A multi-channel liquid return temperature difference cross-talk decoupling control method and system

Abstract

The application relates to the technical field of semiconductor temperature control, and discloses a multi-channel liquid return temperature difference crosstalk decoupling control method and system, wherein the multi-channel liquid return temperature difference crosstalk decoupling control method comprises the following steps: collecting channel data and public cooling water side parameters of each channel, calculating channel heat load and local temperature residual error; calculating a crosstalk influence coefficient for each pair of channels within a preset hysteresis range, and constructing a liquid return temperature difference crosstalk matrix; calculating crosstalk source intensity and disturbed intensity of each channel, marking crosstalk source channels and disturbed channels, generating a public compensation amount, a source channel suppression amount and a disturbed channel compensation amount; forming a final control instruction and delivering the final control instruction to an actuator; and updating the liquid return temperature difference crosstalk matrix; the application represents the influence intensity and action direction between the crosstalk source channels and the target channels through the liquid return temperature difference crosstalk matrix, and adopts different control strategies for different disturbance sources, so that the accuracy of compensation object, compensation direction and compensation opportunity judgment is improved.

Description

Technical Field

[0001] This application relates to the field of semiconductor temperature control technology, and more specifically, to a multi-channel return liquid temperature difference crosstalk decoupling control method and system. Background Technology

[0002] Photolithography, etching, cleaning, and related process equipment in semiconductor manufacturing typically have high requirements for temperature stability. To ensure stable operation of process chambers, stages, chemical circuits, or localized heat load units within a set temperature range, a multi-channel coolant temperature circulation control cabinet is usually used to provide circulating coolant to multiple controlled objects, and the coolant temperature is regulated by temperature sensors, flow regulating valves, heat exchange units, refrigeration units, or heating compensation units.

[0003] Existing multi-channel coolant temperature control methods mostly use the supply temperature, return temperature, or temperature deviation of each channel as feedback quantities, and perform independent PID regulation, fixed flow distribution, or fixed valve position regulation on each channel. This type of method can achieve basic temperature control when the heat load of each channel is stable and the influence between channels is weak.

[0004] However, in actual equipment, multiple coolant channels often share the same cooling water side, heat exchange unit, return flow manifold, or common fluid path space. When the heat load of a certain channel changes rapidly, the return fluid temperature difference of that channel may change abruptly, and this will have a hysteretic temperature effect on other channels through the return flow manifold, heat exchange unit, or common fluid path, causing temperature deviations in other channels even when their own loads have not changed significantly. Changes in the inlet temperature, pressure, or total flow rate of the external cooling water side may also cause synchronous drift in the supply fluid temperature of multiple channels. This type of disturbance from the common cooling source and crosstalk of the return fluid temperature difference between channels are similar in temperature performance, and both may cause the temperature of multiple channels to deviate from the set value. Existing independent channel feedback control usually only corrects the temperature deviation of its own channel, making it difficult to determine whether the temperature deviation is caused by a disturbance from the common cooling source or by a sudden change in the return fluid temperature difference of a certain channel, and it is also difficult to determine the source channel, the affected channel, the direction of the crosstalk, the hysteresis time, and the intensity of the impact.

[0005] Therefore, if an independent temperature feedback control method is still used for each channel during the operation of a multi-channel coolant temperature circulation control cabinet, problems such as inaccurate compensation targets, unreasonable compensation directions, or delayed compensation timing may easily occur. For example, when crosstalk occurs between channels, the system may only passively correct the disturbed channel without suppressing the thermal disturbance of the source channel; it may also misjudge changes in the capacity of the common cooling water side as crosstalk between channels, thus making unnecessary flow or valve position adjustments to local channels. Summary of the Invention

[0006] The purpose of this application is to address the problem that in existing multi-channel coolant temperature control processes, when multiple channels share the same cooling water side, heat exchange unit, or return liquid manifold structure, it is difficult to distinguish between common cooling source disturbances and crosstalk between return liquid temperature differences between channels. Consequently, it is difficult to accurately determine the crosstalk source channel, the disturbed channel, the crosstalk direction, the lag time, and the influence intensity, leading to inaccurate compensation objects, unreasonable compensation directions, and insufficient stability of multi-channel temperature control. Therefore, this application provides a multi-channel return liquid temperature difference crosstalk decoupling control method and system.

[0007] This application provides a multi-channel return liquid temperature difference crosstalk decoupling control method, including the following steps:

[0008] According to the preset control cycle, the channel data and common cooling water side parameters of each channel are collected. The channel heat load is calculated based on the channel data, the common cooling source disturbance is calculated based on the common cooling water side parameters, and the local temperature residual is obtained based on the channel data and the common cooling source disturbance.

[0009] Based on the channel heat load and local temperature residual of each channel, the crosstalk influence coefficient is calculated for each pair of channels within the preset lag range, and the return liquid temperature difference crosstalk matrix is ​​constructed based on the crosstalk influence coefficient.

[0010] Based on the crosstalk matrix of the return liquid temperature difference, calculate the crosstalk source strength and the disturbance strength of each channel, mark the crosstalk source channel and the disturbance channel based on the crosstalk source strength and the disturbance strength, and generate the common compensation amount, the source channel suppression amount and the disturbance channel compensation amount based on the marked crosstalk source channel and the disturbance channel.

[0011] The common compensation amount of each channel, the source channel suppression amount, and the disturbance channel compensation amount are superimposed to form the final control command. The final control command is then subjected to amplitude limiting and rate limiting processing before being sent to the actuator.

[0012] The return liquid temperature difference crosstalk matrix is ​​updated after each preset stable operating condition window ends.

[0013] Furthermore, methods for calculating the disturbance of the common cooling source include:

[0014] Subtract the current values ​​of the inlet temperature, pressure, and total flow rate of the cooling water in the common cooling water side parameters from their respective preset stable historical average values. Divide the difference by the sum of the corresponding preset historical standard deviation and preset small positive number to obtain their respective standardized deviations.

[0015] The standardized deviations of inlet temperature, pressure, and total flow rate are multiplied by their respective preset common disturbance regression coefficients and then summed to obtain the common cooling source disturbance for the current control cycle.

[0016] Furthermore, methods for obtaining local temperature residuals include:

[0017] For each channel, the temperature deviation is obtained by subtracting the set temperature from the current channel data.

[0018] Based on the preset common cooling source disturbance dynamic response coefficient sequence of the channel, and the common cooling source disturbance amount corresponding to the current control cycle and the previous control cycle, a weighted sum is performed to obtain the common disturbance cumulative impact amount of the channel.

[0019] The local temperature residual of the channel at the current moment is obtained by subtracting the cumulative impact of common disturbances from the temperature deviation of the channel at the current moment.

[0020] Furthermore, methods for constructing the return liquid temperature difference crosstalk matrix include:

[0021] The change in channel heat load is obtained by calculating the difference between the current channel heat load and the previous control cycle.

[0022] For any channel pair consisting of a source channel and a target channel, the preset lag time candidate values ​​are traversed within the preset lag range. The correlation coefficient is calculated by aligning the channel heat load change sequence of the source channel with the local temperature residual sequence of the target channel according to each preset lag time candidate value.

[0023] The absolute values ​​of each correlation coefficient are normalized, and the maximum value after normalization is taken as the crosstalk influence amplitude. The crosstalk influence coefficient of the channel pair is obtained by combining the positive and negative signs of the corresponding correlation coefficients.

[0024] The crosstalk influence coefficient of the channel pair is corrected based on the crosstalk influence amplitude;

[0025] The crosstalk influence coefficients for all channel pairs are arranged according to the source channel and target channel numbers to form a return liquid temperature difference crosstalk matrix.

[0026] Furthermore, the method for correcting the crosstalk influence coefficient of the channel pair includes:

[0027] The number of valid samples is determined based on the operating status of the source and target channels;

[0028] Based on the influence of the channel on the number of effective samples and the amplitude of crosstalk within a preset sliding time window, estimate the lower limit of the confidence interval of the channel on the absolute value of the correlation coefficient;

[0029] When the crosstalk effect amplitude is greater than the preset crosstalk judgment threshold, the lower limit of the confidence interval is greater than zero, and the number of effective samples is greater than or equal to the preset minimum number of samples, the crosstalk effect coefficient of the channel pair is retained; otherwise, the crosstalk effect coefficient of the channel pair is set to zero.

[0030] Furthermore, methods for labeling crosstalk source channels and interfered channels include:

[0031] Based on the crosstalk influence coefficients of channel i as the source channel and the change in channel heat load, the crosstalk source strength of channel i is calculated.

[0032] Based on the crosstalk amplitude between all channels j (excluding channel i) and channel i, determine the maximum correlation lag time when channel j is the source channel and channel i is the target channel;

[0033] The disturbance intensity of channel i is calculated by taking channel j as the source channel and channel i as the target channel, taking the crosstalk influence coefficient, maximum correlation lag time, and heat load change.

[0034] When the absolute value of the crosstalk source strength of channel i is greater than the preset source strength threshold and the causal timing lead condition is met, channel i is marked as a crosstalk source channel; channels whose absolute value of the disturbance strength is greater than the preset disturbance strength threshold and which are not marked as crosstalk source channels are marked as disturbed channels.

[0035] Furthermore, methods for determining whether a causal time priority condition is met include:

[0036] For each channel j other than channel i, determine the maximum correlation lag time of channel i as the source channel and channel j as the target channel.

[0037] The timing of the change is determined based on the absolute value of the change in channel heat load or the local temperature residual.

[0038] If the time when the channel heat load change of channel i is triggered is earlier than the time when the local temperature residual of channel j is triggered in the corresponding direction, and the difference between the time interval between the two and the maximum correlation lag time is within the preset time tolerance range, then channel i is considered to satisfy the causal time lead relationship with channel j.

[0039] When channel i satisfies the above causal temporal lead relationship with at least one channel j other than itself, channel i is considered to satisfy the causal temporal lead condition.

[0040] Furthermore, the methods for generating the common compensation quantity, the source channel suppression quantity, and the disturbed channel compensation quantity include:

[0041] The maximum value of the absolute value of the crosstalk source strength of all crosstalk source channels in the current control cycle and the maximum value of the absolute value of the disturbance strength of all disturbed channels are taken as the representative crosstalk strength.

[0042] If the absolute value of the common cooling source disturbance is greater than the product of the crosstalk representative strength and the preset dominance judgment multiple, it is determined that the cooling source is dominant, and the common compensation amount is obtained by multiplying the preset common disturbance compensation gain with the common cooling source disturbance.

[0043] Otherwise, if the crosstalk strength is greater than the smaller of the preset source strength threshold and the preset disturbance strength threshold, then crosstalk is determined to be dominant. The preset source channel suppression gain is multiplied by the crosstalk source strength of the crosstalk source channel to obtain the source channel suppression amount; the preset disturbance channel compensation gain is multiplied by the disturbance strength of the disturbance channel to obtain the disturbance channel compensation amount. The compensation amount not enabled by the current dominant type is zero.

[0044] Furthermore, the method of sending the final control command to the actuator after amplitude limiting and rate limiting processing includes:

[0045] The basic feedback control quantity is generated based on the temperature deviation of the channel;

[0046] The final control command is calculated based on the basic feedback control quantity of the channel, the compensation quantity of the disturbed channel, the suppression quantity of the source channel, and the common compensation quantity.

[0047] The final control command is decomposed into actuator control quantities by using a preset control allocation matrix;

[0048] The control quantities of each actuator are sequentially subjected to amplitude limiting and rate limiting processing to obtain the actual control quantities issued.

[0049] The actual control input is sent to the actuator.

[0050] Furthermore, the method for updating the return liquid temperature difference crosstalk matrix includes:

[0051] After each preset stable operating condition window ends, the observation crosstalk influence coefficient of each channel pair is recalculated based on the channel heat load change sequence and local temperature residual sequence within that window.

[0052] For each crosstalk influence coefficient in the return liquid temperature difference crosstalk matrix, the value obtained by subtracting the preset learning rate is multiplied by the crosstalk influence coefficient before the update, and then the product of the preset learning rate and the recalculated observed crosstalk influence coefficient is added. The sum is used as the updated crosstalk influence coefficient. After all crosstalk influence coefficients are updated, the updated return liquid temperature difference crosstalk matrix is ​​obtained.

[0053] The updated return liquid temperature difference crosstalk matrix replaces the previous return liquid temperature difference crosstalk matrix in the calculation during the next control cycle.

[0054] This application provides a multi-channel return liquid temperature difference crosstalk decoupling control system for executing the aforementioned multi-channel return liquid temperature difference crosstalk decoupling control method; the system includes:

[0055] The data acquisition and calculation module collects channel data and common cooling water side parameters of each channel according to a preset control cycle. It calculates the channel heat load based on the channel data, calculates the common cooling source disturbance based on the common cooling water side parameters, and obtains the local temperature residual based on the channel data and the common cooling source disturbance.

[0056] The temperature difference crosstalk module calculates the crosstalk influence coefficient for each channel based on the channel heat load and local temperature residual within a preset lag range, and constructs the return liquid temperature difference crosstalk matrix based on the crosstalk influence coefficient.

[0057] The crosstalk identification module calculates the crosstalk source strength and the disturbance strength of each channel based on the crosstalk matrix of the return liquid temperature difference, marks the crosstalk source channel and the disturbance channel based on the crosstalk source strength and the disturbance strength, and generates a common compensation amount, a source channel suppression amount and a disturbance channel compensation amount based on the marked crosstalk source channel and the disturbance channel.

[0058] The control execution module superimposes the common compensation amount of each channel, the source channel suppression amount, and the disturbance channel compensation amount to form the final control command. After performing amplitude limiting and rate limiting processing on the final control command, it sends it to the actuator.

[0059] The online update module updates the return liquid temperature difference crosstalk matrix after each preset stable operating condition window ends.

[0060] The beneficial effects of this application are as follows: This application collects channel data and common cooling water side parameters of each channel according to a preset control cycle, calculates channel heat load based on channel data, calculates common cooling source disturbance based on common cooling water side parameters, and obtains local temperature residual from channel data and common cooling source disturbance. This can distinguish between multi-channel synchronous drift caused by external cooling water side and crosstalk between channel return liquid temperature difference, and avoid misjudging common cooling source disturbance as crosstalk of one channel to other channels.

[0061] This application calculates the crosstalk influence coefficient for each channel based on the channel heat load and local temperature residual within a preset hysteresis range, and constructs a return liquid temperature difference crosstalk matrix, enabling the crosstalk relationship between channels to be quantitatively expressed in matrix form. This not only identifies the existence of crosstalk but also characterizes the influence intensity, direction, and hysteresis characteristics between the crosstalk source channel and the target channel, providing a clear basis for subsequent decoupling control.

[0062] This application calculates the crosstalk source strength and disturbance intensity of each channel using a return liquid temperature difference crosstalk matrix, and marks the crosstalk source channel and the disturbed channel accordingly. Then, it generates a common compensation amount, a source channel suppression amount, and a disturbed channel compensation amount, enabling different control strategies to be adopted for different disturbance sources. When the common cooling source disturbance is dominant, common compensation is performed; when inter-channel crosstalk is dominant, source channel suppression and disturbed channel targeted compensation are performed simultaneously, thereby improving the accuracy of the compensation object, compensation direction, and compensation timing.

[0063] This application combines the common compensation amount, the source channel suppression amount, and the disturbed channel compensation amount to form the final control command. After performing amplitude limiting and rate limiting processing on the final control command, it sends it to the actuator. This can achieve decoupling compensation while avoiding excessive actuator movement or rapid changes, and reduce the risks of control overshoot, actuator impact, and mutual interference of multi-channel regulation.

[0064] This application updates the return liquid temperature difference crosstalk matrix after each preset stable operating condition window ends, so that the crosstalk matrix can be adjusted according to changes in channel load, liquid circuit state, heat exchange conditions and long-term operating conditions, avoiding mismatch between fixed crosstalk parameters and actual operating conditions, thereby improving the long-term temperature stability, channel independence and anti-interference capability of the multi-channel coolant temperature circulation control cabinet. Attached Figure Description

[0065] Figure 1 This is a flowchart illustrating a multi-channel return liquid temperature difference crosstalk decoupling control method according to this application;

[0066] Figure 2 This is an example diagram of the construction of the return liquid temperature difference crosstalk matrix for a multi-channel return liquid temperature difference decoupling control method according to this application;

[0067] Figure 3 This is a module example diagram of a multi-channel return liquid temperature difference crosstalk decoupling control system according to this application. Detailed Implementation

[0068] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0069] A multi-channel return liquid temperature difference crosstalk decoupling control method and system includes the following embodiments:

[0070] Example 1:

[0071] A multi-channel return liquid temperature difference crosstalk decoupling control method is disclosed. This method is applied to the controller of a multi-channel coolant temperature circulation control cabinet, applicable to multi-channel coolant circulation temperature control equipment with, but not limited to, 8 to 16 channels. The multi-channel coolant temperature circulation control cabinet includes a coolant supply pipeline, a return liquid pipeline, a return liquid manifold structure, a heat exchange unit, cooling water inlet and outlet, channel flow regulating valves, bypass valves, a circulation pump, a refrigeration unit, a heating compensation unit, a temperature sensor, a flow sensor, a pressure sensor, and a communication module. Multiple coolant channels share a cooling water side, a heat exchange unit, or a return liquid manifold structure. There exists a condition where a sudden change in the return liquid temperature difference of one channel causes hysteretic temperature crosstalk to other channels via the return liquid manifold structure or heat exchange unit. Simultaneously, fluctuations in external cooling water temperature, pressure, or flow rate can cause synchronous drift across multiple channels.

[0072] Before the method is started, the equipment first completes the channel connectivity confirmation, cooling water supply stability confirmation, and effective confirmation of the temperature and flow sensors of each channel. When any channel enters the process operation state, channel heat load switches, return liquid temperature difference exceeds the preset change threshold, or any parameter among the inlet temperature, pressure, and total flow of the common cooling water side deviates from its preset stable historical average value by more than a preset multiple of the corresponding preset historical standard deviation, the controller enters the return liquid temperature difference crosstalk decoupling control state and executes the following steps; the preset multiple is obtained by continuously statistically analyzing the deviation of the inlet temperature, pressure, and total flow of the common cooling water side from their historical average values ​​during the equipment commissioning phase when each channel is in the benchmark stable operating condition, and taking the deviation of each parameter during normal operation as the ratio of the deviation to the corresponding historical standard deviation. The maximum value of the ratio is used as the normal fluctuation boundary. A safety margin is then added to this maximum value to obtain a preset multiple. This multiple is higher than the deviation multiple of each parameter due to random fluctuations during normal steady-state operation, but lower than the typical level of deviation multiple of each parameter when a substantial supply anomaly occurs on the cooling water side. This ensures that decoupling control is not triggered under normal fluctuations, but can be triggered in a timely manner when an anomaly occurs on the cooling water side. The preset change threshold is calibrated during the equipment commissioning phase based on the fluctuation range of the return liquid temperature difference under the reference stable operating conditions of each channel. Its value is higher than the upper limit of the amplitude of normal steady-state fluctuations but lower than the critical value of process anomalies. This makes it insensitive to random measurement noise and normal fluctuations, but can trigger the crosstalk decoupling process in a timely manner for sudden changes in return liquid temperature difference caused by load switching or faults.

[0073] The reference stable operating condition refers to the condition where all control channels are in operation or standby mode, the channel temperature sensors, flow sensors, pressure sensors, and communication status are all valid, the common cooling water side is not in an alarm state, no process load switching occurs in any channel, and the inlet temperature, pressure, and total flow rate of the common cooling water side are all within their stable operating statistical range. The alarm state of the common cooling water side is determined by any parameter of the cooling water inlet temperature, pressure, or total flow rate exceeding the equipment safety threshold or by a sensor malfunction. The stable operating statistical range is determined by stable samples from the equipment commissioning phase or the initial stage of formal operation. During the stable sample collection period, no artificial step disturbances are performed on the cooling water side, no channel load switching is performed, and data corresponding to sensor malfunctions, actuator malfunctions, communication interruptions, and alarm states are excluded.

[0074] After entering the decoupled control state, the controller periodically acquires channel data for each channel according to a preset control cycle. The channel data includes supply temperature, return temperature, set temperature, channel flow rate, valve position, bypass ratio, channel start / stop status, and process operation status. It also acquires common cooling water side parameters, including the inlet temperature, pressure, and total flow rate of the common cooling water side. Furthermore, it acquires the refrigeration unit output power, heating compensation power, and circulation pump frequency. Simultaneously, it records and collects timestamps for each data point to facilitate subsequent lag time alignment and timing analysis. In this embodiment, the channel flow rate is the volumetric flow rate. When a mass flow sensor is used, the preset coolant density is no longer multiplied in the channel heat load calculation.

[0075] The controller also calls upon pre-stored preset coolant density and preset coolant specific heat capacity. These preset coolant density and specific heat capacity are calibrated by the average value of the physical and chemical parameters of the coolant used within the set operating temperature range of the equipment. They are used to combine the return liquid temperature difference with the channel flow rate to convert it into the channel heat load, thereby objectively characterizing the actual heat absorption of each channel at the process load end in terms of energy dimensions. This avoids misjudging the crosstalk source strength caused by the inability to distinguish between small flow rate large temperature difference channels and large flow rate small temperature difference channels using only the return liquid temperature difference or its change. The preset control cycle is determined during the equipment commissioning phase based on the thermal inertia time constant of the coolant temperature circulation system, the sensor sampling response speed, and the actuator action response time. Its value is shorter than the minimum thermal inertia time constant of the system to ensure timely capture of temperature changes, and longer than the shortest time required for sensor sampling and actuator response to ensure the integrity of data acquisition and command execution within each cycle.

[0076] For any channel, the controller subtracts the channel's supply temperature from its return temperature at the current moment. The difference is the return temperature difference, reflecting the temperature rise of the coolant after heat absorption at the process load end. Based on this, the controller multiplies the return temperature difference by the channel's flow rate, preset coolant density, and preset coolant specific heat capacity. The product is the channel heat load, objectively reflecting the actual heat absorption power at the process load end in energy dimensions. Simultaneously, the controller subtracts the channel's set temperature from its supply temperature, resulting in the temperature deviation, reflecting the channel's supply temperature's offset from the set value. The channel heat load and temperature deviation are the foundational data for subsequent common disturbance identification, crosstalk identification, and decoupling control. Using the flow-weighted channel heat load instead of just the return temperature difference avoids distortion in crosstalk source strength assessment caused by channel flow rate differences.

[0077] The above methods are as follows Figure 1 As shown, it includes the following steps:

[0078] Step 100: Collect channel data and common cooling water side parameters for each channel according to the preset control cycle; calculate channel heat load based on channel data; calculate common cooling source disturbance based on common cooling water side parameters; and obtain local temperature residual based on channel data and common cooling source disturbance.

[0079] In each control cycle, the controller calculates the channel heat load of each channel at the current moment in the manner described above, and arranges the channel heat loads of all channels participating in the control in ascending order of channel number to form the channel heat load state vector for the current control cycle. The dimension of this vector is equal to the number of coolant channels currently participating in the control.

[0080] The controller arranges the operating status identifiers of each channel according to the start / stop status and process operation status of each channel, in the same channel numbering order as the channel heat load state vector, to form the channel operating status vector for the current control cycle. The operating status identifier of each channel includes at least five categories: running, standby, bypass, stopped, and abnormal. By arranging the operating status identifiers in the same channel numbering order as the channel heat load state vector, it can be ensured that there will be no channel misalignment when comparing elements by index later, so that the validity mark strictly corresponds one-to-one with the heat load data of the corresponding channel.

[0081] The controller marks the validity of elements in the channel heat load state vector based on the channel operating state vector. For channels with values ​​of "disabled" or "abnormal," the controller does not include them in subsequent crosstalk source judgment, disturbed channel judgment, and correlation calculation. The purpose of the validity marking is that the temperature and flow data of disabled or abnormal channels do not have normal process meaning. If they are included in the crosstalk correlation calculation, it will introduce spurious correlations without physical meaning, resulting in non-zero elements in the crosstalk matrix that do not reflect the real coupling relationship, thus causing subsequent decoupling compensation to produce incorrect actions. However, the channel's occupancy impact on the diversion is retained in the common liquid circuit flow distribution calculation because even if the channel is in a disabled or abnormal state, its pipeline is still physically connected to the common liquid circuit, and it still has an actual diversion effect on the flow distribution of other channels.

[0082] Within the same control cycle of establishing the state vector, the controller reads the inlet temperature, pressure, and total flow rate of the current control cycle from the common cooling water side sensor, and calls the pre-stored preset stable operation historical average, preset historical standard deviation, and preset common disturbance regression coefficient. The preset stable operation historical average includes the historical average of inlet temperature, the historical average of pressure, and the historical average of total flow rate; the preset historical standard deviation includes the historical standard deviation of inlet temperature, the historical standard deviation of pressure, and the historical standard deviation of total flow rate. The preset stable operation historical average and preset historical standard deviation are obtained by continuously collecting and calculating the arithmetic mean and standard deviation of inlet temperature, pressure, and total flow rate during the initial stable operation period after the equipment is put into operation or after formal operation, under the benchmark stable operating conditions of each channel. The collection time covers the typical operating cycle of the equipment to ensure the representativeness and stability of the statistics.

[0083] The controller calculates the common cooling source disturbance for the current control cycle as follows: First, the current inlet temperature is subtracted from the preset stable historical average inlet temperature, and the difference is divided by the sum of the preset historical standard deviation of the inlet temperature and a preset small positive number to obtain the standardized deviation of the inlet temperature. Second, the current pressure is subtracted from the preset stable historical average pressure, and the difference is divided by the sum of the preset historical standard deviation of the pressure and the aforementioned small positive number to obtain the standardized deviation of the pressure. Third, the current total flow rate is subtracted from the preset stable historical average total flow rate, and the difference is divided by the sum of the preset historical standard deviation of the total flow rate and the aforementioned small positive number to obtain the standardized deviation of the total flow rate. The reason for using the standardized deviation instead of the original deviation as the calculation basis is that the physical dimensions and numerical magnitudes of the inlet temperature, pressure, and total flow rate are different. If the original deviations are directly weighted and superimposed, the parameter with a large numerical magnitude will dominate the calculation result of the common cooling source disturbance, while the parameter with a small numerical magnitude but a large actual impact will be submerged, and the actual contribution of each parameter to the synchronous drift of the supply temperature cannot be truly reflected.

[0084] The preset small positive number value is much smaller than the typical magnitude of the corresponding preset historical standard deviation under normal operating conditions. It is used to ensure that the division operation value remains stable without changing the normal value of the standardization deviation in the extreme case where the preset historical standard deviation approaches zero. Its specific value is determined by the controller's floating-point operation accuracy and the sensor's quantization resolution. Finally, the controller multiplies the standardized deviation of the inlet temperature, the standardized deviation of the pressure, and the standardized deviation of the total flow rate by their respective preset common disturbance regression coefficients, and adds the three products together. The result is the common cooling source disturbance amount for the current control cycle. This common cooling source disturbance amount retains its positive and negative signs and is used to characterize the net direction and net intensity of the common cooling water side fluctuation on the synchronous drift of the liquid supply temperature of each channel.

[0085] The preset common disturbance regression coefficients respectively characterize the net contribution direction and intensity of the three cooling water-side characteristics—inlet temperature, pressure, and total flow rate—to the synchronous drift of the supply temperature in each channel. Their values ​​are allowed to be positive or negative, and there is no fixed constraint on the sum of the three. The preset common disturbance regression coefficients are obtained during the equipment commissioning phase under conditions where typical cooling water-side disturbances are artificially applied and each channel is in steady-state operation. The synchronous drift of the supply temperature in each channel is used as the dependent variable, and the standardized deviations of the three characteristics—inlet temperature, pressure, and total flow rate—and their historical values ​​from one lag period to the preset maximum lag period are used as independent variables through multiple linear regression calibration. The reason for using multiple linear regression to calibrate the preset common disturbance regression coefficients is that the influence of inlet temperature, pressure, and total flow rate on the synchronous drift of the supply temperature in each channel is physically approximately linearly superimposed, and there may be some collinearity among the three. Multiple linear regression can simultaneously consider the three characteristics and... Under the condition of their historical lag values, their independent net contributions are separated to avoid the influence of a certain feature being incorrectly attributed to the regression coefficients of other features due to the neglect of a certain feature. The training steps of this regression calibration are as follows: during the equipment commissioning stage, three typical cooling water side disturbances of inlet temperature step, pressure step and total flow step are applied in sequence. When each type of disturbance is applied, the other two types of parameters are kept stable and each channel is in steady state. The synchronous drift of the liquid supply temperature of each channel and the standardized deviation of the three types of features of inlet temperature, pressure and total flow are continuously collected, as well as their historical values ​​from one lag period to the preset maximum regression lag period. After summing all the collected samples, a multiple linear regression model is constructed with the synchronous drift of the liquid supply temperature of each channel as the dependent variable and all the above standardized deviations and their historical values ​​as independent variables. The regression coefficients that minimize the sum of squared residuals between the linear combination of the dependent variable and the independent variable are solved by the least squares method. The obtained regression coefficients are the preset common disturbance regression coefficients.

[0086] The preset maximum regression lag period is determined by rounding up the ratio of the maximum physical lag time of the cooling water side disturbance to the supply temperature of each channel to the preset control period during the equipment commissioning phase. The regression calibration takes into account the physical relationship that an increase in inlet temperature usually leads to a decrease in the common heat exchange capacity and an increase in the supply temperature of each channel, and an increase in pressure or total flow usually leads to an increase in the common heat exchange capacity and a decrease in the supply temperature of each channel. Therefore, the corresponding regression coefficients can have opposite signs. The controller does not require all three regression coefficients to have the same sign or the sum of the three regression coefficients to be one, thereby avoiding the common cooling source disturbance being misjudged as zero or reversed under certain operating conditions due to the mismatch of positive and negative directions. The above processing first standardizes each deviation and then linearly weights and superimposes them according to the signed coefficients, so that the deviations of temperature, pressure and flow, three different dimensions, can be combined into a single common cooling source disturbance according to a unified scale and their respective physical directions.

[0087] The controller records the supply liquid temperature, temperature deviation, and common cooling source disturbance of each channel in the current control cycle and several control cycles corresponding to the length of the preset sliding time window. The length of the preset sliding time window is determined by the equipment commissioning stage based on the maximum physical lag time of the coolant transfer from the common cooling water side to the supply liquid end of each channel through the heat exchange unit and the typical time constant of the temperature response of each channel. Its value covers the complete time span from the occurrence of the common cooling source disturbance to the completion of the synchronous response of the supply liquid temperature of each channel, so as to ensure that the multi-channel unidirectional change process caused by the common disturbance can be completely captured within the window. If the supply liquid temperature of multiple channels changes in the same direction within the same preset sliding time window, that is, rises or falls at the same time, and the direction of the change is consistent with the direction of the synchronous drift component calculated by the common cooling source disturbance through the preset common cooling source disturbance dynamic response, then the controller marks this part of the supply liquid temperature drift as caused by the common cooling source disturbance, and does not directly attribute it to the return liquid temperature difference crosstalk of a certain channel.

[0088] After obtaining the common cooling source disturbance, the controller performs dynamic common disturbance subtraction for each channel to obtain the local temperature residual of that channel. Specifically, the controller subtracts a cumulative common disturbance effect from the temperature deviation of that channel at the current moment, and the resulting difference is the local temperature residual of that channel at the current moment. The calculation method of the cumulative common disturbance effect is as follows: the controller sequentially takes each response coefficient from the zeroth position to the first preset maximum common disturbance lag length position in the preset common cooling source disturbance dynamic response coefficient sequence of that channel, multiplies the response coefficient of this position by the common cooling source disturbance amount of the historical moment corresponding to the corresponding number of control cycles backward from the current moment, and then sums up the products of all positions one by one. The sum obtained is the cumulative common disturbance effect. Among them, the zeroth response coefficient corresponds to the common cooling source disturbance amount at the current moment, the first response coefficient corresponds to the historical common cooling source disturbance amount one control cycle backward, and so on until the response coefficient of the first preset maximum common disturbance lag length position corresponds to the historical common cooling source disturbance amount one control cycle backward.

[0089] The preset common cooling source disturbance dynamic response coefficient sequence is a set of finite-length response coefficients arranged sequentially from zero hysteresis to the preset maximum common disturbance hysteresis length. It is used to approximately describe the first-order inertia and pure hysteresis combined transmission relationship of the common cooling source disturbance through the cooling water side, heat exchange unit and then to the liquid supply end of the channel. The preset common cooling source disturbance dynamic response coefficient sequence is determined by the dynamic regression relationship between the channel temperature deviation and the common cooling source disturbance and its historical values ​​in the equipment commissioning stage or historical stable operation data. Specifically, on the benchmark stable operating condition sample, the channel temperature deviation is used as the dependent variable, and the common cooling source disturbance of the same control cycle and the control cycles of the preset maximum common disturbance hysteresis length in the previous control cycle are used as independent variables to perform least squares dynamic regression. The resulting set of regression coefficients is the preset common cooling source disturbance dynamic response coefficient sequence of the channel.

[0090] The reason for using the least squares dynamic regression method to calibrate the response coefficient sequence is that the transmission process of the common cooling source disturbance through the cooling water side, heat exchange unit, and then to the liquid supply end of each channel has the combined characteristics of first-order inertia and pure time delay. Its influence on the temperature deviation of the channel is distributed across the current moment and multiple historical moments. Least squares dynamic regression can simultaneously estimate the response weights at each historical moment, so that the resulting response coefficient sequence can fully describe the time-domain characteristics of the transmission process, thereby accurately separating the cumulative influence of the common disturbance on the channel at each historical moment in the subsequent convolution subtraction. The training steps of this dynamic regression are as follows: during the equipment commissioning phase, select the operating period of each channel under the baseline stable condition, and continuously collect the temperature deviation of the channel. The common cooling source disturbance amount of the same control cycle and the preset maximum common disturbance lag length of the control cycle is used to construct a multiple linear regression model with the channel temperature deviation as the dependent variable and the common cooling source disturbance amount at each of the above times as the independent variable. The least squares method is used to solve for the set of regression coefficients that minimize the sum of squared residuals between the linear combination of the dependent and independent variables. The obtained regression coefficients are the preset common cooling source disturbance dynamic response coefficient sequence of the channel from the zero position to the preset maximum common disturbance lag length position. Each response coefficient is allowed to be positive or negative to accommodate the physical action direction of the cooling water side inlet temperature, pressure, and total flow rate, as well as the net response direction at each historical time after thermal inertia delay.

[0091] The preset maximum common disturbance hysteresis length is obtained by rounding up the ratio of the longest physical transmission time from the cooling water side to the liquid supply end of each channel to the preset control cycle, and then adding a preset margin order. The preset margin order is determined during the equipment commissioning phase based on the number of additional cycles required for the temperature response tail of each channel to decay to a steady state. Its value is such that the total time span corresponding to the preset maximum common disturbance hysteresis length covers the entire time span from the occurrence of the common disturbance to the completion of a complete response in the channel and decay to a negligible level.

[0092] The local temperature residual is used to represent the local temperature deviation that still exists in the channel after deducting the common influence of the external cooling water side. Compared with static deduction based on the same control cycle, the above-mentioned deduction based on the dynamic response coefficient sequence convolution can separate the cumulative influence of the cooling water side disturbance on the channel in the current and the preset maximum common disturbance lag length of historical cycles from the temperature deviation. This avoids misjudging the cooling water side fluctuation as crosstalk of the return liquid temperature difference of a certain channel to other channels in subsequent steps, and also avoids the common disturbance residue from contaminating subsequent crosstalk identification due to the inertia and lag between the cooling water side disturbance and the channel response.

[0093] The channel heat load state vector, channel operation state vector, common cooling source disturbance and local temperature residual of each channel output in step 100 are used as the state input of the current control cycle and provided to steps 200 to 500.

[0094] Step 200: Based on the channel heat load and local temperature residual of each channel, calculate the crosstalk influence coefficient for each channel within a preset lag range. Construct a return liquid temperature difference crosstalk matrix based on the crosstalk influence coefficient, as detailed below. Figure 2 As shown.

[0095] The controller uses the channel heat load change of any channel in the channel heat load state vector obtained in step 100 as a possible crosstalk source, and the change of the local temperature residual of other channels obtained in step 100 as the disturbed response. Within a preset hysteresis range, it calculates the crosstalk influence coefficient for each pair of channels, retaining the positive and negative signs. In the process of constructing the crosstalk influence relationship in step 200, the channel that generates the channel heat load change is called the source channel, and the channel that receives the influence of the channel heat load change is called the target channel.

[0096] For any source channel, the controller subtracts the channel's heat load at the current moment from the channel's heat load at the time of the previous control cycle. The difference is the change in channel heat load in the current control cycle, where the difference between the current moment and the time of the previous control cycle is equal to a preset control cycle. The change in channel heat load uses the energy flow change after flow weighting as the basic physical quantity of crosstalk source. Therefore, it can avoid the problem that may occur when only the change in return liquid temperature difference is used as the basic physical quantity of crosstalk source, such as misjudging a small flow rate channel with a large temperature difference as a strong crosstalk source and a large flow rate channel with a small temperature difference as a weak crosstalk source.

[0097] For any channel pair consisting of a source channel and a target channel, the controller iterates through preset lag time candidate values ​​within a preset lag range. The controller then aligns the source channel's heat load change sequence with the target channel's local temperature residual sequence on the same historical sample using these preset lag time candidate values ​​before performing correlation calculations. This method of iterating through candidate values ​​within the preset lag range and aligning them one by one is used because crosstalk signals are transmitted from the source channel through the return flow manifold or heat exchange unit to the target channel with an uncertain physical transmission delay. This delay varies between different channel pairs due to factors such as pipe length, flow velocity, and liquid path topology. If lag alignment is not performed and the correlation is directly calculated at the same moment, the resulting delay would be significantly different. If the number is not calculated correctly, the actual crosstalk relationship may be underestimated or even missed due to the time misalignment between the source signal and the disturbed response. Specifically, the local temperature residual sequence of the target channel is delayed relative to the channel heat load change sequence of the source channel by the preset lag time candidate value, and then the correlation coefficient of the two sequences after delay and alignment is calculated. The preset lag time candidate value is selected one by one between the lower limit and the upper limit of the preset lag range according to the preset step size. The preset step size is determined by the equipment commissioning stage according to the preset control cycle and the time resolution requirements of pipeline transmission lag. Its value is equal to the preset control cycle or an integer multiple of the preset control cycle to ensure a balance between the search accuracy and calculation efficiency of the lag time.

[0098] The correlation coefficient is calculated as follows: For each sample point, the value of the source channel heat load change at that sample point is subtracted from its mean value over the entire window. The value of the target channel local temperature residual at the sample point corresponding to the current preset lag time candidate value is also subtracted from its mean value over the entire window. These two differences are multiplied, and the sum of these products over all sample points within the window is obtained to get the numerator. The square root of the sum of the squares of the differences between the source channel heat load change and the mean value within the window is used as the first factor. The square root of the sum of the squares of the differences between the target channel local temperature residual and the mean value within the window is used as the second factor. The first and second factors are multiplied together and then added to the preset small positive number to obtain the denominator. Finally, the numerator is divided by the denominator to obtain the correlation coefficient, retaining the positive and negative signs for the current preset lag time candidate value. A positive sign indicates that the target channel local temperature residual shows a positive-increasing coupling relationship when the source channel heat load change increases, while a negative sign indicates that the target channel local temperature residual shows a negative-decreasing coupling relationship when the source channel heat load change increases.

[0099] After obtaining the correlation coefficient with positive and negative signs for any preset lag time candidate value, the controller divides the absolute value of the correlation coefficient by a sum of a suppression factor determined by a preset liquid circuit coupling penalty factor and one, to obtain the normalized correlation amplitude for that preset lag time candidate value. At the same time, the sign of the correlation coefficient is retained separately as the correlation direction indicator for the corresponding lag time candidate value. The reason for introducing the preset liquid circuit coupling penalty factor for normalization is that calculations based purely on statistical correlation cannot distinguish between real crosstalk caused by physical coupling of the liquid circuit and spurious correlation caused by external common excitation or coincidental timing. Even if a high statistical correlation coefficient occurs between channel pairs that are physically far apart in the liquid circuit structure, their actual crosstalk transmission capacity is much lower than that of channel pairs that are physically close due to pipeline attenuation and mixing dilution. Therefore, it is necessary to physically constrain the statistical correlation amplitude according to the liquid circuit topology so that the final crosstalk influence coefficient reflects both statistical evidence and physical feasibility.

[0100] The larger the value of the preset liquid path coupling penalty factor, the less tightly the connection between the source channel and the target channel is in the liquid path structure. This suppression factor is used to suppress the crosstalk amplitude of channel pairs that are physically far apart in the liquid path structure. The preset liquid path coupling penalty factor is determined by the device before leaving the factory based on the geometric connection relationship between the source channel and the target channel in the coolant pipeline topology and the manifold structure. The controller traverses the above normalized correlation amplitudes within the entire preset hysteresis range and takes the maximum value as the crosstalk influence amplitude between the source channel and the target channel. At the same time, the correlation direction identifier corresponding to the hysteresis candidate value that obtains the maximum value is taken as the crosstalk influence direction sign of the channel pair. The crosstalk influence coefficient between the crosstalk influence amplitude and the crosstalk influence direction sign is the value that retains the positive or negative sign. This crosstalk influence coefficient can characterize both the intensity of the crosstalk and the direction of the crosstalk, avoiding the subsequent misjudgment of the compensation direction caused by simply taking the maximum correlation value and losing the negative coupling relationship.

[0101] The controller simultaneously records the candidate value of the preset lag time corresponding to the maximum value of the normalized correlation amplitude. This value is the maximum correlation lag time from the source channel to the target channel, representing the lag time of heat load change from the source channel to the target channel. The lower and upper limits of the preset lag range correspond to the minimum and maximum physical lag time of return liquid from the source channel to the target channel through the manifold or heat exchange unit. The preset lag range is obtained by estimating the minimum physical transmission time based on the pipeline geometry and the rated flow rate of the coolant before the equipment leaves the factory, and estimating the maximum physical transmission time based on the lowest operating flow rate and the farthest liquid coupling path, and then subtracting a preset lower limit margin in the direction of the lower limit and adding a preset upper limit margin in the direction of the upper limit. The preset lower limit margin and the preset upper limit margin are determined by the equipment commissioning stage based on the maximum negative deviation and the maximum positive deviation of the actual transmission lag under each operating condition relative to the rated operating condition transmission lag, so as to cover the fluctuation range of the actual transmission lag under each operating condition.

[0102] Before recording the crosstalk influence coefficient of the channel pair into the subsequent crosstalk matrix, the controller first performs crosstalk verification. The reason for introducing crosstalk verification is that, under limited sample conditions, the crosstalk influence amplitude estimated based on the correlation coefficient may be affected by measurement noise, multi-channel common process cycle drive, or random data fluctuations, resulting in false non-zero values. If it is recorded as the real crosstalk relationship into the crosstalk matrix without statistical verification, it will cause the subsequent decoupling control to apply unnecessary compensation actions to channel pairs that do not actually have crosstalk, thus introducing additional control disturbances and reducing system stability. The effective sample number refers to the total number of control cycles within a preset sliding time window in which both the source channel and the target channel are in an effective operating state and the corresponding channel heat load change and local temperature residual are valid acquisition values. This excludes cycles in which the source channel or target channel is in a stopped or abnormal state, cycles in which the acquisition values ​​are invalid due to sensor failure, communication interruption, or data abnormality, and cycles in which pairing cannot be performed due to exceeding the window boundary after hysteresis alignment. The remaining number of cycles that can simultaneously provide valid pairing data for the source channel heat load change and the target channel local temperature residual is the effective sample number.

[0103] The specific crosstalk verification method is as follows: The controller first estimates the lower limit of the confidence interval for the absolute value of the correlation coefficient of the channel based on the number of valid samples and the amplitude of the crosstalk effect. The estimation method is as follows: The controller subtracts two from the number of valid samples to obtain the degrees of freedom, then multiplies the square of the crosstalk effect amplitude by one minus the square of the crosstalk effect amplitude, multiplies the quotient by the degrees of freedom, and takes the square root to obtain the test statistic. The controller compares the test statistic with the preset one-sided critical value corresponding to the degrees of freedom. If the test statistic is greater than the preset one-sided critical value, the controller calculates the lower limit of the confidence interval as follows: the quotient is obtained by subtracting the preset one-sided critical value from the crosstalk effect amplitude, dividing by the number of valid samples minus one, and taking the square root. This difference is the absolute value of the correlation coefficient. The approximate estimate of the lower limit of the confidence interval is used to characterize the lowest true level that the channel's crosstalk influence amplitude may reach at a given confidence level. The reason for subtracting two from the effective sample size to obtain the degrees of freedom is that the calculation of the correlation coefficient involves estimating the mean of the source channel heat load change sequence and the target channel local temperature residual sequence. The estimation of the two means consumes two independent constraints. Therefore, the amount of independent information that can freely vary in the effective sample size is the effective sample size minus two. This degree of freedom determines the shape parameter of the sampling distribution that the test statistic follows. When the test statistic is less than or equal to the preset one-sided critical value, the lower limit of the confidence interval is set to zero, and the crosstalk influence coefficient of the channel pair is set to zero.

[0104] The reason for calculating the test statistic and comparing it with the preset one-sided critical value is that, under the null hypothesis that there is no real crosstalk correlation between the source channel and the target channel, the observed crosstalk amplitude may only be generated by random fluctuations of a finite sample. The test statistic converts the crosstalk amplitude into a standardized quantity that follows a known sampling distribution when the null hypothesis is true. By comparing this standardized quantity with the preset one-sided critical value, it can be determined whether the observed crosstalk amplitude exceeds the range that random fluctuations can explain under the null hypothesis. If it exceeds, the null hypothesis is rejected and the crosstalk relationship is determined to exist. If it does not exceed, the possibility that the crosstalk amplitude is generated by random fluctuations cannot be ruled out. The preset one-sided critical value is calculated by the equipment commissioning stage according to the confidence level and degrees of freedom required by the project. The corresponding confidence level is determined by the equipment commissioning stage based on a comprehensive trade-off between the project costs of misjudging and missing crosstalk.

[0105] As an example, the formula for calculating the degrees of freedom is expressed as follows:

[0106] ;

[0107] in, For degrees of freedom; The effective sample size is denoted by 2; the constant subtracted is the number of independent constraints consumed by the two estimated parameters: the mean of the source channel heat load change sequence and the mean of the target channel local temperature residual sequence.

[0108] The formula for calculating the above test statistic is expressed as follows:

[0109] ;

[0110] in, To test the statistic; Source channel To the target channel The amplitude is affected by crosstalk; For degrees of freedom; The square of the crosstalk effect amplitude; This is one minus the square of the crosstalk effect amplitude; The quotient is the square of the crosstalk effect amplitude divided by one minus the square of the crosstalk effect amplitude; this quotient is related to the degrees of freedom. The test statistic is obtained by multiplying and then taking the square root. .

[0111] The formula for calculating the lower limit of the above confidence interval is expressed as follows:

[0112] ;

[0113] in, This represents the lower limit of the confidence interval for the absolute value of the correlation coefficient. Source channel To the target channel The amplitude is affected by crosstalk; For degrees of freedom Preset one-sided threshold values ​​corresponding to confidence levels; The number of valid samples; Subtract one from the number of valid samples; The square root of the result after subtracting one from the number of valid samples; The quotient is the square root of the preset one-sided critical value divided by the number of valid samples minus one.

[0114] The controller simultaneously checks whether the number of valid samples within the window is greater than or equal to the preset minimum number of samples. Only when all three conditions are met—the crosstalk amplitude is greater than the preset crosstalk threshold, the lower limit of the confidence interval is greater than zero, and the number of valid samples is greater than or equal to the preset minimum number of samples—does the controller determine that there is real return liquid temperature difference crosstalk between the source channel and the target channel, and retains the amplitude and sign of the crosstalk influence coefficient of that channel, as well as the corresponding maximum correlation lag time. Otherwise, the crosstalk influence coefficient of that channel is set to zero to avoid spurious correlations caused by measurement noise, multi-channel shared process cycle time, or insufficient samples being mistakenly identified as actual crosstalk. The preset crosstalk judgment threshold is determined during the equipment commissioning phase under baseline stable operating conditions by statistically analyzing the noise distribution characteristics of the normalized correlation amplitude between each channel pair, and taking the level corresponding to the upper envelope of the noise fluctuation as the preset crosstalk judgment threshold. This threshold is higher than the random correlation caused by noise but lower than the correlation level when crosstalk actually exists. The preset minimum sample size is determined during the equipment commissioning phase, under the premise of ensuring that the statistical confidence of the correlation coefficient estimation meets the engineering requirements, based on the preset sliding time window length, preset control cycle, and channel response time. This prevents the correlation coefficient from being estimated with too few samples when the channel has just switched over, when the sample size is insufficient, or when the excitation is insufficient.

[0115] After performing the above calculations for all valid channel pairs, the controller will arrange all crosstalk influence coefficients, retaining both positive and negative signs, in the order of row number corresponding to the source channel number and column number corresponding to the target channel number. This forms a return temperature difference crosstalk matrix with both the number of rows and columns equal to the number of coolant channels currently under control. The controller will then set all diagonal elements of the same channel pair in this matrix to zero to exclude channel-specific temperature feedback. The reason for using a matrix structure to organize the crosstalk influence coefficients is that in a multi-channel system, any channel may simultaneously act as a crosstalk source affecting multiple target channels, or simultaneously act as a target channel being affected by multiple source channels. It can completely and without omission express the bidirectional coupling relationship between all channel pairs in a row and column index manner, so that subsequent steps can obtain the crosstalk source strength by row summation and the disturbance strength by column summation, thereby supporting the systematic hierarchical decoupling control of complex crosstalk patterns with multiple sources and multiple targets. Each non-zero element of the return liquid temperature difference crosstalk matrix retains a positive or negative sign. Positive elements indicate that an increase in the heat load of the source channel will cause an increase in the local temperature residual of the target channel, and negative elements indicate that an increase in the heat load of the source channel will cause a decrease in the local temperature residual of the target channel, thereby providing clear directional information for subsequent hierarchical decoupling control.

[0116] The return liquid temperature difference crosstalk matrix with positive and negative signs retained and the lag time matrix composed of the maximum correlation lag time of each channel pair output by step 200 are used by steps 300 to 500. The row number of the return liquid temperature difference crosstalk matrix corresponds to the source channel number and the column number corresponds to the target channel number. In step 300, the controller uses this matrix to calculate the crosstalk source strength and the disturbance strength of each channel, and officially marks the crosstalk source channel and the disturbance channel accordingly.

[0117] Step 300: Calculate the crosstalk source strength and disturbance intensity of each channel based on the crosstalk matrix of the return liquid temperature difference; mark the crosstalk source channel and the disturbance channel based on the crosstalk source strength and the disturbance intensity; and generate a common compensation amount, a source channel suppression amount and a disturbance channel compensation amount based on the marked crosstalk source channel and the disturbance channel.

[0118] Based on the return liquid temperature difference crosstalk matrix obtained in step 200 and the channel heat load change of each channel in the current control cycle, the controller calculates the crosstalk source strength of any channel as follows: Iterates through all other channels j except for channel i itself. For each channel j, the crosstalk influence coefficient corresponding to channel i as the source channel to channel j as the target channel in the return liquid temperature difference crosstalk matrix is ​​multiplied by the channel heat load change of channel i in the current control cycle. Then, the products corresponding to all channels j are added together. The sum is the crosstalk source strength of channel i in the current control cycle, retaining the positive and negative signs. It is used to characterize the net direction and net intensity of the influence of channel i as a crosstalk source on other channels. The larger the absolute value of the crosstalk source strength, the stronger the crosstalk effect. Its sign indicates whether the local temperature residual of other channels tends to increase or decrease as a whole.

[0119] As an example, the formula for calculating the crosstalk source strength is expressed as follows:

[0120] ;

[0121] in, For channel In the Each control cycle retains the strength of the crosstalk source with positive and negative signs; This represents the number of coolant channels currently under control. For the channels in the return liquid temperature difference crosstalk matrix As a source channel to channel As the crosstalk influence coefficient corresponding to the target channel; For channel In the The change in channel heat load for each control cycle; Taking a positive sign indicates a channel The overall change in heat load in one channel causes the local temperature residual in other channels to tend to increase, while a negative sign indicates a tendency to decrease.

[0122] If the absolute value of the crosstalk source strength of channel i is greater than the preset source strength threshold, and channel i satisfies the causal timing lead condition, then the controller will mark channel i as the crosstalk source channel of the current control cycle. The causal timing lead condition is introduced as an additional criterion for marking crosstalk sources because the absolute value of the crosstalk source strength exceeding the threshold alone cannot rule out the pseudo-causal relationship between the channel heat load change and the local temperature residual change of other channels as synchronous responses to the same external excitation. Only when the channel heat load change of the channel is indeed earlier than the local temperature residual change of other channels in time and the time interval is consistent with the physical transfer lag can it be confirmed that the channel is the real initiator of crosstalk rather than the synchronous victim, thereby avoiding the mistaken marking of the disturbed channel or synchronous response channel as a crosstalk source, which would lead to the wrong direction of subsequent suppression actions.

[0123] The determination method for the causal timing lead condition is as follows: For each channel j other than channel i, the controller reads the maximum relevant lag time from channel i as the source channel to channel j as the target channel from the lag time matrix. Then, it compares the triggering time of the channel heat load change of channel i at the current moment with the triggering time of the corresponding direction of the local temperature residual change of channel j. The corresponding direction refers to the fact that the sign of the change direction of the local temperature residual of the target channel is consistent with the sign of the crosstalk influence coefficient of the channel pair in the return liquid temperature difference crosstalk matrix after multiplying it with the current channel heat load change of the source channel. That is, when the sign is positive, the local temperature residual of the target channel increases, and when the sign is negative, the local temperature residual of the target channel decreases. If the triggering time of the channel heat load change of channel i is earlier than the current channel heat load change, the controller will determine the lead condition. When a corresponding directional trigger change occurs in the local temperature residual of channel j, and the time interval between the two deviates from the maximum correlation lag time within a preset timing tolerance range, then channel i is considered to satisfy a causal timing lead relationship with channel j. The trigger change refers to the moment when the absolute value of the channel heat load change or the local temperature residual is greater than the trigger term within a preset sliding time window. The trigger term is the mean of the absolute value of the channel heat load change or the local temperature residual within the preset sliding time window plus a product term. The product term is the preset trigger change multiple multiplied by the standard deviation of the absolute value of the channel heat load change or the local temperature residual within the preset sliding time window. The preset trigger change multiple and the preset timing tolerance range are both calibrated during the equipment commissioning phase based on the fluctuation range of sensor sampling accuracy and pipeline transmission lag.

[0124] When channel i satisfies the above causal timing lead relationship with at least one channel j other than itself, it is considered that channel i satisfies the causal timing lead condition. The preset source strength threshold is calculated by taking the upper limit of random fluctuation of the absolute value of crosstalk source strength of each channel under the reference stable operating conditions during the equipment commissioning phase. This preset source strength threshold is higher than the occasional intensity fluctuation caused by noise and measurement error under normal operating conditions, but lower than the intensity level when the dominant crosstalk source actually appears, thereby avoiding misjudgment.

[0125] Meanwhile, the controller calculates the disturbance intensity of any channel i as follows: it iterates through all other channels j except for channel i itself. For each channel j, the controller first reads the maximum correlation lag time from channel j as the source channel to channel i as the target channel from the lag time matrix. Then, it traces back from the current time to determine the historical time corresponding to the maximum correlation lag time as the lag alignment time of channel j. Next, it extracts the channel heat load change at the lag alignment time from the historical sequence of channel heat load change of channel j. Finally, it multiplies the crosstalk influence coefficient corresponding to channel j as the source channel to channel i as the target channel in the return liquid temperature difference crosstalk matrix with the channel heat load change of channel j at the lag alignment time to obtain the single-channel disturbance contribution of channel j to channel i.

[0126] The channel heat load change is calculated using the lag alignment time rather than the current time because there is a physical delay in the transmission of crosstalk signals from the source channel to the target channel. The crosstalk effect felt by the target channel at the current time actually originates from the thermal disturbance emitted by the source channel several control cycles ago. If the channel heat load change of the source channel at the current time is used, it will cause a time misalignment, making the calculation result of the disturbance intensity unable to correctly reflect the crosstalk effect actually borne by the target channel. The controller adds up the single-channel disturbance contribution of all channels j to channel i one by one. The sum is the disturbance intensity of channel i in the current control cycle, retaining the positive and negative signs. It is used to characterize the net direction and net intensity of the local temperature residual after channel i is subjected to crosstalk from the return liquid temperature difference of other channels in the next stage. The above processing of retrieving historical time according to the maximum correlation lag time is used to align the arrival time of the source disturbance to the disturbed channel. The summation method of retaining the positive and negative signs ensures that the disturbance intensity can reflect both the net effect of being pushed up and the net effect of being pushed down.

[0127] As an example, the formula for calculating the above disturbance intensity is expressed as:

[0128] ;

[0129] in, For channel In the Each control cycle retains the strength of the disturbance with positive or negative signs; This represents the number of coolant channels currently under control. For the channels in the return liquid temperature difference crosstalk matrix As a source channel to channel As the crosstalk influence coefficient corresponding to the target channel; For channel In the The maximum correlation lag time is traced back to the previous control cycle. The change in channel heat load at the corresponding hysteresis alignment time; For channels in the lag time matrix As a source channel to channel The maximum relevant lag time for the target channel; Taking a positive sign indicates a channel The local temperature residual is net pushed up, and a negative sign indicates net pushed down.

[0130] If the absolute value of the disturbance intensity of channel i is greater than the preset disturbance intensity threshold, and channel i itself is not marked as a crosstalk source channel, then the controller marks channel i as the disturbed channel in the current control cycle. The preset disturbance intensity threshold is calculated by statistically analyzing the random fluctuation upper limit of the absolute value of the disturbance intensity of each channel under the baseline stable operating conditions during the equipment commissioning phase. This threshold is higher than the occasional disturbance intensity fluctuation caused by noise and measurement errors under normal operating conditions, but lower than the disturbance intensity level presented by the disturbed channel when crosstalk transmission actually exists, thereby avoiding misjudging normal fluctuations as disturbances.

[0131] After identifying the crosstalk source channel and the disturbed channel, the controller generates a common compensation quantity, a source channel suppression quantity, and a disturbed channel compensation quantity based on the common cooling source disturbance quantity obtained in step 100 and the aforementioned crosstalk source strength and disturbed channel strength, which retain their positive and negative signs. All three are quantities that retain their positive and negative signs so that subsequent steps can uniquely determine the action direction of each actuator. The reason for using a hierarchical generation of three independent control components—common compensation quantity, source channel suppression quantity, and disturbed channel compensation quantity—is that the multi-channel coolant temperature circulation system simultaneously experiences two different types of disturbances: global synchronization drift caused by fluctuations on the common cooling water side and local differential offset caused by crosstalk between channel return fluid temperature differences. If only a single control quantity is used to uniformly address these disturbances, it will be impossible to distinguish the different mechanisms and objects of action of global disturbances and local crosstalk, which may easily lead to local compensation for global disturbances or local crosstalk. The directional error of global adjustment is addressed by generating three types of control components in a hierarchical manner. This allows the common compensation quantity to specifically target global synchronous drift for overall correction via a common actuator, the source channel suppression quantity to specifically target crosstalk sources for source suppression via a local actuator of that channel, and the disturbed channel compensation quantity to specifically target disturbed channels for directional compensation via a local actuator of that channel. Each component performs its own function without interfering with the others. In this embodiment, the following notation conventions are adopted: a positive sign for the final control command of each channel indicates an enhanced cooling effect on that channel, while a negative sign indicates a weakened cooling effect on that channel. A positive sign for the common cooling source disturbance quantity indicates that fluctuations on the cooling water side cause the supply temperature of each channel to tend to increase, while a negative sign indicates that fluctuations on the cooling water side cause the supply temperature of each channel to tend to decrease. The meanings of the signs for crosstalk source strength and disturbed strength are consistent with the crosstalk influence coefficient in step 200.

[0132] When the cooling source is dominant, the controller prioritizes adjusting the common cooling capacity, common heating compensation, or cooling water side adjustment. The determination of cooling source dominance is as follows: the controller takes the maximum value of the absolute crosstalk source strength of all channels marked as crosstalk sources and the maximum value of the absolute disturbance strength of all channels marked as disturbed as crosstalk sources in the current control cycle as the representative crosstalk strength. When there are no crosstalk source channels, the maximum absolute value of the crosstalk source strength is 0; when there are no disturbed channels, the maximum absolute value of the disturbance strength is 0. If the common cooling source is disturbed... If the absolute value of the momentum is greater than the product of the crosstalk intensity and the preset dominance determination multiple, then the cooling source is determined to be dominant. The preset dominance determination multiple is a positive value greater than one, which is calibrated during the equipment commissioning phase under the conditions of applying two typical operating conditions: common cooling source disturbance and inter-channel crosstalk, with the goal of correctly distinguishing between the common cooling source dominant condition and the inter-channel crosstalk dominant condition. At this time, the controller multiplies the preset common disturbance compensation gain with the common cooling source disturbance amount of the current control cycle, and the resulting product is the common compensation amount that retains the positive or negative sign for the current control cycle.

[0133] When the common cooling source disturbance is positive, meaning the cooling water side fluctuation causes the liquid supply temperature of each channel to tend to rise, the common compensation amount is also positive. After superposition, the final control command tends to enhance common cooling, corresponding to increasing the cooling water side adjustment, increasing the common cooling output, or reducing the common heating compensation. When the common cooling source disturbance is negative, the common compensation amount is also negative, corresponding to the actuator action in the opposite direction. The preset common disturbance compensation gain is taken as a positive value. It is obtained by calibrating the response sensitivity of the common cooling unit, common heating compensation unit, or cooling water side adjustment component under the condition of artificially applying typical disturbances on the cooling water side during the equipment commissioning stage, with the goal of eliminating the synchronous drift of the liquid supply temperature of each channel caused by the common cooling source disturbance. This makes the product of this gain and the common cooling source disturbance amount approximately cancel out the synchronous drift component.

[0134] When crosstalk is dominant, the controller performs source suppression on the crosstalk source channel and directional compensation on the disturbed channel. The method for determining crosstalk dominance is to compare the absolute value of the common cooling source disturbance in the current control cycle with the representative strength of the crosstalk. If the absolute value of the common cooling source disturbance does not meet the above-mentioned cooling source dominance determination condition, and the representative strength of the crosstalk is greater than the minimum of a preset source strength threshold or a preset disturbed strength threshold, then crosstalk is determined to be dominant. For a channel marked as a crosstalk source, the controller multiplies the preset source channel suppression gain by the crosstalk source strength of that channel; the result is the crosstalk source strength of that channel in the current control cycle. The source channel suppression amount retains a positive or negative sign during the control cycle, which is used to reduce the net crosstalk effect of this channel on other channels from the source. When the crosstalk source strength is positive, that is, when the overall heat load change of this channel causes the local temperature residual of other channels to tend to rise, the source channel suppression amount is also positive. After superposition, the final control command is directed to enhance the cooling of this source channel, correspondingly reducing the bypass ratio of this channel, increasing the flow rate of this channel, or increasing the valve opening of this channel, thereby reducing the return liquid temperature of this channel to reduce its thermal shock to the common liquid circuit or heat exchange unit. When the crosstalk source strength is negative, the source channel suppression amount is also negative, corresponding to the actuator action in the opposite direction.

[0135] The preset source channel suppression gain is set to a positive value. This value is determined during equipment commissioning when a single channel is artificially set as the dominant crosstalk source. The goal is to reduce the net impact of the suppressed crosstalk source channel on the local temperature residuals of other channels to noise levels. The sensitivity of the channel's flow rate, valve position, or bypass actuator is calibrated to ensure that the product of this gain and the crosstalk source strength matches the net impact of the channel's coolant on the common fluid path or heat exchange unit. For channels marked as disturbed, the controller multiplies the preset disturbed channel compensation gain by the disturbance strength of that channel. The result is the disturbance compensation amount for that channel in the current control cycle, retaining its positive or negative sign, used to compensate for the hysteresis temperature crosstalk experienced by that channel. When the disturbance strength is positive, meaning the local temperature residual of the channel is pushed up, ... The compensation amount for the disturbed channel is also positive. After superposition, the final control command points to enhance the cooling of the disturbed channel, which corresponds to increasing the valve position of the channel, increasing the local flow rate of the channel, lowering the target liquid supply temperature of the channel, or increasing the cooling distribution of the channel. When the disturbance intensity is negative, the compensation amount for the disturbed channel is also negative, and the actuator moves in the opposite direction. The preset compensation gain for the disturbed channel is taken as a positive value. It is obtained by calibrating the response sensitivity of the correction amount of the valve position, local flow rate, or target liquid supply temperature of the disturbed channel under the condition of artificially creating typical crosstalk between channels during the equipment commissioning stage, with the goal of making the amplitude of the local temperature residual of the disturbed channel no longer significantly increase after the lag time arrives. This makes the product of the gain and the disturbance intensity approximately offset the local temperature deviation caused by the lag temperature crosstalk borne by the disturbed channel.

[0136] Step 300 outputs the crosstalk source strength, the disturbance strength, the crosstalk source channel marker, the disturbance channel marker, the common compensation amount, the source channel suppression amount, and the disturbance channel compensation amount for each channel in the current control cycle, for use by step 400 in synthesizing the final control command; for channels that are neither crosstalk sources nor disturbance channels, the corresponding source channel suppression amount and disturbance channel compensation amount are set to zero.

[0137] Step 400: The common compensation amount of each channel, the source channel suppression amount, and the disturbance channel compensation amount are superimposed to form the final control command. The final control command is then subjected to amplitude limiting and rate limiting processing before being sent to the actuator.

[0138] The controller superimposes the basic feedback control quantity for each channel with the hierarchical decoupling control quantity obtained in step 300 to form the final control command for that channel, retaining its positive or negative sign in the current control cycle. The method of superimposing the basic feedback control quantity and the hierarchical decoupling control quantity to generate the final control command is used because the basic feedback control quantity is responsible for eliminating temperature deviations caused by the channel's own temperature tracking and load changes, while the hierarchical decoupling control quantity is responsible for offsetting external interference caused by fluctuations from the common cooling water side and crosstalk from other channels. Since their targets are different and orthogonal, their superposition enables the channel to simultaneously possess the ability to quickly track its own deviations and actively suppress external crosstalk. Using either component alone cannot simultaneously address both types of disturbances. The elimination is achieved by adding the product of the basic feedback control quantity of the channel, the product of the preset common disturbance compensation allocation coefficient and the common compensation quantity, the product of the preset source channel suppression allocation coefficient and the source channel suppression quantity of the channel, and the product of the preset disturbed channel compensation allocation coefficient and the disturbed channel compensation quantity of the channel. The sum is the final control command of the channel before it is limited in the current control cycle. If the channel is not marked as a crosstalk source, its source channel suppression quantity is zero; if the channel is not marked as a disturbed channel, its disturbed channel compensation quantity is zero. The above three preset allocation coefficients are used to adjust the weights of the common disturbance compensation component, the source channel suppression component, and the disturbed channel compensation component relative to the basic feedback control quantity under different disturbance dominance conditions.

[0139] The preset common disturbance compensation allocation coefficient, preset source channel suppression allocation coefficient, and preset disturbed channel compensation allocation coefficient are all greater than zero and less than one. They are calibrated during the equipment commissioning phase under three typical disturbances: common cooling source disturbance, single channel crosstalk source mutation, and typical disturbed response. The goal is to minimize the deviation of the liquid supply temperature of each channel from the set temperature after the final control command is applied, and to prevent repeated reversals of the control direction. During calibration, the three preset allocation coefficients are made to work synergistically with the basic feedback control quantity, so that the weight of each decoupled component relative to the basic feedback control quantity can effectively offset the corresponding disturbance without causing overshoot or oscillation. Furthermore, the preset common disturbance compensation allocation coefficient is dominant under the condition that common disturbance is dominant, and the preset source channel suppression allocation coefficient and the preset disturbed channel compensation allocation coefficient are dominant under the condition that channel crosstalk is dominant.

[0140] The aforementioned basic feedback control quantity is generated by the channel based on its temperature deviation. The generation method is as follows: multiply the channel's preset proportional gain by the channel's temperature deviation at the current moment to obtain the proportional component; multiply the channel's preset integral gain by the integral of the channel's temperature deviation over time within a preset integral effective window to obtain the integral component; multiply the channel's preset differential gain by the derivative of the channel's temperature deviation with respect to time at the current moment to obtain the differential component; and add the proportional, integral, and differential components together to obtain the basic feedback control quantity of the channel at the current moment.

[0141] The basic feedback control quantity is generated by combining proportional, integral, and derivative components. The proportional component provides immediate correction to the current temperature deviation, enabling rapid system response. The integral component eliminates accumulated historical deviations, ensuring no steady-state error in the system. The derivative component predicts the trend of temperature deviation changes, allowing the system to suppress deviations before they escalate. The synergistic effect of these three components balances response speed, steady-state accuracy, and dynamic stability. Using any single component or any combination of two components alone cannot simultaneously meet these three performance requirements. The length of the preset effective integral window is determined during equipment commissioning based on the open-loop response time of the channel's temperature control loop. Its value covers the time span required for the temperature deviation to converge after the basic feedback control quantity is applied, ensuring that the integral component can eliminate steady-state deviations without excessive accumulation of historical deviations due to an excessively long window.

[0142] To prevent the integral component from accumulating continuously when the actuator is saturated or the decoupling component is at a high amplitude for a long time, causing integral saturation and runaway, the controller, after the final control command of the channel is controlled, distributed and limited, detects in real time whether any actuator in the channel has reached the limiting boundary or the rate limiting boundary. If it has been reached and the integral component continues to accumulate in the same direction, it will further deepen the saturation. Therefore, the accumulation of the integral component by the temperature deviation of the channel is stopped during the saturation period. If the actuator exits saturation or the temperature deviation direction is reversed, the normal accumulation of the integral component is restored, thereby achieving anti-integral saturation. The preset proportional gain, preset integral gain and preset derivative gain are all taken as positive values. They are tuned during the equipment commissioning phase for each channel under the condition that the decoupling component is turned off, with the goal of the liquid supply temperature of the channel quickly converging to the set temperature under the set temperature step or small load step without obvious overshoot and continuous oscillation.

[0143] The controller decomposes the final control command, which retains the sign of each channel, into actuator control quantities through a preset control allocation matrix, based on the magnitude and sign of the command. These actuator control quantities include the valve position adjustment of the channel flow regulating valve, the proportional adjustment of the channel bypass valve, the frequency adjustment of the branch pump or circulation pump, the output adjustment of the refrigeration unit, the output adjustment of the heating compensation unit, and the adjustment of the common cooling water side regulating components. This ensures that the direction of action of each actuator is uniquely consistent with the temperature regulation direction required by the final control command, avoiding conflicts where multiple actuators have opposite directions and cancel each other out.

[0144] The reason for using a preset control allocation matrix to decompose a single final control command into multiple actuator control quantities is that the temperature regulation capability of each channel is provided by multiple physical actuators. Each actuator has different response speeds, regulation ranges, coupling degrees to other channels, and physical constraints. If the final control command is directly issued to a single actuator, it may exceed the actuator's regulation capability or cause secondary coupling to other channels. However, by allocating the control commands rationally according to the characteristics of each actuator using an allocation matrix, the coordinated regulation capability of multiple actuators can be fully utilized, achieving the requirements of the final control command while satisfying physical constraints. Temperature regulation effect; The preset control allocation matrix is ​​calibrated before the equipment leaves the factory based on the action direction, response speed, physical constraints and mutual coupling relationship of each actuator. Specifically, the static gain and dynamic response time of each actuator under typical working conditions are tested on bench or in the field. Then, the corresponding allocation ratio is assigned to each actuator according to the priority order of response speed from fast to slow and coupling relationship from weak to strong. This ensures that the demand for enhanced cooling direction is preferentially undertaken by actuators with fast response speed and weak coupling to other channels, while the action of the common cooling water side adjustment component is reserved as the main control means under the common disturbance dominant working condition.

[0145] When the final control command is positive and indicates enhanced cooling, the opening of the corresponding channel flow regulating valve increases, the channel flow increases, the bypass ratio decreases, the cooling output increases, or the heating compensation decreases. When the final control command is negative and indicates weakened cooling or to avoid overcompensation, the opening of the corresponding channel flow regulating valve decreases, the channel flow decreases, the bypass ratio increases, the cooling output decreases, or the heating compensation increases. When the absolute value of the common compensation is greater than the sum of the absolute values ​​of the source channel suppression and the absolute values ​​of the disturbed channel compensation, the controller prioritizes adjusting the common refrigeration unit or the cooling water-side regulating component to avoid miscompensation of local channels. The control allocation matrix ensures that the action directions of each actuator are consistent with each other at any given time and do not conflict with each other.

[0146] Before outputting control commands to the actuators, the controller sequentially performs amplitude limiting and rate limiting processing on the actuator control quantities after control allocation for each of the above channels. The reason for performing amplitude limiting and rate limiting processing on the actuator control quantities is that, under certain extreme conditions, the decoupled control algorithm may calculate control quantities that exceed the physical range of the actuator or generate excessive control quantity jumps between adjacent control cycles. If these are directly sent to the actuators without constraints, it may lead to mechanical damage to the actuators, overpressure impact on the coolant pipelines, or damage to the process load due to sudden temperature changes. Amplitude limiting processing ensures that the control quantity is always within the range allowed by the physical capabilities of the actuators and process safety, while rate limiting processing ensures that the rate of change of the control quantity does not exceed the maximum rate of change allowed by the mechanical response capability of the actuators and the pressure bearing capacity of the pipelines. Together, they constitute a safety protection mechanism for the control output.

[0147] The limiting method is as follows: if the actuator control quantity is less than its corresponding preset minimum control quantity, it is set to the preset minimum control quantity; if the actuator control quantity is greater than its corresponding preset maximum control quantity, it is set to the preset maximum control quantity; if the actuator control quantity is between the preset minimum control quantity and the preset maximum control quantity, it remains unchanged. The rate limiting method is as follows: the change amplitude of the actuator control quantity in this control cycle relative to the control quantity issued in the previous control cycle is limited to its corresponding preset maximum change rate. If the calculated change amplitude exceeds the preset maximum change rate, the change amplitude is limited to the product of the preset maximum change rate and the preset control cycle, to avoid the actuator from exhibiting excessively large step movements in a single control cycle. The result obtained after the above limiting and rate limiting processing is the actual control quantity issued by the channel to the actuator. The preset minimum control quantity and the preset maximum control quantity... The control quantities are the lower and upper limits allowed by the actuator, determined by the actuator's physical range and process safety limits. Specifically, for each channel, the lower and upper limits of the factory physical range of the channel's flow regulating valve position, bypass valve ratio, circulating pump frequency, refrigeration unit output, and heating compensation unit output are taken. These limits are then tightened in conjunction with the process safety limits of the channel's coolant, piping components, and process load, such as the minimum flow rate, maximum temperature, and maximum pressure. The more stringent of the two is the preset minimum control quantity and preset maximum control quantity of the corresponding actuator, ensuring that the controlled quantity after the limit is neither exceeded by the actuator's physical capacity nor breaches the process safety boundary. The preset maximum rate of change is determined during the equipment commissioning phase based on the actuator's mechanical response capability, the coolant piping pressure-bearing capacity, and the requirement to avoid process shocks, ensuring that the change in the control quantity is within the actuator's safe response range and does not cause excessive disturbance to the process load.

[0148] The controller simultaneously monitors the symbol sequence of the final control command of each channel within the most recent consecutive control cycles of the preset oscillation monitoring window. If the number of reversals of the final control command of a certain channel within the preset oscillation monitoring window reaches or exceeds the preset oscillation reversal threshold, the controller determines that the channel has a risk of decoupling oscillation. The reason for introducing the oscillation monitoring mechanism is that the estimation error of the crosstalk influence coefficient, the identification deviation of the lag time, or the closed-loop coupling between multiple channels in the hierarchical decoupling control may cause overcompensation or directional deviation between the decoupling compensation amount and the actual crosstalk amount, which in turn causes the final control command of a certain channel to frequently reverse between enhanced cooling and weakened cooling, forming oscillation. If the oscillation is not detected and suppressed in time, the liquid supply temperature of the channel will continue to fluctuate and cannot converge. At the same time, the oscillation will be propagated to other channels through the crosstalk path, causing the overall system to become unstable.

[0149] The preset oscillation reversal number threshold is determined during the equipment commissioning phase based on the maximum normal frequency of the final control command sign reversal during normal adjustment. Its value is higher than the upper limit of the sign reversal number caused by temperature tracking or small load fluctuations during normal adjustment, but lower than the number of sign reversals within the same window when decoupling oscillation actually occurs. This threshold distinguishes between normal adjustment and oscillation. When a channel is determined to have a risk of decoupling oscillation, the controller reduces the preset source channel suppression allocation coefficient and the preset disturbed channel compensation allocation coefficient corresponding to that channel to the product of their original values ​​and the preset oscillation suppression attenuation factor. The preset oscillation suppression attenuation factor is greater than zero and less than one, determined during the equipment commissioning phase under conditions of artificially created decoupling oscillation. The reduced allocation coefficient is calibrated with the goal of enabling the oscillation to converge within a preset oscillation monitoring window. Simultaneously, the proportion of the basic feedback control quantity in the final control command of this channel is temporarily increased, causing the channel's control behavior to revert to being primarily based on basic feedback control. During the switching process of the allocation coefficient decreasing from its original value to a new value and then returning to its original value, the controller smoothly changes the allocation coefficient using a linear transition or first-order filtering method according to a preset smooth transition duration. Simultaneously, the cumulative reference of the integral component in the basic feedback control quantity is synchronously corrected, ensuring that the overall final control command remains continuous at the moment of switching, without any abrupt changes, thus achieving a disturbance-free switching. After the oscillation risk disappears, the controller restores the allocation coefficient to its original preset value using the same smoothing method.

[0150] The preset oscillation monitoring window length is determined during the equipment commissioning phase based on the oscillation cycle and control response time that may occur under typical decoupled control conditions. Its value covers at least two complete potential oscillation cycles to ensure that the oscillation mode of frequent control direction reversal can be reliably identified without misjudging the normal adjustment process as oscillation due to an excessively short window. The preset smooth transition time is calibrated during the equipment commissioning phase under typical operating conditions with the goal of ensuring that the final control command does not exhibit a perceptible step and does not introduce new oscillations during the allocation coefficient switching process.

[0151] The actual control quantities output in step 400 are used as control commands to be sent to each actuator, driving the flow regulating valve, bypass valve, circulating pump, refrigeration unit, heating compensation unit and common cooling water side regulating components to operate.

[0152] Step 500: Update the return liquid temperature difference crosstalk matrix after each preset stable operating condition window ends.

[0153] After each preset stable operating condition window ends, the controller compares the local temperature residuals of each channel obtained after the actual decoupling compensation performed in steps 100 to 300 within that window with the local temperature residuals of each channel at the beginning of the window, and recalculates the observed crosstalk influence coefficient of each channel pair within that window. The reason for periodically updating the return liquid temperature difference crosstalk matrix online is that the actual crosstalk relationship between each channel will slowly drift due to factors such as pipeline aging, heat exchanger scaling, coolant performance degradation, changes in channel flow distribution, or changes in process load characteristics during long-term operation of the equipment. If the initial crosstalk matrix calibrated during the equipment commissioning phase is always used, it will deviate more and more from the actual crosstalk relationship as the operating time increases, causing the direction or magnitude of decoupling compensation to gradually become inaccurate. Online updates enable the crosstalk matrix to continuously track changes in the actual crosstalk relationship, ensuring that decoupling control remains effective throughout the entire life cycle of the equipment.

[0154] The calculation of the observed crosstalk influence coefficient is based on the channel heat load change sequence and local temperature residual sequence within the window. The correlation coefficient calculation, normalization based on a preset liquid circuit coupling penalty factor, and crosstalk verification based on a preset crosstalk judgment threshold and a preset minimum sample size are performed as described in step 200 to obtain the observed crosstalk influence coefficient corresponding to the window. Unlike the crosstalk influence coefficient calculated based on the original uncontrolled data in step 200, the local temperature residual sequence here is calculated using channel data and common cooling water side parameters collected after the actual control quantity was issued to each actuator in step 400, following step 100. Therefore, this observed crosstalk influence coefficient reflects... Under the current controller parameters and control actions already implemented, the residual effect of the source channel heat load change on the local temperature residual of the target channel is used to evaluate the actual effect of decoupling compensation and provide a basis for subsequent online updates. The observed crosstalk influence coefficient also retains its positive and negative signs. The length of the preset stable operating condition window is determined during the equipment commissioning phase based on the time required for the system to complete a full response cycle of crosstalk propagation and decoupling compensation under typical stable operating conditions. Its value covers the entire time span from the occurrence of crosstalk to the full manifestation of the decoupling compensation effect, so as to ensure that the compensation effect can be effectively evaluated at the end of the window and that the compensation effect is not judged before it is fully manifested due to the window being too short.

[0155] If, after decoupling compensation, the local temperature residual of a certain disturbed channel decreases in the direction predicted by the original crosstalk influence coefficient, and the observed crosstalk influence coefficient has the same sign as the original crosstalk influence coefficient, then the controller considers that the crosstalk influence coefficient of the corresponding channel pair obtained in step 200 is consistent with the actual crosstalk relationship in both direction and amplitude, retains the corresponding crosstalk influence coefficient, and corrects it according to the following online update formula; if, after compensation, the local temperature residual of the disturbed channel does not decrease in the predicted direction or increases in the opposite direction, or the observed crosstalk influence coefficient has a different sign than the original crosstalk influence coefficient, then the controller considers that the corresponding crosstalk influence coefficient is too large in amplitude. If there is a misjudgment in direction, the original crosstalk influence coefficient is corrected by the observed crosstalk influence coefficient according to the online update formula below, so that its amplitude is reduced or its sign is changed; the controller updates each element in the return liquid temperature difference crosstalk matrix online in the following way: the value obtained by subtracting the preset learning rate from one is multiplied by the crosstalk influence coefficient with the sign retained before the update, and then the product of the preset learning rate and the observed crosstalk influence coefficient with the sign retained recalculated in the current running window is added. The sum is the crosstalk influence coefficient with the sign retained after the update. After all elements are updated, the updated return liquid temperature difference crosstalk matrix is ​​obtained.

[0156] The crosstalk influence coefficient is updated online using an exponentially weighted moving average because this method can gradually absorb new observation results while retaining historical calibration information. This ensures that the updated crosstalk influence coefficient will not experience large jumps due to single observation noise, nor will it fail to track the slow drift of the actual crosstalk relationship by completely ignoring new observations, thus achieving a balance between tracking speed and noise immunity. The preset learning rate is greater than zero and less than one. During the equipment commissioning phase, under typical stable operating conditions, the tracking error of the actual crosstalk relationship is continuously converged within multiple consecutive preset stable operating condition windows using the crosstalk matrix, without large jumps in parameters caused by single occasional disturbances. The objective is to select the optimal value that balances tracking speed and noise immunity by comparing different candidate values ​​on historical data under stable operating conditions. The training steps for this online update process are as follows: during the equipment debugging phase, historical operating data covering multiple typical stable operating conditions and typical crosstalk operating conditions are collected. For each candidate preset learning rate value, the online update process of multiple consecutive preset stable operating condition windows is simulated. The cumulative tracking error and single update jump amplitude of the crosstalk matrix to the actual crosstalk relationship under each candidate value are calculated. The candidate value with the smallest tracking error that meets the convergence requirement and whose single update jump amplitude does not exceed the preset jump tolerance limit is selected as the final preset learning rate.

[0157] To avoid misusing observation results and contaminating the crosstalk matrix when identification conditions are poor, the controller temporarily reduces the preset learning rate to a preset conservative learning rate, or excludes the observation results of the current window from the crosstalk matrix update, under any of the following learning freeze conditions. The reason for introducing the learning freeze mechanism is that the reliability of the observation crosstalk influence coefficient on which online updates rely depends on the steady state and sufficiency of the data within the observation window. When the channel start / stop status changes, the common cooling source experiences severe disturbances, the actuator remains saturated for a long time, or the signal excitation is insufficient, the local temperature residual changes within the observation window may not be caused by inter-channel crosstalk but by the aforementioned non-steady-state factors. In this case, the calculated observation crosstalk influence coefficient cannot truly reflect the crosstalk relationship between channels. If it is used to update the crosstalk matrix, it will cause... Elements in the matrix deviate from their true values ​​or even exhibit directional errors, leading to erroneous actions in subsequent decoupling compensation. Learning freeze conditions include: changes in channel start / stop status within the window; changes in the number of channels within the window; the amplitude or rate of change of the common cooling source disturbance exceeding a preset severe disturbance threshold; any actuator being at the amplitude-limiting boundary or rate-limiting boundary for a cumulative time exceeding half the length of the preset stable operating condition window within the window; the sample variance of any channel's heat load change sequence or local temperature residual sequence within the window being lower than a preset minimum variance threshold, insufficient to support reliable correlation estimation; the number of effective samples within the window being lower than the preset minimum sample number; or the lower limit of the confidence interval corresponding to the observed crosstalk influence coefficient in step 200 is not greater than zero.

[0158] The preset conservative learning rate is greater than zero and less than the preset learning rate. It is determined during the equipment commissioning phase on typical operating condition samples with poor identification conditions, with the goal of preventing directional misjudgment of the crosstalk matrix after conservative update due to a single abnormal observation. During any of the above-mentioned learning freeze conditions, the controller does not use the observation results of the current window to update the liquid temperature difference crosstalk matrix, or only performs conservative updates according to the preset conservative learning rate, in order to avoid residual changes caused by factors such as controller and load self-recovery, disappearance of common disturbances, or release of actuator limiting being incorrectly attributed to a certain crosstalk influence coefficient. After the freeze condition is completely lifted, the controller re-enters the crosstalk identification window and re-executes steps 200 to 300. The preset severe disturbance threshold and the preset minimum variance threshold are both statistically obtained during the equipment commissioning phase under typical stable operating conditions and typical disturbance operating conditions, with the goal of reliably distinguishing between steady-state excitations that can be identified online and severe disturbances or weak excitations that are not suitable for online identification.

[0159] The updated return liquid temperature difference crosstalk matrix replaces the previous return liquid temperature difference crosstalk matrix in the next control cycle and participates in the calculation of steps 200 to 300, forming a closed-loop online adaptive learning process.

[0160] While updating the crosstalk matrix, the controller outputs the common disturbance level, main crosstalk source channels, main affected channels, crosstalk lag time, crosstalk influence intensity, crosstalk influence direction, and executed decoupling control actions for the current control cycle to the operation screen or host computer communication interface. The common disturbance level is determined by classifying the absolute value of the common cooling source disturbance and by the degree of deviation of the common cooling water side parameters from their preset stable historical averages. Specifically, a low common disturbance level is defined as the absolute value of the difference between the common cooling water side inlet temperature, pressure, and total flow rate and their respective preset stable historical averages, all of which do not exceed the product of the corresponding preset historical standard deviation and a preset multiple, and the common cooling water side is not in an alarm state. A low common disturbance level is defined as the absolute value of the difference between any of the inlet temperature, pressure, or total flow rate and its preset stable historical average, exceeding the product of the corresponding preset historical standard deviation and a preset multiple. The product of multiples, but before triggering the alarm state on the common cooling water side, is determined to be of medium common disturbance level; when any parameter of inlet temperature, pressure or total flow exceeds the corresponding equipment safety threshold, or when a sensor failure causes the common cooling water side to be in alarm state, it is determined to be of high common disturbance level; the main crosstalk source channel is determined by the channel with the largest absolute value of crosstalk source strength in step 300 and marked as the crosstalk source; the main disturbed channel is determined by the channel with the largest absolute value of disturbed strength in step 300 and marked as the disturbed channel; the crosstalk lag time corresponds to the maximum correlation lag time from the main crosstalk source channel to the main disturbed channel; the crosstalk influence intensity corresponds to the absolute value of the crosstalk influence coefficient from the main crosstalk source channel to the main disturbed channel; the crosstalk influence direction is determined by the sign of the crosstalk influence coefficient; the executed decoupling control action is the valve position, flow rate, pump frequency, and cooling and heating compensation adjustment content corresponding to the actual control quantities issued by each channel in step 400.

[0161] The controller continuously collects statistics on the above information. When a channel is repeatedly marked as a high-intensity crosstalk source within a preset continuous monitoring period, that is, when the absolute value of the crosstalk source intensity of the channel continuously exceeds the preset source intensity threshold within the preset continuous monitoring period, or when the absolute value of the local temperature residual of a disturbed channel cannot recover to within the preset residual stability upper limit within the preset continuous monitoring period after continuous decoupling compensation, the controller generates a maintenance prompt to prompt the inspection of channel flow, valve status, heat exchanger blockage, coolant contamination, or abnormal cooling water supply.

[0162] The reason for introducing the maintenance prompt mechanism is that, under normal circumstances, decoupled control can suppress the crosstalk source intensity and the local temperature residual of the disturbed channel to an acceptable level within a limited time. If a channel continues to exhibit high-intensity crosstalk or the residual cannot converge within a preset continuous monitoring period exceeding the system's normal response time, it indicates that the anomaly has exceeded the compensation capability of the control algorithm. The root cause may be a physical fault at the hardware level rather than a transient disturbance that can be eliminated by adjusting the actuator. In this case, it is necessary to prompt the operation and maintenance personnel to perform physical inspection and maintenance. The preset residual stability upper limit is determined during the equipment commissioning phase when each channel is in a steady-state operating condition and decoupled. Under normal compensation conditions, the steady-state fluctuation range of the absolute value of the local temperature residual of each channel is statistically analyzed and its upper envelope is taken to determine the level to which the local temperature residual of each channel should fall back after the decoupling compensation is normal. The preset continuous monitoring duration is determined by the equipment debugging stage based on the longest time required for the system to fully take effect and restore steady state under normal operating conditions from the occurrence of crosstalk. Its value is greater than the time required for the system to complete a complete decoupling compensation closed loop under the most unfavorable conditions, so as to ensure that maintenance prompts are only triggered when the decoupling compensation cannot eliminate the anomaly, and to avoid false alarms due to the transient process not yet being completed.

[0163] At this point, the controller completes the multi-channel return liquid temperature difference crosstalk decoupling control process for this control cycle. When entering the next control cycle, it returns to step 100, re-collects the temperature, flow rate, and cooling water side parameters of each channel, and repeats steps 100 to 500 to achieve continuous directional crosstalk identification and hierarchical decoupling control.

[0164] Example 2:

[0165] See Figure 3 As shown, a multi-channel return liquid temperature difference crosstalk decoupling control system is provided for executing the aforementioned multi-channel return liquid temperature difference crosstalk decoupling control method. The system includes:

[0166] The data acquisition and calculation module 101 acquires channel data and common cooling water side parameters of each channel according to a preset control cycle, calculates channel heat load based on channel data, calculates common cooling source disturbance based on common cooling water side parameters, and obtains local temperature residual based on channel data and common cooling source disturbance.

[0167] The temperature difference crosstalk module 102 calculates the crosstalk influence coefficient for each channel based on the channel heat load and local temperature residual within a preset lag range, and constructs the return liquid temperature difference crosstalk matrix based on the crosstalk influence coefficient.

[0168] The crosstalk identification module 103 calculates the crosstalk source strength and the disturbance strength of each channel based on the crosstalk matrix of the return liquid temperature difference, marks the crosstalk source channel and the disturbance channel based on the crosstalk source strength and the disturbance strength, and generates a common compensation amount, a source channel suppression amount and a disturbance channel compensation amount based on the marked crosstalk source channel and the disturbance channel.

[0169] The control execution module 104 superimposes the common compensation amount of each channel, the source channel suppression amount, and the disturbance channel compensation amount to form the final control command, and then sends the final control command to the actuator after performing amplitude limiting and rate limiting processing.

[0170] The online update module 105 updates the return liquid temperature difference crosstalk matrix after each preset stable operating condition window ends.

[0171] The embodiments of this application have been described above, but these embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments based on the guidance of these embodiments, and all of them are within the protection scope of these embodiments.

Claims

1. A multi-channel return liquid temperature difference crosstalk decoupling control method, characterized in that, Includes the following steps: According to the preset control cycle, the channel data and common cooling water side parameters of each channel are collected. The channel heat load is calculated based on the channel data, the common cooling source disturbance is calculated based on the common cooling water side parameters, and the local temperature residual is obtained based on the channel data and the common cooling source disturbance. Based on the channel heat load and local temperature residual of each channel, the crosstalk influence coefficient is calculated for each pair of channels within the preset lag range, and the return liquid temperature difference crosstalk matrix is ​​constructed based on the crosstalk influence coefficient. Based on the crosstalk matrix of the return liquid temperature difference, calculate the crosstalk source strength and the disturbance strength of each channel, mark the crosstalk source channel and the disturbance channel based on the crosstalk source strength and the disturbance strength, and generate the common compensation amount, the source channel suppression amount and the disturbance channel compensation amount based on the marked crosstalk source channel and the disturbance channel. The common compensation amount of each channel, the source channel suppression amount, and the disturbance channel compensation amount are superimposed to form the final control command. The final control command is then subjected to amplitude limiting and rate limiting processing before being sent to the actuator. The return liquid temperature difference crosstalk matrix is ​​updated after each preset stable operating condition window ends.

2. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 1, characterized in that, Methods for calculating disturbances from common cooling sources include: Subtract the current values ​​of the inlet temperature, pressure, and total flow rate of the cooling water in the common cooling water side parameters from their respective preset stable historical average values. Divide the difference by the sum of the corresponding preset historical standard deviation and preset small positive number to obtain their respective standardized deviations. The standardized deviations of inlet temperature, pressure, and total flow rate are multiplied by their respective preset common disturbance regression coefficients and then summed to obtain the common cooling source disturbance for the current control cycle.

3. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 2, characterized in that, Methods for obtaining local temperature residuals include: For each channel, the temperature deviation is obtained by subtracting the set temperature from the current channel data. Based on the preset common cooling source disturbance dynamic response coefficient sequence of the channel, and the common cooling source disturbance amount corresponding to the current control cycle and the previous control cycle, a weighted sum is performed to obtain the common disturbance cumulative impact amount of the channel. The local temperature residual of the channel at the current moment is obtained by subtracting the cumulative impact of common disturbances from the temperature deviation of the channel at the current moment.

4. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 1, characterized in that, Methods for constructing the return liquid temperature difference crosstalk matrix include: The change in channel heat load is obtained by calculating the difference between the current channel heat load and the previous control cycle. For any channel pair consisting of a source channel and a target channel, the preset lag time candidate values ​​are traversed within the preset lag range. The correlation coefficient is calculated by aligning the channel heat load change sequence of the source channel with the local temperature residual sequence of the target channel according to each preset lag time candidate value. The absolute values ​​of each correlation coefficient are normalized, and the maximum value after normalization is taken as the crosstalk influence amplitude. The crosstalk influence coefficient of the channel pair is obtained by combining the positive and negative signs of the corresponding correlation coefficients. The crosstalk influence coefficient of the channel pair is corrected based on the crosstalk influence amplitude; The crosstalk influence coefficients for all channel pairs are arranged according to the source channel and target channel numbers to form a return liquid temperature difference crosstalk matrix.

5. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 4, characterized in that, The method for correcting the crosstalk influence coefficient of the channel pair includes: The number of valid samples is determined based on the operating status of the source and target channels; Based on the influence of the channel on the number of effective samples and the amplitude of crosstalk within a preset sliding time window, estimate the lower limit of the confidence interval of the channel on the absolute value of the correlation coefficient; When the crosstalk effect amplitude is greater than the preset crosstalk judgment threshold, the lower limit of the confidence interval is greater than zero, and the number of effective samples is greater than or equal to the preset minimum number of samples, the crosstalk effect coefficient of the channel pair is retained; otherwise, the crosstalk effect coefficient of the channel pair is set to zero.

6. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 5, characterized in that, Methods for marking crosstalk source channels and scrambled channels include: Based on the crosstalk influence coefficients of channel i as the source channel and the change in channel heat load, the crosstalk source strength of channel i is calculated. Based on the crosstalk amplitude between all channels j (excluding channel i) and channel i, determine the maximum correlation lag time when channel j is the source channel and channel i is the target channel; The disturbance intensity of channel i is calculated by taking channel j as the source channel and channel i as the target channel, taking the crosstalk influence coefficient, maximum correlation lag time, and heat load change. When the absolute value of the crosstalk source strength of channel i is greater than the preset source strength threshold and the causal timing lead condition is met, channel i is marked as a crosstalk source channel; channels whose absolute value of the disturbance strength is greater than the preset disturbance strength threshold and which are not marked as crosstalk source channels are marked as disturbed channels.

7. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 6, characterized in that, Methods for determining whether a causal time priority condition is met include: For each channel j other than channel i, determine the maximum correlation lag time of channel i as the source channel and channel j as the target channel. The timing of the change is determined based on the absolute value of the change in channel heat load or the local temperature residual. If the time when the channel heat load change of channel i is triggered is earlier than the time when the local temperature residual of channel j is triggered in the corresponding direction, and the difference between the time interval between the two and the maximum correlation lag time is within the preset time tolerance range, then channel i is considered to satisfy the causal time lead relationship with channel j. When channel i satisfies the above causal temporal lead relationship with at least one channel j other than itself, channel i is considered to satisfy the causal temporal lead condition.

8. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 6, characterized in that, Methods for generating common compensation, source channel suppression, and disturbed channel compensation include: The maximum value of the absolute value of the crosstalk source strength of all crosstalk source channels in the current control cycle and the maximum value of the absolute value of the disturbance strength of all disturbed channels are taken as the representative crosstalk strength. If the absolute value of the common cooling source disturbance is greater than the product of the crosstalk representative strength and the preset dominance judgment multiple, it is determined that the cooling source is dominant, and the common compensation amount is obtained by multiplying the preset common disturbance compensation gain with the common cooling source disturbance. Otherwise, if the crosstalk strength is greater than the smaller of the preset source strength threshold and the preset disturbance strength threshold, then crosstalk is determined to be dominant. The preset source channel suppression gain is multiplied by the crosstalk source strength of the crosstalk source channel to obtain the source channel suppression amount; the preset disturbance channel compensation gain is multiplied by the disturbance strength of the disturbance channel to obtain the disturbance channel compensation amount. The compensation amount not enabled by the current dominant type is zero.

9. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 8, characterized in that, Methods for sending final control commands to actuators after amplitude and rate limiting include: The basic feedback control quantity is generated based on the temperature deviation of the channel; The final control command is calculated based on the basic feedback control quantity of the channel, the compensation quantity of the disturbed channel, the suppression quantity of the source channel, and the common compensation quantity. The final control command is decomposed into actuator control quantities by using a preset control allocation matrix; The control quantities of each actuator are sequentially subjected to amplitude limiting and rate limiting processing to obtain the actual control quantities issued. The actual control input is sent to the actuator.

10. The multi-channel return liquid temperature difference crosstalk decoupling control method according to claim 1, characterized in that, The method for updating the return liquid temperature difference crosstalk matrix includes: After each preset stable operating condition window ends, the observation crosstalk influence coefficient of each channel pair is recalculated based on the channel heat load change sequence and local temperature residual sequence within that window. For each crosstalk influence coefficient in the return liquid temperature difference crosstalk matrix, the value obtained by subtracting the preset learning rate is multiplied by the crosstalk influence coefficient before the update, and then the product of the preset learning rate and the recalculated observed crosstalk influence coefficient is added. The sum is used as the updated crosstalk influence coefficient. After all crosstalk influence coefficients are updated, the updated return liquid temperature difference crosstalk matrix is ​​obtained. The updated return liquid temperature difference crosstalk matrix replaces the previous return liquid temperature difference crosstalk matrix in the calculation during the next control cycle.

11. A multi-channel return liquid temperature difference crosstalk decoupling control system, characterized in that, A system for executing a multi-channel return liquid temperature difference crosstalk decoupling control method as described in any one of claims 1-10; the system includes: The data acquisition and calculation module collects channel data and common cooling water side parameters of each channel according to a preset control cycle. It calculates the channel heat load based on the channel data, calculates the common cooling source disturbance based on the common cooling water side parameters, and obtains the local temperature residual based on the channel data and the common cooling source disturbance. The temperature difference crosstalk module calculates the crosstalk influence coefficient for each channel based on the channel heat load and local temperature residual within a preset lag range, and constructs the return liquid temperature difference crosstalk matrix based on the crosstalk influence coefficient. The crosstalk identification module calculates the crosstalk source strength and the disturbance strength of each channel based on the crosstalk matrix of the return liquid temperature difference, marks the crosstalk source channel and the disturbance channel based on the crosstalk source strength and the disturbance strength, and generates a common compensation amount, a source channel suppression amount and a disturbance channel compensation amount based on the marked crosstalk source channel and the disturbance channel. The control execution module superimposes the common compensation amount of each channel, the source channel suppression amount, and the disturbance channel compensation amount to form the final control command. After performing amplitude limiting and rate limiting processing on the final control command, it sends it to the actuator. The online update module updates the return liquid temperature difference crosstalk matrix after each preset stable operating condition window ends.

Images