolute value of P2(t) - P1(t) is ΔP1(t). In the same way, ΔP2(t), ΔP3(t)... ΔP5(t) can be calculated. The above differences form a set of difference sequences {ΔPi(t)}. This difference sequence will form key-value pairs with the time stamp t and be stored in the database in ascending order of sampling time. In the case of sampling once per minute, 60 sets of difference sequences will be formed within 1 hour, and 1440 sets of difference data will be recorded for 24-hour periodic sampling, and they will be stored in the "branch pressure difference database" through unified numbering and corresponding time stamps. In an actual engineering project, taking the water supply system on the 6th floor of an office building as an example, assume that the data collected at a certain time point t = 10:00 is P1 = 0.42 MPa, P2 = 0.44 MPa, P3 = 0.45 MPa, P4 = 0.43 MPa, P5 = 0.41 MPa, P6 = 0.39 MPa. Then the pressure differences between adjacent branches are ΔP1 = 0.02 MPa, ΔP2 = 0.01 MPa, ΔP3 = 0.02 MPa, ΔP4 = 0.02 MPa, ΔP5 = 0.02 MPa respectively, and they are recorded and stored in the database. The specific execution of the above "calculation" action is: subtract P(i)(t) from P(i + 1)(t) and take its absolute value. The process cannot be simplified to a function call but must perform specific numerical operations. For example, if P(i + 1) = 0.44 and P(i) = 0.42, the calculation formula is |0.44 - 0.42| = 0.02. When "judging" whether there is a branch with a significantly offset pressure difference, it is necessary to set a static pressure difference reference threshold ΔP th = 0.03 MPa. If ΔP(i)(t) ≥ ΔP th then it is determined that the branch has an abnormal offset; the setting of this threshold refers to the daily average pressure difference change range under the normal operation state of the branch, and the maximum daily fluctuation value is 0.027 MPa obtained by statistical analysis of the monitoring results in 6 months. To ensure the analysis sensitivity, set ΔP th to 0.03 MPa, and if it exceeds, it will trigger the offset judgment; in the above sampling data, ΔP(i)(t) < ΔP th, so no abnormal offset was found at this moment; Table 1 shows the static pressure and pressure difference of the branch under different time nodes on a certain day.

[0024] Table 1: Monitoring Table of Branch Static Pressure and Pressure Difference (Unit: MPa)

[0025] As shown in Table 1, the value of ΔP fluctuates at different times, but before it exceeds 0.03 MPa, it is regarded as stable operation; if the frequency of the value of ΔP3 or ΔP4 reaching or exceeding 0.03 MPa exceeds 5 times / hour, it will be regarded that a certain branch has an offset trend. By sorting and recording the time series, a complete offset trend data sequence can be obtained, laying a foundation for subsequent filtering processing.

[0026] S102: Call the offset of the branch static pressure, apply the moving average filtering algorithm, and based on the adaptive time window of the branch pressure fluctuation characteristics, perform convolution operation on the sequence data to generate the filtered offset sequence; The moving average filtering algorithm is an adaptive sliding filtering algorithm combined with the branch topology structure, and the window length is set according to the static pressure sampling frequency and noise characteristics; According to the offset of the branch static pressure, the original data sequence is processed by the moving average filtering algorithm. In actual operation, first, the collected static pressure difference sequence needs to be input into the filtering algorithm, and a suitable sliding window is selected. The length of this sliding window depends on the sampling frequency and noise characteristics of the static pressure data. For example, in practical applications, if the sampling frequency is set to once per minute, the length of the sliding window can be adjusted according to the historical data fluctuations. Assuming the initial setting is 5 minutes, then the filtering window contains 5 sampling points. The goal of the moving average filtering is to reduce the abnormal data fluctuations caused by environmental noise, making the data more stable and facilitating subsequent trend analysis.

[0027] The specific execution process is as follows. Assume that at a certain moment t, the difference sequence from ΔP1(t) to ΔP5(t) is [0.02, 0.01, 0.02, 0.02, 0.02]. First, these data are processed by a sliding window with a window size of 5, and the goal is to average the current data point with the previous 4 data points to obtain a smoothed value. For example, at t = 10:00, the first calculation is ΔP avg =(0.02 + 0.01 + 0.02 + 0.02 + 0.02) / 5 = 0.018 MPa; then the sliding window continues to move forward, and each time the average value of the current position data and the previous 4 data in the window is calculated to obtain the smoothed ΔP sequence. In this process, the length of the sliding window is an important parameter, which not only depends on the sampling frequency but also needs to be adjusted according to the actual pressure fluctuation characteristics. If it is found that the noise is large during data analysis, the window length can be increased, and vice versa.

[0028] After the static pressure difference sequence is smoothed, the adaptive sliding filtering algorithm can be used to further optimize the data. At this time, the filtering window will be adaptively adjusted according to the data fluctuation, so that the window size can change dynamically. For example, when the operation is relatively stable, the window can be reduced to reduce unnecessary calculations; while during the period of large pressure fluctuations, the window will automatically increase to better track the data changes and eliminate the interference of instantaneous outliers on the trend judgment.

[0029] S103: Based on the filtered offset sequence, establish a cubic polynomial function relationship, calculate the trend component value corresponding to the time node by the least square method, connect the component coordinate points in the continuous time interval, and generate a pressure difference trend curve; On the basis of the filtered data, further construct a trend curve to reflect the long-term change trend of the pressure difference. For this purpose, a cubic polynomial function is used to establish the mathematical relationship between the pressure difference and time, and then the trend component corresponding to the time node is calculated by the least square method. The core idea of this method is to fit a smooth curve to describe the main trend changes in the data sequence and exclude the influence of local noise and short-term fluctuations.

[0030] The specific implementation steps are as follows. Assume that the ΔP sequence after moving average filtering is [0.02, 0.01, 0.02, 0.02, 0.02]. First, take time t as the independent variable and ΔP as the dependent variable to construct a polynomial equation. For example, assume that we choose a cubic polynomial to fit the data, and the standard form of the cubic polynomial is: , Next, calculate the coefficients a3, a2, a1, a0 of this equation by the least square method. The specific implementation process of the least square method includes: substituting the time point t and the corresponding ΔP value into the polynomial equation, and minimizing the residuals of the points through an optimization algorithm, that is, adjusting a3, a2, a1, a0 to make the fitting curve closest to the real data. Assume that the ΔP sequence [0.02, 0.01, 0.02, 0.02, 0.02] and the corresponding time points t = [1, 2, 3, 4, 5] are used. We substitute these data into the polynomial equation and use the least square method to solve it to obtain the fitted polynomial function.

[0031] Through the obtained coefficients, the fitted polynomial equation is obtained as follows: ; According to this equation, the ΔP value at a future time point can be further predicted. For example, the predicted ΔP value at t = 6 is: ; Connect the trend component values of the calculated time nodes to generate a complete differential pressure trend curve.

[0032] Specifically, the steps of S2 are as follows: S201: Based on the differential pressure trend curve, construct multiple groups of time sliding windows according to a preset time scale, calculate the differential pressure difference between the start and end times of the window, calculate the change rate based on the difference and the time span, record the corresponding change rate and the change direction interval, and generate a normalized differential pressure change rate sequence; Based on the differential pressure trend curve, first select the historical differential pressure data sequence of a certain heating branch. The data is sampled at a frequency of 1 minute and is denoted as the sequence , within the time scale minutes, construct a sliding window group with a sliding interval of 5 minutes. Each window has a length of 20 minutes. For each sliding window , record that there are sampling points in it, and determine the starting differential pressure value of the window. For any th sampling point within the window, its differential pressure difference . For example: if the starting differential pressure of a certain window is 162.4 kPa and the differential pressure of the 5th point within the window is 161.6 kPa, then the corresponding differential pressure difference is kPa. Then, weight the differences within the entire window. To reduce human intervention, equal weight processing is used. Let the weighting coefficient of each sampling point be , and the average value of the differences within the window is . For example: if the differences within the window are for a total of 20 points, and its average value is assumed to be kPa. Then, calculate the weighted sum of the weighted offset of the weights, , and then find the weighted time interval sum within the window. Let the sampling time interval be 60 seconds, and the time interval between the th point and the starting point. For example: for the 5th point, seconds. Therefore, the weighted sum of a certain window is: ; Calculate the differential pressure change rate using the formula: ; Among them, represents the normalized differential pressure change rate within the th time sliding window, represents the differential pressure difference between the differential pressure value at the th moment and the starting moment of the window within the th time sliding window, with the unit of Pa. represents the arithmetic mean of the differential pressure differences within the th time sliding window, with the unit of Pa. represents the weighted coefficient corresponding to the th moment within the th time sliding window. represents the time span length of the th time sliding window, with the unit of s. represents the time interval between the th moment and the start moment of the window within the th time sliding window, with the unit of s. represents the number of time points included in the th time sliding window.

[0033] Substitute into the calculation: ; The result here is 0, indicating that the differential pressure within this window shows a uniform decrease. If a perturbation is added and set as , then: ; Continue to calculate the sliding window in the above manner , to obtain the original differential pressure change rate sequence . Perform normalization on it. Let the overall maximum value be , and the minimum value be . Then the normalized differential pressure change rate of a certain window is: ; If the normalized result is less than 0.5, it is marked as "decrease", and combined with the start time and end time of this window, record the result to form the content shown in the following table.

[0034] Table 2: Normalized differential pressure change rate sequence table

[0035] As shown in Table 2, the normalized change rates within the differential time windows are all positive but the corresponding original change rates are in the negative direction, so they are uniformly marked as "decrease". The field values in the table are directly calculated from the measured differential pressure and time series. The normalization process uses linear interpolation to make the window change rate within the [0, 1] standard range.

[0036] S202: Call the normalized differential pressure change rate sequence, identify continuous decline sections, extract section temperature data, compare each point of the temperature sequence against the thermal equilibrium reference value, screen out temperature points whose offset exceeds the reference interval, set a dynamic temperature offset threshold in combination with the branch heat capacity parameter, screen out abnormal offset points, and generate a steady-state offset interval identification value; The thermal equilibrium reference value is set according to the interval formed by expanding the standard deviation by k times above and below the average value; After calling the normalized differential pressure change rate sequence [0.11, 0.20, 0.34], sequentially compare the change rate direction values [-1, -1, -1] of adjacent sliding windows to determine whether there is a continuously negative area. Since the directions here are all negative, it is determined that the period from [00:00, 00:30] is a continuously declining section. Extract the temperature sequence [65.2, 64.8, 64.5, 63.9, 63.5, 63.0] within the corresponding time section, and set the thermal equilibrium reference interval accordingly. Calculate the mean value of this sequence , standard deviation . Adopt Set the thermal equilibrium interval as . Compare each point of the temperature sequence to see if it falls within this interval. It is found that all data points are within the interval, and initially it is judged that there is no obvious offset.

[0037] Further set to 1, and adjust the thermal equilibrium interval to . At this time, the temperature values 63.0, 63.5, and 63.9 in the sequence are lower than or close to the lower limit of the interval, and initially screen out offset points. Introduce the heat capacity to judge the actual offset degree. Set the branch heat capacity , temperature offset influence amount threshold , and obtain the temperature offset threshold . Taking as the reference, screen out the temperature points that satisfy , that is, retain the temperature values less than 63.55 or greater than 64.75. Among them, 63.0 and 63.5 meet the conditions, corresponding to the times 00:25 and 00:30 respectively, and are marked as the steady-state offset interval. Finally, the offset interval identification value array formed is [0, 0, 0, 0, 1, 1].

[0038] S203: Based on the steady-state offset interval identification value, trace the corresponding window index, screen out the overlapping index set that satisfies the conditions of continuous decline of differential pressure and temperature offset, summarize it in index order to form an instruction list, and output a diversion start instruction; The condition for continuous decline of differential pressure is to perform linear fitting on the differential pressure values using the least squares method within the sliding window length, and the slope of the fitting line is negative; Based on the steady-state offset interval identification value [0, 0, 0, 0, 1, 1], combined with the normalized differential pressure change direction value [-1, -1, -1], the index positions corresponding to the intersection of the two are extracted as the last two items of the sequence, with indices 4 and 5, corresponding to times [00:25, 00:30]. Therefore, an overlapping section that satisfies the two conditions of "continuous decrease in differential pressure" and "stable temperature offset" is identified, and a diversion start instruction list [4, 5] is summarized and generated. According to the start time of the sliding window pointed to by the index, the trigger time set for the diversion start action is finally formed as {00:25, 00:30}, serving as the execution instruction node for the control system to respond to temperature offset and abnormal differential pressure trend.

[0039] Specifically, the steps of S3 are as follows: S301: Based on the diversion start instruction, obtain the current opening value of the relay valve of the target branch, compare the current opening value with the set full-open threshold. If the current opening value is less than the full-open threshold, call the current opening value and the set step size for cumulative operation, update the real-time opening value of the relay valve of the target branch, and generate the opening change rate of the target branch; After receiving the diversion start instruction, read the current opening value θ = 72% through the digital valve positioner installed on the target branch B1, call the preset full-open threshold θ max = 85%, compare 72% with 85%. When 72% < 85%, calculate the set step size δ = 100% × 0.3 / 5 = 6% (taking the maximum valve opening of 100%, response coefficient β = 0.3, response delay time T = 5 seconds), perform the opening cumulative operation: 72% + 6% = 78%, update the valve opening value to 78%, and calculate the opening change rate Δθ = 6% / step. For example, in the feed system of a chemical reactor, when the main reaction channel needs to increase the flow rate, the control valve is gradually opened in steps of 6%. After 3 adjustments, the opening value reaches 90% (72% + 3 × 6%). At this time, the change rate remains constant at 6% / step until the full-open threshold is reached. If the opening value of 94% exceeds 85% during the 4th adjustment, the cumulative operation is stopped, and the final opening change rate is generated.

[0040] S302: According to the opening change rate of the target branch, detect the opening value of the relay valve of the current non-target branch, compare the detected opening value with the locking threshold. If the opening value is higher than the locking threshold, perform a decreasing calculation on the current opening value and the set step size, update the opening state of the relay valve of the non-target branch, and generate the opening change rate of the non-target branch; According to the opening change rate of the target branch, detect the opening value of the relay valve of the non-target branch. When the current opening value of branch S2 is detected to be 65% ( ), set the locking threshold to 10% ( ), and collect the real-time opening values of the valves V1 and V2 associated with branch S2 through the pressure sensor as 45% and 35% respectively ( , ) Determine the adjustment weight coefficient according to the fluid simulation experimental data , , calculate the weighted sum of associated valves: , obtain the average opening value of branch S2 ( ) = , measure the response delay time through a timer seconds, set the delay correction factor according to the valve specification manual , calculate the denominator term: , finally calculate the opening change rate , compare the calculation result with the locking threshold. When the change rate exceeds the locking threshold, perform a step size decreasing operation. For example, the initial opening of 65% is reduced to 50% after 3 adjustments ( ), generate the opening change rate of the non-target branch, using the formula: ; where, represents the opening change rate of the th non-target branch in the th adjustment, in % / step, represents the current opening value of the relay valve of the th non-target branch in the th adjustment, in percentage, represents the set step size of the relay valve of the th non-target branch, in percentage, represents the adjustment weight coefficient of the th associated valve of the th non-target branch on the regulation of this branch, obtained by fitting fluid simulation and experimental data based on the physical coupling relationship of the valves, represents the real-time opening value of the th associated valve of the th non-target branch, in percentage, represents the locking threshold of the th non-target branch, in percentage, represents the average opening value of the associated valves of the th non-target branch, in percentage, represents the number of associated valves in the non-target branch, represents the current adjustment step number, in step, represents the number of the non-target branch, represents branch corresponding valve response delay correction factor, in 1 / s, represents branch The valve response delay time, in seconds.

[0041] Calculate ; The result shows that each adjustment of the non-target branch S2 needs to reduce the opening by 3.2%. After 3 adjustments, the opening value decreases from 65% to 50% ( ), and the adjustment stops when the detected opening value is lower than the locking threshold of 10%. The formula accurately reflects the multi-valve coupling effect by introducing a weight coefficient and improves the response accuracy by combining a delay correction factor to ensure that the opening synchronization adjustment is completed within a 2-second delay time.

[0042] S303: Call the opening change rate of the target non-branch and the opening change rate of the target branch, determine whether the opening value of the target branch has reached the fully open threshold, and determine whether the opening value of the non-target branch is lower than the locking threshold. If both conditions are met, generate a valve adjustment completion signal; The step size is calculated as the maximum valve opening × response coefficient / response delay time. The response coefficient is an adjustment sensitivity parameter determined empirically, ranging from 0.1 to 0.5, and the response delay time is in seconds; Call the opening change rate of 6% / step of the target branch B1 and the change rate of 3.2% / step of the non-target branch S2, and continuously monitor the current opening value θ of the target branch target = 90%, and compare it numerically with the fully open threshold θ max = 85%. When 90% ≥ 85%, it is determined to meet the standard. At the same time, detect the opening value θ of the non-target branch S2 non-target = 8%, and compare it with the locking threshold θ min = 10%. When 8% ≤ 10%, it is determined to meet the standard. In the heat pipe network balance adjustment, when the main heating branch reaches 90% opening after 5 adjustments (initial 60% + 5 × 6%), the non-target branch drops to 8% after 3 adjustments (initial 65% - 3 × 5% × 3.2%). At the same time, when both θ target ≥ θ max and θ non-target ≤ θ min These two conditions are met, triggering the DCS system to output a 4 - 20mA valve adjustment completion signal.

[0043] Specifically, the steps of S4 are as follows: S401: Monitor the temperature at the end of the branch based on the valve adjustment completion signal, extract the continuous monitoring time and corresponding temperature data, calculate the temperature difference between adjacent moments to construct a change amount sequence, and obtain the temperature change amount time series value; Based on the valve adjustment completion signal, the terminal temperature of the branch is monitored. First, the status feedback of the valve actuator in the temperature control system is received through the signal processing unit. When the feedback signal is marked as the "adjustment completed" state, the terminal temperature of the branch is collected in real time. The temperature data acquisition module sets the sampling period to 60 seconds. During the continuous monitoring process, multiple groups of time and corresponding temperature value combinations are extracted. Each group contains a timestamp and the temperature data at the corresponding time. For example, in the actual overtime operation environment, the monitoring point is located at the end of the air-conditioning branch on the west side of the third floor of the office building. The monitoring period is 5 minutes and a total of 5 data points are collected. The numerical difference between two consecutive temperature values ​​is calculated. The difference processing unit uses the direct difference method to calculate the temperature change at adjacent moments, that is, the temperature value at the next moment is subtracted from the temperature value at the previous moment to form a temperature change sequence. For example, the monitoring data are 27.5℃, 2 8.0℃, 28.6℃, 29.3℃, 30.1℃, and the corresponding changes are 0.5℃, 0.6℃, 0.7℃, and 0.8℃, respectively. Each difference is recorded and combined into a time series array to further extract the temperature change trend. In the above process, in order to ensure that the difference reflects the real change trend, the temperature data of the monitoring point needs to be initially smoothed and the abnormal fluctuation value is eliminated. The abnormal identification condition is the record with a change greater than 2℃ / min or less than -2℃ / min. After identification, the data of this section is eliminated and the difference is recalculated. If the data is normal, the constructed temperature change time series is [0.5, 0.6, 0.7, 0.8], which is used as the basis for temperature dynamic judgment in subsequent steps. In order to show the temperature behavior in the differentiated change stage, the above sequence is stored in the form of a vector, and the data frame structure is constructed and input into the subsequent processing module, as shown in Table 3: Table 3: Branch end temperature change table

[0044] As shown in Table 3, through the five sets of temperature data, it can be seen that their continuous change trend is steadily rising. The change sequence constituted by them is archived into the database one by one, and the sequence is input into the next stage of processing flow as an important basic data for the dynamic evolution of temperature state, and finally the temperature change time series value is obtained.

[0045] S402: calling the temperature change time series value to input into the grey prediction model, generating a cumulative sequence and performing mean smoothing, identifying the time node that exceeds the high temperature threshold for the first time in the prediction sequence, and obtaining the predicted over-temperature time point value; Call the time series value of the temperature change amount and input it into the grey prediction model. First, read the temperature change amount sequence [0.5, 0.6, 0.7, 0.8] constructed in S401 through the data cache module. Then, perform cumulative processing on the sequence in turn. Using the method of sequential summation, the first item is 0.5, the second item is obtained by adding 0.5 and 0.6 to get 1.1, the third item is obtained by adding 0.7 to 1.1 to get 1.8, and the fourth item is obtained by adding 0.8 to get 2.6, thus constructing the cumulative sequence [0.5, 1.1, 1.8, 2.6]. Next, use the three-point moving average method to smooth the cumulative sequence. Calculate the average value of the previous item, the current item, and the next item for the middle data items 1.1 and 1.8 respectively, which are (0.5 + 1.1 + 1.8) / 3 = 1.13 and (1.1 + 1.8 + 2.6) / 3 = 1.83. For the boundary items 0.5 and 2.6, use the two-point average method, that is, (0.5 + 1.1) / 2 = 0.55 and (1.8 + 2.6) / 2 = 2.2. Finally, obtain the smoothed sequence [0.55, 1.13, 1.83, 2.2], and this sequence is used as the basic data for model prediction and input into the next step of deduction; based on this sequence, establish a time axis with a time step of 60 seconds and perform an extended prediction operation. Predict that the cumulative temperature changes at the next three time nodes are 2.9, 3.6, and 4.5 respectively, and the corresponding prediction point times are 300 seconds, 360 seconds, and 420 seconds. On this basis, judge whether to enter the high-temperature warning area. Use the current actual end temperature value of 30.1°C as the reference temperature, and add it to the predicted change values in turn to obtain the predicted temperature values of 33.0°C, 33.7°C, and 34.6°C respectively. The set high-temperature threshold is 30.5°C, and this threshold is set in the following way: in an overtime simulation experiment, the initial indoor temperature is 27.5°C, the temperature rise rate is 0.8°C / min, and after 4 minutes, that is, 240 seconds, the expected temperature is 27.5 + 0.8×4 = 30.7°C. Combining the investigation of the upper tolerance limit of the temperature sensor threshold in the overtime working environment, choosing 30.5°C slightly lower than this point as the risk threshold has a greater safety margin; compare 30.5°C with the predicted temperature results one by one. The first overlimit occurs at the predicted time point of 300 seconds, that is, 33.0°C first exceeds 30.5°C, and thus the predicted over-temperature time point value is identified as 300 seconds; the experimental basis for the thermal dynamic stage in this step is as follows: under the operating conditions of the air conditioning system at the end of the office building, the experiment is designed to simulate the actual overtime environment. After the air conditioning adjustment is completed, the refrigeration port is closed after 60 seconds, and the temperature response process is started to be recorded. The sampling period is 60 seconds, and 8 groups of temperature data are collected within 420 seconds. The temperature curve shows three typical stages. The temperature rise is slow from 0 to 120 seconds, which is the initial stable stage, and the average temperature change amount is less than 0.3°C. From 120 to 300 seconds, it enters the linear temperature rise stage, and the average change amount is 0.6°C, with a maximum of 0.8°C. After 300 seconds, it is the stable over-temperature stage, which is significantly higher than the preset threshold; therefore, the extracted temperature change amount sequence [0.5, 0.6, 0.The range of [7, 0.8] is in the most representative linear temperature rise stage, with thermal response sensitivity, capable of reflecting the true indoor temperature rise trend, and suitable as training data for the prediction model, thus ensuring that the model can accurately identify that the predicted over-temperature time point value is 300 seconds.

[0046] S403: Calculate the time offset according to the predicted over-temperature time point value and the monitoring timestamp, set the interval extension width, and then combine to form the integrated value of the complete time period to obtain the remaining cooling failure time window; Calculate the time offset according to the predicted over-temperature time point value and the monitoring timestamp. Taking the current temperature monitoring end timestamp of 240 seconds as the reference point, obtain the predicted over-temperature time point value of 300 seconds, and calculate the offset Δt = 60 seconds. Superimpose and expand this offset with the threshold trigger expected lag. To construct the actual control window, set the interval extension width. In this example, the window is set to extend 60 seconds forward and 120 seconds backward. Therefore, the integrated value of the complete time period extends from 240 seconds to 420 seconds, and the remaining cooling failure time window [240s, 420s] is integrated in combination with this window time period; In the actual overtime scenario, if the end-air-conditioning branch of this office building fails to achieve the cooling effect during this time window, the compensation mechanism or alarm module will be automatically triggered; The above offset calculation process is as follows: Subtract the last monitoring time point of 240 seconds from the predicted time point value of 300s to get the offset Δt = 60s. This offset is used to evaluate the predicted early response time. To ensure the timely response of the control strategy, a 60s forward compensation period is set to cope with short-cycle errors, and a 120s backward period is set as the heat inertia cooling capacity dissipation window to form the time period [240s, 420s]; Conduct an actual evaluation of the rationality of the time window. If the temperature rise rate is 0.8℃ per minute, the expected temperature rise within 3 minutes in the window is 2.4℃. Calculated from the current temperature of 30.1℃ to a maximum of about 32.5℃, which significantly exceeds the indoor comfort limit, verifying that this window setting has a temperature rise response space. Finally, the constructed remaining cooling failure time window is [240s, 420s].

[0047] Specifically, the steps of S5 are as follows: S501: Detect the branch flow based on the remaining cooling failure time window, extract the branch flow change value per unit time, determine the branch carrying capacity, and screen the branches with fluctuations within the controllable range. Introduce the expected information of the distribution of overtime personnel workstations and regional concentration, combine with the branch flow change characteristics, identify the branches with long-term stability and adjacent to the concentrated area, and establish a stable interval of flow change; After determining the failure of the residual cooling, enter the branch flow stability screening process. Set the starting point of the residual cooling failure detection as 180 seconds after the failure, sample once every 60 seconds, and continuously collect the instantaneous flow data within three time periods. Taking branch numbers 3, 7, and 11 as examples, their sampling values are as follows: for branch 3, they are 0.13, 0.14, and 0.12 m³ / min; for branch 7, they are 0.17, 0.16, and 0.18 m³ / min; for branch 11, they are 0.12, 0.13, and 0.11 m³ / min. First, calculate the flow change sequence of the branch. For example, the changes in branch 3 are +0.01 and -0.02 m³ / min, the changes in branch 7 are -0.01 and +0.02 m³ / min, and the changes in branch 11 are +0.01 and -0.02 m³ / min, and the change values of the branches are obtained in turn. The set judgment criterion is that the absolute values of two adjacent changes are both less than 0.05 m³ / min, which is regarded as controllable fluctuation and enters the candidate branch set. After the initial screening, branches 3, 7, and 11 meet this condition and are listed in the candidate set.

[0048] Subsequently, further determine the flow stability through the standard deviation. Set the upper limit standard deviation to 0.02 m³ / min, and calculate the standard deviation of the branch flow change values. The standard deviation of branch 3 is approximately 0.015 m³ / min, the standard deviation of branch 7 is 0.015 m³ / min, and the standard deviation of branch 11 is 0.015 m³ / min. All three are lower than the set threshold, so they are confirmed as stable branches. Finally, output the stable interval branch number sequence [3, 7, 11], and combine the spatial aggregation characteristics of the overtime area to make an association mark for the proximity of the spatial positions of these branches to the concentrated area, which serves as the input basis for the subsequent path load ratio and pilot valve control module.

[0049] S502: According to the branch numbers in the stable interval of the flow change, call the branch position data and the flow load label, judge the path connectivity and screen the branches that meet the load range, combine the overtime personnel concentration strategy, calculate the geographical coupling degree between the available branches and the preset concentrated area, and screen the optimal path group with the cold quantity response timeliness as the weight to generate the available path load ratio distribution value; According to the branch numbers in the stable interval of flow rate change, first process the branch number sequence [3, 7, 11] in the stable interval, sequentially read the corresponding physical position data and form a coordinate array [(12.3, 45.6), (18.2, 50.4), (21.9, 48.1)]. Use each branch number as a query index to extract its spatial coordinate values in the position database. After the coordinate extraction is completed, record it in the form of a two-dimensional vector to prepare for subsequent spatial connectivity judgment. Subsequently, extract the average flow rate change per unit time of each branch in the past 5 minutes through the flow label interface, which are 0.13, 0.17, and 0.12 m³ / min respectively, corresponding to the numbers 3, 7, and 11 in sequence. Compare its flow rate value with the set maximum carrying flow rate threshold one by one. The threshold for branch 3 is 0.25 m³ / min, for branch 7 is 0.3 m³ / min, and for branch 11 is 0.2 m³ / min. Calculate the remaining carrying value between the current flow rate and the threshold and divide it by its maximum carrying value to obtain the available carrying ratio of the corresponding branch. The calculation is as follows: The remaining carrying of branch 3 is 0.25 - 0.13 = 0.12 m³ / min, and the ratio is 0.12 / 0.25 = 0.48; for branch 7 it is 0.3 - 0.17 = 0.13 m³ / min, and the ratio is 0.13 / 0.3 ≈ 0.43; for branch 11 it is 0.2 - 0.12 = 0.08 m³ / min, and the ratio is 0.08 / 0.2 = 0.4. Set the minimum available carrying ratio threshold to 30%, that is, when the ratio ≥ 0.3, it is determined as an available branch. After sequentially judging that the above branches all meet the available requirements, enter the branch spatial connectivity judgment step. Taking branch 3 as a reference benchmark, calculate its Euclidean spatial distances to branch 7 and branch 11. After sequentially performing the square operation of the vector coordinate difference and summing and then taking the square root, calculate the distance from branch 3 to 7 is 7.61 m, and from branch 3 to 11 is 9.88 m. Set the connectivity judgment threshold to 10 m. If the distance between any pair of branches is not greater than 10 m, it is determined as spatially connected. Branches 3 and branches 7 and 11 both meet this condition, and confirm its spatial connectivity status as "1", and the execution value is δ （3，7） = 1, δ （3，11） = 1. Based on this, generate a connected path branch sequence [3, 7, 11]. At the same time, check the spatial overlap degree within the distribution coordinates of the overtime area of this path group and incorporate it into the coupling degree calculation weight to improve the matching efficiency of cold quantity allocation. Subsequently, calculate the respective values of the path carrying ratio, using the formula: ; Where, represents the available path carrying ratio distribution value of branch and path and is used as a quantitative index for determining the priority of branch backflow, and adjust the opening of the guiding valve for dynamic flow guidance. represents branch The average flow rate change value per unit time within the current cycle, with the unit of m 3 / s, which is processed by Z-score normalization into a dimensionless parameter, representing the path the current cycle average flow load adjustment coefficient corresponding to the branch group, representing the branch and the path the connectivity judgment parameter between them. If connected, take 1; if not connected, take 0, representing the branch and the path the original average load flow value, with the unit of m 3 / s, which is processed by Z-score normalization into a dimensionless parameter, representing the total flow from the branch to the path within the current cycle, with the unit of m 3 , which is processed by Z-score normalization into a dimensionless parameter, representing the branch corresponding to the path the configured maximum number of branches, representing the path the standard deviation of the load flow of the branch under the path 3 / s within the current cycle.

[0050] is the average flow rate change value per unit time of branch i within the current cycle. After normalization, they are 3.61×10 -3 、4.72×10 -3 、3.33×10 -3 , is the current cycle average flow load adjustment coefficient under path j. In this example, it is uniformly taken as 0.9, is 1, is the original average load flow value of the path. After normalization, the values are 0.1, 0.15, 0.08, is the total flow from the branch to the path within the current cycle, which are 0.021, 0.026, 0.019 m³ respectively. After normalization, the values are 0.25, 0.30, 0.22, is the configured number of branches under the path, uniformly taken as 2, is the standard deviation of the load flow within the current cycle, uniformly set as 0.004 m³ / s. After conversion, they are substituted into the formula and calculated as follows: ; ; ; The results show that the path carrying ratio values of branches 3, 7, and 11 are 218.01, 195.16, and 230.27 respectively. An array of path carrying ratios can be generated based on these values for subsequent regulating valve adjustment and control operations.

[0051] S503: Calculate the required opening degree of the regulating valve and adjust the adjustment coefficient based on the available path carrying ratio distribution value and the matching flow offset percentage. Considering the real-time cooling load demand in the concentrated area and the dynamic change trend during overtime periods, revise the opening degree of the regulating valve and the adjustment priority in real time to generate a residual cooling compensation execution instruction.

[0052] After obtaining the path carrying ratio array [218.01, 195.16, 230.27], it is used as a reference index for the available path capacity of branches in the current cycle. At the same time, set the total adjustment amount of the overall residual cooling compensation flow to 0.3 m³ / min and the offset ratio to 20%. Calculate that the absolute flow offset amount that needs to be redistributed is 0.06 m³ / min. Allocate the adjustment flow proportionally according to the path carrying ratio. The adjustment flow obtained by branch 3 is approximately 0.022 m³ / min, branch 7 is approximately 0.020 m³ / min, and branch 11 is approximately 0.018 m³ / min. This allocation result is used to generate the adjustment target for the branch, and the adjustment increment is calculated by combining the existing flow state.

[0053] Next, call the regulating valve control module to convert the adjustment amount of the branch into an increment of the regulating valve opening degree. Set that each gear of the regulating valve corresponds to a flow increase of 0.01 m³ / min. Then, branch 3 needs to be adjusted by approximately 2.2 gears, branch 7 by approximately 2.0 gears, and branch 11 by approximately 1.8 gears. Round the gear values of the regulating valve to 2 gears. Prioritize the allocation of adjustment values to the branches near the areas with high overtime occurrences, and sort the control strategies of the regulating valve accordingly to ensure preferential response to the dynamic demands of the concentrated cooling load area. Finally, form a compensation control instruction set {3: 2, 7: 2, 11: 2}, and send an execution command through the control module to physically adjust the opening degree of the regulating valve of the branch, thereby completing the actual flow redistribution process of the residual cooling compensation in this cycle.

[0054] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A control method for a water-cooled air-conditioning system applicable to overtime work, characterized in that, It includes the following steps: S1: Collect the static pressure value at the end of the branch through a differential pressure sensor, calculate the static pressure offset between adjacent branches, use a moving average filtering algorithm to denoise the offset, perform trend fitting according to the time series, and generate a differential pressure trend curve; S2: Conduct multi-window dynamic gradient analysis on the differential pressure trend curve, extract the normalized differential pressure change rate, and generate a diversion start command when the steady-state pressure decays in consecutive windows and the temperature of the target branch deviates from the thermal equilibrium reference value; S3: According to the diversion start command, identify the opening value of the relay valve in the target branch. If it does not reach the full-open threshold, increase it step by step to full open. If the opening of the non-target branch exceeds the locking threshold, decrease it step by step to locking, and generate a valve adjustment completion signal; S4: Based on the valve adjustment completion signal, monitor the temperature at the end of the branch, collect the temperature change amount and input it into a grey prediction model, output the time point when the temperature reaches the high-temperature threshold, and generate a remaining cooling failure time window; S5: Based on the remaining cooling failure time window, detect the branch flow rate, determine the area in combination with the personnel concentration strategy, select the branch that meets the load-bearing conditions to construct a priority return path, calculate the opening of the guiding valve through the flow offset, and generate a remaining cooling compensation execution command.

2. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 1, characterized in that, The full-open threshold is the upper limit value of the controllable adjustment range of the valve, and the locking threshold is the lower limit value of the controllable adjustment range of the valve. It is set according to the adjustable flow rate of the valve, and the step size is set based on the degree of opening change within the adjustment period; The high-temperature threshold is dynamically adjusted based on factors including instantaneous load, original load trend, regional cooling and heating demand ratio, external environment parameters, air-conditioning load parameter types, and weight coefficients; The grey prediction model uses a single-variable grey prediction model, and the sampling period is set according to the thermal response characteristics of the monitored object, and the data length covers a complete thermal dynamic stage; The load-bearing condition is that the current flow rate is lower than the designed load-bearing capacity; The differential pressure trend curve includes the differential pressure change amplitude, change rate characteristics, and trend stability index. The diversion start command includes the diversion trigger condition type, pressure decay judgment logic, and temperature difference response level. The valve adjustment completion signal includes the relay valve response status, valve action period, and adjustment status identifier. The remaining cooling failure time window includes the temperature control response time limit, prediction confidence interval, and heat capacity correction coefficient. The remaining cooling compensation execution signal includes the return path priority, flow distribution ratio, and guiding strategy parameters.

3. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 1, characterized in that The specific steps of S1 include: S101: Collect the static pressure value at the end of the branch through a differential pressure sensor, calculate the absolute difference between the static pressure values of adjacent branches, sort and store the difference sequence according to the timestamp, and generate a branch static pressure offset; S102: Call the branch static pressure offset, apply a moving average filtering algorithm, and perform convolution operation on the sequence data based on the adaptive time window of the branch pressure fluctuation characteristics to generate a filtered offset sequence; The moving average filtering algorithm is an adaptive sliding filtering algorithm combined with the branch topology structure, and the window length is set according to the static pressure sampling frequency and noise characteristics; S103: Based on the filtered offset sequence, establish a cubic polynomial function relationship, calculate the trend component values corresponding to the time nodes by the least squares method, connect the component coordinate points in the continuous time interval, and generate a differential pressure trend curve.

4. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 3, characterized in that, The specific steps of S2 include: S201: Based on the differential pressure trend curve, construct multiple groups of time sliding windows according to the preset time scale, calculate the differential pressure difference between the start and end moments of the window, calculate the change rate according to the difference and the time span, record the corresponding change rate and the change direction interval, and generate a normalized differential pressure change rate sequence; S202: Call the normalized differential pressure change rate sequence, identify the continuously decreasing section, extract the temperature data of the section, compare each point of the temperature sequence with the thermal balance reference value point by point, screen out the temperature points whose offset exceeds the reference interval, set a dynamic temperature offset threshold in combination with the branch heat capacity parameter, screen out the offset abnormal points, and generate a steady-state offset interval identification value; The thermal balance reference value is set according to the interval formed by expanding the standard deviation by k times above and below the average value; S203: Based on the steady-state offset interval identification value, track the corresponding window index, screen out the overlapping index set that meets the conditions of continuous differential pressure decrease and temperature offset, summarize it in the index order to form an instruction list, and output a diversion start instruction; The condition for the continuous decrease of the differential pressure is that the differential pressure value is linearly fitted by the least squares method within the sliding window length, and the slope of the fitted straight line is negative.

5. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 4, characterized in that, Calculate the differential pressure change rate using the formula: ； Among them, represents the rate of change of differential pressure within the th time sliding window, represents the differential pressure difference between the differential pressure value at the th moment and the differential pressure value at the start moment of the window within the th time sliding window, with the unit of Pa, represents the arithmetic mean of the differential pressure differences within the th time sliding window, with the unit of Pa, represents the weighting coefficient corresponding to the th moment within the th time sliding window, represents the time span length of the th time sliding window, with the unit of s, represents the time interval between the th moment and the start moment of the window within the th time sliding window, with the unit of s, represents the number of time points included within the th time sliding window.

6. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 4, characterized in that, The specific steps of S3 include: S301: Based on the diversion start instruction, obtain the current opening value of the relay valve of the target branch, compare the current opening value with the set fully open threshold. If the current opening value is less than the fully open threshold, then call the current opening value and the set step size for cumulative operation, update the real-time opening value of the relay valve of the target branch, and generate a target branch opening change rate; S302: According to the target branch opening change rate, detect the opening value of the relay valve of the current non-target branch, compare the detected opening value with the locking threshold. If the opening value is higher than the locking threshold, then perform a decreasing calculation on the current opening value and the set step size, update the opening state of the relay valve of the non-target branch, and generate a non-target branch opening change rate; S303: Call the target non-branch opening change rate and the target branch opening change rate, judge whether the opening value of the target branch has reached the fully open threshold, and judge whether the opening value of the non-target branch is lower than the locking threshold. If both conditions are met, generate a valve adjustment completion signal.

7. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 6, wherein, Calculate the non-target branch opening change rate using the formula: ； Among them, represents the opening change rate of the th non-target branch in the th adjustment, with the unit of % / step, represents the current opening value of the relay valve of the th non-target branch in the th adjustment, with the unit of percentage, represents the set step size of the relay valve of the th non-target branch, with the unit of percentage, represents the adjustment weight coefficient of the th associated valve on the th non-target branch affecting the adjustment of this branch, which is obtained by fitting fluid simulation and experimental data based on the physical coupling relationship of the valves, represents the real-time opening value of the th associated valve on the th non-target branch, with the unit of percentage, represents the locking threshold of the th non-target branch, with the unit of percentage, represents the average opening value of the associated valves of the th non-target branch, with the unit of percentage, represents the number of associated valves in the non-target branch, represents the current adjustment step number, with the unit of step, represents the number of the non-target branch, represents the branch corresponding valve response delay correction factor, with the unit of 1 / s, represents the branch valve response delay time, with the unit of s.

8. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 6, characterized in that, The specific steps of S4 include: S401: Based on the valve adjustment completion signal, monitor the temperature at the end of the branch, extract the continuous monitoring time and the corresponding temperature data, calculate the temperature difference between adjacent moments to construct a change amount sequence, and obtain the temperature change amount time series value; S402: Call the temperature change amount time series value and input it into the grey prediction model, generate an accumulated sequence and perform mean smoothing processing, identify the time node when the predicted sequence first exceeds the high temperature threshold, and obtain the predicted over-temperature time point value; S403: Calculate the time offset based on the predicted over-temperature time point value and the monitoring timestamp, set the interval extension width, and then combine them to form a complete time period integration value to obtain the after-cooling failure time window.

9. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 8, characterized in that, The specific steps of S5 include: S501: Detect the branch flow based on the after-cooling failure time window, extract the branch flow change value per unit time, determine the branch carrying capacity, and screen the branches with fluctuations within the controllable range. Introduce the work station distribution of overtime workers and the expected information of regional concentration, and combine the branch flow change characteristics to identify the branches with long-term stability and adjacent to the concentration area, and establish a stable interval of flow change. S502: According to the branch numbers in the stable interval of flow change, call the branch position data and flow load labels, judge the path connectivity, and screen the branches that meet the carrying range. Combine the overtime worker concentration strategy, calculate the geographical coupling degree between the available branches and the preset concentration area, and screen the optimal path group with the cold quantity response time limit as the weight to generate the available path carrying ratio distribution value. S503: Based on the available path carrying ratio distribution value and the matching flow offset percentage, calculate the required opening degree of the pilot valve and adjust the adjustment coefficient. Consider the real-time cold load demand in the concentration area and the dynamic change trend during overtime periods, and real-time correct the opening degree of the pilot valve and the adjustment priority to generate the after-cooling compensation execution instruction.

10. The control method of the water-cooled air-conditioning system applicable to overtime work according to claim 9, characterized in that, Calculate the path carrying ratio distribution value using the formula: ； Among them, represents the branch and the available path carrying ratio distribution value of the path is used as a quantitative index for judging the priority of branch reflux, and the opening of the pilot valve is adjusted to conduct dynamic flow guidance. represents the branch The average flow rate change value per unit time within the current cycle, with the unit of m 3 / s, and is processed into a dimensionless parameter through Z-score standardization. represents the path The current cycle average flow load adjustment coefficient corresponding to the branch group. represents the branch and the path The connectivity judgment parameter between them, taking 1 for connected and 0 for unconnected. represents the branch and the path The original average load flow value, with the unit of m 3 / s, and is processed into a dimensionless parameter through Z-score standardization. represents the total flow of the branch to the path within the current cycle, with the unit of m 3 , and is processed into a dimensionless parameter through Z-score standardization. represents the branch corresponding to the path The configured maximum number of branches. represents the path The load flow standard deviation of the branch under the path within the current cycle, with the unit of m 3 / s.

Images