Cooling tower constant pressure PID self-adaptive control method and system based on process monitoring

Abstract

The application provides a cooling tower constant pressure PID adaptive control method and system based on process monitoring, and relates to the technical field of PID control.The method comprises the following steps: decoupling a working condition characteristic index group and a deviation feedback index group; constructing an operating condition vector and matching an initial operating scene group; calling a historical state transition chain, performing scene membership probability correction based on scene state transition prediction, and outputting a real-time operating scene; performing disturbance-free smooth coarse adjustment of the cooling tower constant pressure control; executing fine-grained PID fine adjustment of the cooling tower constant pressure; and performing scene switching analysis under incremental monitoring until the operating scene is changed, triggering fine-grained PID fine adjustment under disturbance-free smooth coarse adjustment. The application solves the technical problem that the prior art usually performs control parameter setting based on single or low-dimensional characteristics, cannot accurately identify the current operating condition category, and thus leads to PID parameter update lag and difficulty in timely matching the actual system dynamic characteristics when the operating condition rapidly changes.

Description

Technical Field

[0001] This invention relates to the field of PID control technology, and specifically to a constant-pressure PID adaptive control method and system for cooling towers based on process monitoring. Background Technology

[0002] Cooling tower constant pressure control systems are widely used in HVAC and industrial circulating water systems. Their core objective is to maintain stable water supply pressure under different ambient temperatures, load changes, and hydraulic disturbances. Traditional control methods typically employ fixed-parameter PID control or empirically tuned piecewise PID control, which adjusts pressure error through proportional, integral, and derivative actions. However, this single fixed-parameter PID control is difficult to adapt to complex operating environments over long periods.

[0003] Specifically, existing technologies typically tune control parameters based on single or low-dimensional features, such as pressure deviation or flow rate changes. They lack the ability to structure and classify multi-dimensional operating conditions, making it impossible to accurately identify the current operating condition category. This results in control parameter adjustments being based on local information rather than the global state. Furthermore, when operating conditions change rapidly, PID parameter updates lag behind, making it difficult to match the actual dynamic characteristics of the system in a timely manner. Summary of the Invention

[0004] This application provides a constant-pressure PID adaptive control method and system for cooling towers based on process monitoring. It aims to solve the technical problem that existing technologies typically tune control parameters based on single or low-dimensional features, which cannot accurately identify the current operating condition category. As a result, when the operating condition changes rapidly, the PID parameter update lags and it is difficult to match the dynamic characteristics of the actual system in a timely manner.

[0005] The first aspect disclosed in this application provides a constant-pressure PID adaptive control method for cooling towers based on process monitoring. The method includes: decoupling a set of operating condition characteristic indicators and a set of deviation feedback indicators from a process monitoring data pool; constructing an operating condition vector based on the set of operating condition characteristic indicators and matching an initial operating scenario group in a scenario mode library; using the acquisition timestamp of the set of operating condition characteristic indicators as the backtracking starting point, retrieving the historical state transition chain of a preset time window, and performing scenario membership probability correction on the initial operating scenario group based on scenario state transition prediction to output a real-time operating scenario; matching benchmark PID control parameters according to the real-time operating scenario to perform non-disturbance smooth coarse adjustment of the cooling tower constant-pressure control; using the benchmark PID control parameters as the online control starting point, executing fine-grained PID fine adjustment of the cooling tower constant-pressure control driven by the deviation feedback indicator group; performing scenario switching analysis under incremental monitoring on the set of operating condition characteristic indicators until the operating scenario changes, triggering a fine-grained PID fine adjustment loop under non-disturbance smooth coarse adjustment.

[0006] The second aspect of this application discloses a cooling tower constant pressure PID adaptive control system based on process monitoring. The system is used in the aforementioned cooling tower constant pressure PID adaptive control method based on process monitoring. The system includes: an index decoupling module for decoupling operating condition characteristic index groups and deviation feedback index groups from a process monitoring data pool; a scenario matching module for constructing an operating condition vector based on the operating condition characteristic index groups and matching an initial operating scenario group in a scenario pattern library; and a probability correction module for retrieving historical state transition chains within a preset time window, using the collection timestamp of the operating condition characteristic index groups as the backtracking starting point, and correcting the initial... The initial running scenario group performs scenario membership probability correction based on scenario state transition prediction and outputs the real-time running scenario; the smoothing coarse adjustment module is used to perform non-disturbance smoothing coarse adjustment of cooling tower constant pressure control according to the benchmark PID control parameters matched with the real-time running scenario; the PID fine adjustment module is used to perform fine-grained PID fine adjustment of cooling tower constant pressure driven by the deviation feedback index group, with the benchmark PID control parameters as the starting point of online control; the fine adjustment loop module is used to perform scenario switching analysis under incremental monitoring of the operating condition characteristic index group until the running scenario changes, triggering a fine-grained PID fine adjustment loop under non-disturbance smoothing coarse adjustment.

[0007] One or more technical solutions provided in this application have at least the following beneficial effects: By decoupling process monitoring data into a set of operating condition characteristic indicators and a set of deviation feedback indicators, the structural separation of environmental driving information and control error information is achieved, enabling the system to possess dual-channel capabilities of feedforward identification and feedback adjustment. By matching the operating condition vector with a scenario pattern library, continuous monitoring data is mapped to discrete semantic scenarios, upgrading the control system from numerical control to scenario control, significantly improving the ability to express operating conditions and the transferability of control strategies. Introducing a historical state transition chain based on timestamp backtracking and constructing a dynamic empirical transition matrix ensures that scenario identification results no longer rely on single-point operating condition matching but integrate historical evolution patterns. Through a probability correction mechanism, the influence of instantaneous noise or abnormal data on scenario judgment can be effectively suppressed, improving the stability and consistency of scenario identification, achieving an upgrade from static matching to dynamic prediction. Through a multi-source PID parameter fusion mechanism between real-time running scenarios and K candidate scenarios, multi-model one-to-one control parameter fusion is achieved. Consistency correction avoids PID abrupt changes caused by misjudgment in a single scenario. Through probability weighting and intersection consistency constraints, parameter switching is made gradual and continuous, thereby significantly reducing control shock and improving the system's stability and safety during scenario switching. A feedback adjustment mechanism is built based on pressure deviation and its rate of change, enabling PID parameters to have real-time dynamic correction capabilities. At the same time, pressure condition labels are introduced to realize differentiated adjustment strategies under different pressure ranges, allowing control behavior to adaptively adjust with the operating range, thereby effectively suppressing overshoot, reducing oscillations, accelerating convergence speed, and improving local control accuracy. Through incremental monitoring of operating condition characteristic index groups, continuous detection and dynamic updates of operating scenarios are realized. When structural changes occur in the operating conditions, coarse and fine adjustment cycles are automatically triggered, enabling the system to continuously adapt to operating condition drift and load changes in long-term operation, avoiding control strategy failure or lag, and realizing full life cycle adaptive control.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0009] Figure 1 A schematic flowchart of a constant-pressure PID adaptive control method for cooling towers based on process monitoring, provided in an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of the pressure response curve change in the constant pressure PID adaptive control method for cooling towers based on process monitoring provided in the embodiments of this application.

[0011] Figure 3 A schematic diagram of the structure of a cooling tower constant pressure PID adaptive control system based on process monitoring provided in this application embodiment.

[0012] Figure labeling: Index decoupling module 10, Scene matching module 20, Probability correction module 30, Smoothing coarse adjustment module 40, PID fine adjustment module 50, Fine adjustment loop module 60. Detailed Implementation

[0013] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structure, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0014] Example 1, as Figure 1 As shown in the embodiments of this application, a constant-pressure PID adaptive control method for cooling towers based on process monitoring is provided. The method includes: Decouple the operating condition characteristic index group and the deviation feedback index group from the process monitoring data pool.

[0015] Multi-source operational data is extracted in real time from the cooling tower operation monitoring data pool and decoupled structurally. The monitoring data pool includes multi-dimensional time-series data such as temperature, flow rate, pressure, and equipment status. Through preset data field mapping rules, variables reflecting changes in the external operating environment and load are divided into operating condition characteristic index groups, such as wet-bulb temperature, water supply flow rate, and temperature difference changes. Variables directly reflecting the degree of deviation from the control target are divided into deviation feedback index groups, including real-time pressure at the controlled point, pressure deviation, and its rate of change. Operating condition characteristics describe the system's external operating state, while deviation feedback characterizes the dynamics of control errors. This decoupling process achieves structural separation of feedforward and feedback information, providing different information channels to support subsequent scenario identification and control adjustment.

[0016] Based on the aforementioned set of operating condition characteristic indicators, an operating condition vector is constructed, and an initial operating scenario group is matched in the scenario pattern library.

[0017] The decoupled operating condition characteristic index group is standardized and normalized to construct an operating condition vector representing the current operating state. This vector uniformly describes the current external load and environmental combination characteristics of the cooling tower. This vector is then input into a pre-constructed scenario pattern library for matching. This scenario pattern library is formed from historical operating experience or expert rules, with each scenario corresponding to a set of typical operating condition characteristic combinations. The matching process adopts a hierarchical scenario model tree structure. Through coarse-grained similarity filtering and fine-grained feature consistency verification, multiple candidate scenario sets that meet the conditions are selected, forming the initial operating scenario group. Here, the scenario pattern library represents a typical operating condition classification knowledge base, and the operating condition vector uniformly expresses the current system state mathematically, thereby achieving a mapping from data to semantic scenarios.

[0018] Using the collection timestamp of the aforementioned working condition characteristic index group as the starting point for backtracking, the historical state transition chain of the preset time window is retrieved, and the initial running scenario group is corrected based on the scenario state transition prediction scenario membership probability, and the real-time running scenario is output.

[0019] Using the current operating condition feature data collection timestamp as the time anchor, a continuous historical state transition chain is constructed by tracing back the operation records within a preset time window from historical data. This chain represents the actual switching path between scenarios. Based on this transition chain, the transition frequency between different scenarios is statistically analyzed, and a dynamic empirical transition matrix is ​​constructed to describe the probabilistic migration relationship between scenarios. Furthermore, each candidate scenario in the initial operating scenario group is input into this transition matrix, and its membership probability in the current time context is calculated. The original matching results are then subjected to Bayesian probability correction. This correction process essentially introduces time-related constraints, ensuring that scenario judgment not only depends on the similarity of the current operating condition but also incorporates historical evolution trends, thereby outputting the real-time operating scenario with the highest probability and achieving more stable scenario recognition results.

[0020] Based on the real-time operating scenario, the baseline PID control parameters are matched to perform a smooth, non-disruptive coarse adjustment of the cooling tower constant pressure control.

[0021] After obtaining the real-time operating scenario, the baseline PID control parameters corresponding to that scenario are extracted from a preset PID parameter mapping table. These baseline parameters represent the empirically optimal control combination for long-term stable operation under that scenario. K high-probability candidate scenarios and their corresponding PID parameters are introduced, and the parameters are fused and corrected using a probability-weighted method to form corrected PID control parameters. In this process, perturbation-free smooth coarse-tuning avoids abrupt changes in control output during parameter switching. This is achieved by imposing continuity constraints and rate-of-change limits on the proportional, integral, and derivative parameters to ensure a smooth transition of the control signal and prevent shocks to the pump or valve. This step essentially involves a global control strategy switch based on the scenario identification results, but the multi-candidate fusion mechanism reduces the risk of control fluctuations caused by misjudgments in a single scenario.

[0022] Using the aforementioned baseline PID control parameters as the starting point for online control, the constant pressure fine-grained PID fine-tuning of the cooling tower driven by the aforementioned deviation feedback index group is executed.

[0023] Based on the coarsely tuned PID parameters, this serves as the starting point for online control, initiating a feedback-driven fine-tuning phase. This phase uses a set of deviation feedback indicators as the core input. The controlled point pressure deviation reflects the control error amplitude, and the deviation change rate reflects the dynamic trend of the error. Simultaneously, real-time pressure values ​​are combined to determine the current pressure operating range, used to distinguish different load intervals. Through a preset fine-tuning rule library, the deviation-change rate is mapped to the incremental adjustment of the PID parameters. This increment is then biased and corrected according to the pressure operating range, adapting the adjustment strategy to the characteristics of different operating intervals. The increment is limited and smoothed to avoid high-frequency oscillations, and finally superimposed on the current PID parameters to achieve fine-grained dynamic optimization, thereby improving pressure stability and response speed within a local range.

[0024] The scenario switching analysis under incremental monitoring is performed on the aforementioned working condition characteristic index group until the operating scenario changes, triggering a fine-grained PID fine-tuning loop under non-disruptive smooth coarse adjustment.

[0025] During continuous operation of the control system, a sliding window incremental monitoring method is used to monitor the operating condition characteristic index group. By detecting the changing trends of key indicators such as ambient wet-bulb temperature, flow rate changes, and load feedback, it is determined whether the current system is experiencing operating condition drift or structural changes. If the operating condition change exceeds a preset threshold, a scenario switching analysis mechanism is triggered, remapping the current operating condition vector to the scenario mode library to determine if a new optimal matching scenario has emerged. When a change in the operating scenario is detected, the system re-enters the coarse-tuning process, re-executes PID parameter matching and bumpless switching, while retaining the current fine-tuning state as the initial condition, achieving a cyclical connection between coarse and fine-tuning. This mechanism forms an event-driven closed-loop control structure, enabling the system to maintain stability and adaptability during long-term changes in operating conditions.

[0026] Furthermore, the method of constructing a working condition vector based on the aforementioned working condition feature index group and matching an initial working scenario group in a scenario pattern library includes: Multiple combinations of operational feature conditions for various typical operational scenarios are predefined; based on feature universality sorting, the multiple combinations of operational feature conditions are hierarchically divided to construct a scenario hierarchical model tree, and the multiple typical operational scenarios are associated with multiple leaf nodes of the scenario hierarchical model tree; the operational condition vector is mapped to the scenario hierarchical model tree, and joint pruning under hierarchical progressive search is performed to obtain a matching node tree; the typical operational scenarios associated with multiple matching nodes are extracted from the matching nodes to form the initial operational scenario group.

[0027] Based on long-term historical operating data of cooling towers, expert experience, and control target constraints, several typical operating scenarios are predefined, each corresponding to a set of operating characteristic condition combinations. These combinations characterize the discrimination boundaries of the scenarios, such as high-temperature high-load conditions, low-temperature low-load conditions, and fluctuating load transition conditions. Each combination consists of several key feature thresholds or ranges, including wet-bulb temperature range, supply and return water temperature difference range, flow rate level, and load feedback level. Essentially, these operating characteristic condition combinations serve as a standardized description template for the scenarios, replacing single numerical judgments and improving the structure and scalability of scenario expression. This step forms a basic scenario knowledge set, providing standardized input for subsequent hierarchical organization and rapid matching.

[0028] The generality of features is evaluated for multiple predefined combinations of operational features. Feature generality refers to the degree of sharing and distinguishing ability of a feature across different scenarios. For example, wet-bulb temperature is a highly generalized feature, while local load mutation is a low-generalized feature. Based on this ranking, features are progressively divided from highly generalized to low-generalized, constructing a multi-level scenario hierarchy model tree. Upper-level nodes represent macroscopic operational state categories, while lower-level nodes are progressively refined into specific operating condition branches. Specific typical operational scenarios are associated with the leaf nodes of this tree structure, making each leaf node correspond to one or more executable control strategy templates. This structure achieves a coarse-to-fine scenario representation organization, providing a hierarchical index foundation for subsequent rapid pruning and matching.

[0029] The real-time constructed runtime condition vectors are input into the scene hierarchical model tree for mapping and matching. First, coarse-grained similarity calculations are performed at high-level nodes, using rapid distance metrics or rule-based filtering to eliminate obviously mismatched branches, achieving the first round of pruning. Then, in intermediate-level nodes, further filtering is performed based on multi-dimensional feature consistency, jointly pruning branches that do not meet the constraints to reduce the search space. Finally, fine-grained matching verification is performed at low-level nodes, retaining the set of node paths that satisfy the multi-dimensional feature constraints, forming a matching node tree. Joint pruning refers to simultaneously using multiple feature conditions to collaboratively eliminate branches that do not meet the conditions, rather than filtering step-by-step using a single feature, thereby improving matching efficiency and reducing the risk of misjudgment propagation.

[0030] After hierarchical progressive search and pruning, the resulting matching node tree contains multiple end nodes or leaf node paths that meet the conditions. Typical operating scenarios associated with these matching nodes are extracted; each node may correspond to one or more historical or predefined scenario labels. Since a single operating condition vector may exhibit ambiguity or overlap in actual operation, a unique scenario is not directly selected. Instead, multiple candidate typical operating scenarios are retained to form an initial operating scenario group. This scenario group serves as the input basis for subsequent probability correction and PID parameter matching, providing the control strategy with a certain degree of redundancy. This avoids control strategy deviations due to single-point mismatches, thereby improving the overall system stability and adaptability.

[0031] Furthermore, the method involves mapping the operating condition vector to the scene hierarchical model tree, performing joint pruning under hierarchical progressive search, and obtaining a matching node tree. The operating condition vector is mapped to the scene hierarchy model tree, and coarse-grained node filtering is performed under breadth-first search to prune the scene hierarchy model tree as candidate subtrees; fine-grained node verification is performed on the candidate subtrees under depth-first search of the operating condition vector, and the candidate subtrees are pruned to form the matching node tree.

[0032] Using the current operating condition vector as input features, a breadth-first search traversal is performed on the scene hierarchy model tree. This process starts from the root node and expands layer by layer, performing a rapid consistency evaluation on all nodes within the same layer. The evaluation is based on the coarse-grained similarity between the operating condition vector and the combination of scene feature conditions represented by the node, including coverage of key environmental indicators, load level matching, and traffic trend consistency. When a node at a certain layer does not meet the basic matching threshold, its subtree branch is directly removed, thus achieving large-scale pruning and reducing subsequent computational complexity. Through this coarse-grained screening process, the original scene hierarchy model tree is compressed into candidate subtrees containing only potential matching regions, reducing the search space from a global tree structure to a locally highly relevant substructure, improving the efficiency of subsequent fine-grained matching.

[0033] After constructing the candidate subtree, a depth-first search traversal is performed to verify each branch of potential matching paths in depth. This stage uses the multidimensional consistency of the combination of runtime vectors and node feature conditions as the core evaluation criterion. Based on coarse-grained screening, fine-grained feature constraints are further introduced, including continuous variable deviation, feature weight matching degree, and the degree of fulfillment of combination rules. During the depth-first search, the search expands downwards layer by layer along a single branch, and a review is performed at each node. Once a node that does not meet the fine-grained constraints is found, the subsequent paths of that branch are backtracked and pruned. Through this fine-grained review mechanism, the candidate subtree is further shrunk and optimized, ultimately retaining the set of nodes that meet the complete multidimensional constraints, forming a matching node tree, thus transforming from "potentially relevant" to "high-confidence matching."

[0034] Furthermore, taking the collection timestamp of the aforementioned working condition characteristic index group as the starting point for backtracking, the historical state transition chain of a preset time window is retrieved, and the initial running scenario group is corrected based on the scenario state transition prediction to determine the scenario membership probability, thereby outputting the real-time running scenario. The method includes: Based on the scene switching frequency, the preset time window is dynamically adjusted; using the collection timestamp as the backtracking starting point, a backtracking interval is constructed in combination with the preset time window, and the historical state transition chain is retrieved from the circular log buffer; multiple typical running scenarios are enumerated to construct multiple scene jump state pairs, and the frequency of the historical state transition chain is traversed to obtain multiple empirical transition probabilities, thus constructing a dynamic empirical transition matrix; the initial running scenario group is mapped to the dynamic empirical transition matrix to obtain a scene membership probability group; the initial running scenario group is sorted in descending order according to the scene membership probability group to locate the real-time running scenario with high confidence.

[0035] By monitoring the number of scene changes per unit time in real time, a scene switching frequency index is generated and used as the basis for adjusting the time window. When the scene switching frequency is high, it indicates that the system is in a rapidly fluctuating condition. In this case, the time window is shortened to enhance the sensitivity to short-term state changes. When the scene switching frequency is low, the time window is appropriately expanded to cover state evolution information over a longer period, thereby improving statistical stability. The preset time window is used to limit the scope of historical data backtracking. Its dynamic adjustment mechanism ensures that subsequent state transition modeling neither over-relies on short-term noise nor ignores long-term trends, achieving time scale adaptation.

[0036] Using the current operating condition characteristic data collection timestamp as the time reference point, and combining it with a dynamically adjusted preset time window length, a backtracking interval is formed by tracing back along the historical timeline. This backtracking interval is used to limit the range of historical data participating in the statistics. In the system's circular log buffer, the sequence of continuously recorded scene state changes within this interval is extracted in chronological order to construct a historical state transition chain. This transition chain describes the continuous evolution path of the cooling tower's operating scene in the time dimension, including the scene ID and its adjacent jump relationships. Through this process, the scene identification results at discrete time points are transformed into a continuous state sequence, providing basic data support for subsequent transition probability modeling.

[0037] Multiple typical operating scenarios are combined and enumerated to construct all possible scenario transition state pairs, i.e., the set of directed state pairs from the source scenario to the target scenario. The historical state transition chain is traversed, and each actual scenario transition is matched and counted, calculating the frequency of each state pair in historical data. Based on this, the frequency is normalized to obtain the empirical transition probabilities between different scenarios, forming a dynamic empirical transition matrix. This matrix essentially describes the actual transition tendency between operating scenarios within the current backtracking window, thus introducing temporal correlation and system inertia information, enabling scenario recognition to no longer rely solely on instantaneous features but to integrate historical evolution patterns.

[0038] Each candidate scenario in the initial running scenario group obtained in the previous steps is used as an input index and mapped to the dynamic empirical transition matrix for probability querying and updating. Specifically, based on the current state and combined with historical transition probabilities, the probability of each candidate scenario occurring in the current system evolution context is calculated, thus obtaining the corresponding scenario membership probability value. This process is equivalent to further incorporating the working condition similarity matching results into the time transition prior correction, so that the credibility of each candidate scenario depends not only on the current feature matching degree but also on the degree of support from historical transition paths, ultimately forming a scenario membership probability group.

[0039] After determining the probability group to which a scene belongs, the candidate scenes in the initial running scene group are sorted in descending order according to their corresponding probability values. A higher probability indicates a stronger matching reliability of the scene under the combined effects of the current operating conditions and historical evolution. The scene with the highest probability is selected as the current real-time running scene for subsequent PID parameter matching and control strategy generation. This step achieves the final decision output from multiple candidate scenes through a probability ranking mechanism. Compared to a single-rule matching method, this effectively reduces the risk of misjudgment and enhances the stability of scene recognition results under complex operating condition fluctuations.

[0040] Furthermore, based on the benchmark PID control parameters matched to the real-time operating scenario, a non-disruptive, smooth coarse adjustment of the cooling tower constant pressure control is performed. The method includes: The scenario membership probability group is filtered based on a preset membership probability threshold to retrieve K candidate operating scenarios from the initial operating scenario group; the baseline PID control parameters and the K candidate PID control parameters are retrieved from the scenario parameter mapping table using the scenario IDs of the real-time operating scenario and the K candidate operating scenarios as indices; the K scenario membership probabilities of the K candidate operating scenarios are used as contribution coefficients to perform intersection consistency correction on the K candidate PID control parameters and the baseline PID control parameters to obtain corrected PID control parameters; the corrected PID control parameters are used for disturbance-free smooth coarse adjustment of the cooling tower constant pressure control.

[0041] Based on scenario membership probability groups, a preset membership probability threshold is set as a screening threshold to distinguish between high-confidence and low-confidence scenarios. All candidate scenarios are filtered according to their membership probabilities, retaining only those exceeding the threshold. Then, considering quantity constraints, the top K candidate scenarios are selected. These K scenarios are not simply ranked results, but rather subsets that simultaneously satisfy probability significance and scenario coverage, ensuring diversity and representativeness in subsequent PID fusion. The membership probability threshold controls the confidence range of scenarios participating in fusion, and the K candidate scenarios form the basis for multi-model parameter fusion, thereby avoiding control offset caused by misjudgment of a single scenario.

[0042] Using the currently determined real-time operating scenario as the primary index, the corresponding baseline PID control parameters are retrieved from the scenario parameter mapping table. These parameters represent the combination of the three PID parameters (proportional, integral, and derivative) after historical stable operation or optimized tuning under that scenario. Using the selected K candidate operating scenario IDs as indexes, the corresponding candidate PID control parameter sets are simultaneously queried. The scenario parameter mapping table is essentially a "scenario-control parameter" relational database, used to achieve a direct mapping from operating condition semantics to control law parameters.

[0043] The membership probabilities of K candidate operating scenarios are used as weighted contribution coefficients and introduced into the PID parameter fusion process. First, the difference between each candidate PID parameter and the baseline PID parameter is calculated, and the difference is weighted and constrained based on the membership probabilities. Under the intersection consistency correction mechanism, only the overlapping portions of multiple candidate parameters and the baseline parameter within the same direction of change or the same constraint range are retained, suppressing parameter offsets with large dispersion, thereby avoiding the influence of a few abnormal candidate scenarios on the overall parameters. The core of this mechanism is to use the baseline PID parameter as a stable anchor point, and through the dual constraints of probability weighting and consistency screening, ensure that the fusion result can absorb information from multiple scenarios without causing control strategy instability due to excessive scenario dispersion, ultimately obtaining smooth and reliable corrected PID control parameters.

[0044] The obtained corrected PID control parameters are used as new coarse-tuning parameters for the current control system to update the PID control law of the cooling tower constant pressure control loop. During parameter switching, a smooth, disturbance-free switching mechanism is introduced. By limiting the rate of change of the PID parameters and combining this with output continuity constraints, sudden changes in control quantities are prevented from impacting actuators such as water pumps and valves. Simultaneously, during the coarse-tuning phase, priority is given to ensuring the overall stability and response consistency of the system, allowing the control system to smoothly transition from the old scenario parameters to the new fused parameter state. After this step is completed, the system enters a stable coarse-tuning operating state, providing a stable initial control benchmark for subsequent deviation feedback-driven fine-grained tuning.

[0045] In one specific embodiment, to verify the perturbation-free coarse-tuning effect of the PID based on multi-scenario fusion, an experimental model was performed using a cooling tower constant pressure system as the object. The baseline PID parameters were set as Kp0=1.20, Ki0=0.35, Kd0=0.08. K=3 candidate scenarios were selected, with membership probabilities of P1=0.62, P2=0.25, and P3=0.13, corresponding to PID parameters of S1:(1.35,0.40,0.10), S2:(1.10,0.30,0.06), and S3:(1.50,0.45,0.12). The weighted normalization fusion formula was used: W i =P i ·exp(-||Δθ i ||), and W ’ i =W i / ΣW i , where Δθ i This represents the Euclidean deviation between the candidate PID parameters and the baseline parameters. The calculated corrected contribution weight is approximately: W ’ 1 = 0.58, W ’ 2 = 0.27, W ’ 3 = 0.15. The final corrected PID parameters are calculated using the anchor-weighted model: Kp = Kp0 + Σ(W ’ i·(Kp i -Kp0)) yields: Kp≈1.29, Ki≈0.37, Kd≈0.086.

[0046] like Figure 2 As shown, in this embodiment, the pressure response curve changes are recorded: the initial system pressure is set to 0.50 MPa. The overshoot during traditional single-scene switching is approximately 8.6%, while using this method, the overshoot is reduced to 2.3%, and the steady-state time is shortened from 42 seconds to 28 seconds. The corresponding pressure change trend can be expressed as follows: the pressure response curve rises rapidly in the first 10 seconds, then enters a slow convergence range and stabilizes near approximately 0.5 MPa.

[0047] The results show that the fusion mechanism of "probability weighting plus parameter anchor consistency correction" can significantly reduce the impact of PID parameter switching and improve the dynamic stability and convergence speed of the cooling tower constant pressure system under fluctuating operating conditions.

[0048] Furthermore, the method involves using the K scenario membership probabilities of the K candidate operating scenarios as contribution coefficients to perform intersection consistency correction on the K candidate PID control parameters and the baseline PID control parameters, thereby obtaining the corrected PID control parameters. The method includes: Calculate the K parameter deviations of the K candidate PID control parameters relative to the benchmark PID control parameters; perform consistency fusion on the K scenario membership probabilities and the K parameter deviations to obtain K correction contribution weights; use the benchmark PID control parameters as the benchmark anchor point, and perform weighted correction on the K candidate PID control parameters according to the K correction contribution weights to obtain the corrected PID control parameters.

[0049] Using the obtained baseline PID control parameters as a reference anchor, differentiated calculations are performed on the PID parameters corresponding to each candidate operating scenario. Specifically, the K candidate PID control parameters are compared with the baseline PID control parameters item by item along three dimensions: proportional (Kp), integral (Ki), and derivative (Kd), to obtain the corresponding parameter deviation vector. This parameter deviation is used to characterize the degree of control strategy deviation of different candidate scenarios relative to the current real-time scenario; its magnitude reflects the direction and amplitude of control intensity adjustment. Through this step, the probabilistic differences at the scenario level are transformed into a quantifiable expression of control parameter deviation, providing a unified numerical basis for subsequent consistency fusion.

[0050] The membership probability of each candidate scenario is used as a confidence weight input, while the corresponding PID parameter deviation is introduced as an adjustment factor for consistency fusion calculation. Consistency fusion refers to not only considering the probability of scenario occurrence but also determining whether the parameter offset direction is consistent with the baseline control strategy. When a candidate scenario has a high probability and corresponds to a small parameter deviation, its contribution weight is enhanced; conversely, when a high probability is accompanied by a large discrete deviation, its weight is suppressed, thereby avoiding excessive influence of extreme scenarios on the control results. Through this process, K corrected contribution weights are obtained, enabling a unified expression of probability information and control parameter consistency information.

[0051] Using the baseline PID control parameters as a stability anchor point—the control center where the system is currently operating reliably—a weighted correction mechanism of K candidate PID parameters is introduced. Specifically, the correction contribution weights are used as weighting coefficients to weight and sum the offsets of each candidate PID parameter relative to the baseline parameters, limiting the overall correction magnitude to ensure continuous and controllable changes in the final control parameters. The core of this mechanism lies in the "anchor point constraint," meaning all corrections revolve around the baseline PID, rather than directly replacing it, thus avoiding control strategy drift due to large dispersion in candidate scenarios. The resulting corrected PID control parameters possess both stability and adaptability, and can be used for subsequent smooth, disturbance-free control switching.

[0052] Furthermore, the operating condition characteristic index group consists of ambient wet-bulb temperature, cooling water supply main flow rate, first-order rate of change of flow rate, supply and return water temperature difference, terminal total load feedback signal, and instantaneous water replenishment flow rate; the deviation feedback index group consists of real-time pressure value of the controlled point, pressure deviation of the controlled point, and first-order rate of change of deviation.

[0053] The set of operating condition characteristic indicators characterizes the external operating environment and load status of the cooling tower system. Among them, wet-bulb temperature reflects the environmental thermal boundary conditions, water supply flow rate and its rate of change characterize the dynamic changes of system load, supply and return water temperature difference reflects the heat exchange efficiency status, terminal total load feedback describes the overall energy demand level, and instantaneous water replenishment flow rate reflects the changes in system water balance.

[0054] The deviation feedback index group describes the degree of deviation from the control target and its dynamic trend. The real-time pressure value of the controlled point reflects the actual control state of the current system, the pressure deviation measures the magnitude of the error between the control point and the set target, and the first-order rate of change of the deviation describes the error change trend and convergence speed, thereby providing dynamic feedback basis for PID fine-tuning.

[0055] Furthermore, using the aforementioned benchmark PID control parameters as the starting point for online control, the method of performing fine-grained PID tuning of the cooling tower under constant pressure driven by the deviation feedback index group includes: Based on the real-time pressure value of the controlled point in the deviation feedback index group, a pressure condition label is determined; using the pressure deviation and first-order change rate of the controlled point in the deviation feedback index group as input, the PID parameter increment is matched in the preset fine-tuning rule library; after biasing and superimposing the PID parameter increment according to the pressure condition label, amplitude limiting and smoothing processing is performed to obtain an effective fine-tuning increment; the effective fine-tuning increment is superimposed on the base PID control parameter to obtain the fine-tuned PID control parameter for constant pressure control of the cooling tower.

[0056] Using the real-time pressure value of the controlled point in the deviation feedback index group as the core state input, the current constant pressure operation state of the cooling tower is zoned. Multiple pressure condition zones are predefined, such as low-pressure recovery zone, stable operation zone, and high-pressure suppression zone, each corresponding to different control sensitivities and PID adjustment strategies. When the real-time pressure value enters a certain zone, the corresponding pressure condition zone label is automatically assigned. This label represents the current pressure operating range state of the control system, thus providing a zoning basis for subsequent PID incremental correction and achieving regionalized adaptive adjustment of the control strategy.

[0057] Pressure deviation (error amplitude) and the first-order rate of change of deviation (error trend) are used as two-dimensional dynamic input features, and matched against a preset fine-tuning rule base. This fine-tuning rule base is a mapping set constructed based on historical control experience or optimization results, and its core function is to establish a mapping relationship between error states and PID incremental adjustments. Pressure deviation reflects the degree of deviation from the current steady-state of the system, while the rate of change of deviation is used to determine the error convergence speed or divergence trend. By combining these two, different control states such as "rapidly approaching steady state," "oscillation correction," or "slow deviation" can be distinguished, thereby outputting the corresponding PID parameter increments and realizing the generation of a dynamic fine-tuning strategy.

[0058] Based on the PID increment output from the rule base, pressure condition labels are introduced for zoned bias correction. Different pressure condition bands correspond to different control sensitivity coefficients; for example, the integral action is enhanced in the low-pressure recovery zone, and proportional gain fluctuations are suppressed in the high-pressure suppression zone. This allows for directional or amplitude bias superposition of the original PID increment. The corrected increment is then limited to prevent parameter changes from exceeding the system's safety or stability thresholds. Furthermore, a smoothing filtering mechanism, such as moving average or first-order inertial filtering, is introduced to suppress high-frequency disturbances. Ultimately, an effective fine-tuned increment is obtained, ensuring it meets the control requirements of the current pressure range while avoiding abrupt changes in the control output.

[0059] The effective fine-tuning increment is used as a dynamic correction term and superimposed on the current baseline PID control parameters to form the final fine-tuned PID control parameters. This superposition process uses a continuous update mechanism in time, that is, a recursive update is performed in each control cycle, thereby achieving online adaptive adjustment of the parameters. The final PID parameters inherit the stability of the baseline parameters and integrate the dynamic correction capability of real-time deviation feedback, which is used to directly drive the constant pressure control loop of the cooling tower, achieving a unity of pressure stability control and dynamic response optimization.

[0060] Furthermore, if the matching result of the operating condition vector in the scenario mode library is an empty set, then the operating condition vector is used for fault mode matching to trigger a cooling tower fault alarm.

[0061] When the matching result of the operating condition vector in the scenario pattern library is an empty set, the system does not directly return a no-result state, but instead switches to fault identification mode. In this mode, the current operating condition vector is input into a preset fault pattern library and matched with predefined abnormal operating condition features, such as typical fault feature combinations like sensor drift, abnormal flow changes, or heat exchange efficiency failure. When the matching degree exceeds the fault threshold, it is determined to be the corresponding fault type, and the cooling tower fault alarm mechanism is triggered. At the same time, the fault occurrence timestamp and feature vector are recorded for subsequent diagnosis and maintenance analysis. This mechanism realizes the transformation from "failure to identify normal operating conditions" to "proactive identification of abnormal states," improving system safety and fault tolerance.

[0062] Example 2 is based on the same inventive concept as the cooling tower constant pressure PID adaptive control method based on process monitoring in the previous examples, such as... Figure 3 As shown in the embodiment of this application, a constant pressure PID adaptive control system for cooling towers based on process monitoring is provided. The system includes: The indicator decoupling module 10 is used to decouple the operating condition characteristic indicator group and the deviation feedback indicator group from the process monitoring data pool; the scenario matching module 20 is used to construct an operating condition vector based on the operating condition characteristic indicator group and match the initial operating scenario group in the scenario pattern library; the probability correction module 30 is used to retrieve the historical state transition chain of the preset time window with the collection timestamp of the operating condition characteristic indicator group as the backtracking starting point, and perform scenario membership probability correction based on scenario state transition prediction on the initial operating scenario group to output the real-time operating scenario; the smoothing coarse adjustment module 40 is used to perform non-disturbance smoothing coarse adjustment of the cooling tower constant pressure control according to the benchmark PID control parameters matched with the real-time operating scenario; the PID fine adjustment module 50 is used to execute the cooling tower constant pressure fine-grained PID fine adjustment driven by the deviation feedback indicator group with the benchmark PID control parameters as the online control starting point; the fine adjustment loop module 60 is used to perform scenario switching analysis under incremental monitoring of the operating condition characteristic indicator group until the operating scenario changes, triggering the fine-grained PID fine adjustment loop under non-disturbance smoothing coarse adjustment.

[0063] Furthermore, the scene matching module 20 is used to perform the following operation steps: Multiple combinations of operational feature conditions for various typical operational scenarios are predefined; based on feature universality sorting, the multiple combinations of operational feature conditions are hierarchically divided to construct a scenario hierarchical model tree, and the multiple typical operational scenarios are associated with multiple leaf nodes of the scenario hierarchical model tree; the operational condition vector is mapped to the scenario hierarchical model tree, and joint pruning under hierarchical progressive search is performed to obtain a matching node tree; the typical operational scenarios associated with multiple matching nodes are extracted from the matching nodes to form the initial operational scenario group.

[0064] Furthermore, the scene matching module 20 is used to perform the following operation steps: The operating condition vector is mapped to the scene hierarchy model tree, and coarse-grained node filtering is performed under breadth-first search to prune the scene hierarchy model tree as candidate subtrees; fine-grained node verification is performed on the candidate subtrees under depth-first search of the operating condition vector, and the candidate subtrees are pruned to form the matching node tree.

[0065] Furthermore, the probability correction module 30 is used to perform the following operation steps: Based on the scene switching frequency, the preset time window is dynamically adjusted; using the collection timestamp as the backtracking starting point, a backtracking interval is constructed in combination with the preset time window, and the historical state transition chain is retrieved from the circular log buffer; multiple typical running scenarios are enumerated to construct multiple scene jump state pairs, and the frequency of the historical state transition chain is traversed to obtain multiple empirical transition probabilities, thus constructing a dynamic empirical transition matrix; the initial running scenario group is mapped to the dynamic empirical transition matrix to obtain a scene membership probability group; the initial running scenario group is sorted in descending order according to the scene membership probability group to locate the real-time running scenario with high confidence.

[0066] Furthermore, the smoothing coarse adjustment module 40 is used to perform the following operation steps: The scenario membership probability group is filtered based on a preset membership probability threshold to retrieve K candidate operating scenarios from the initial operating scenario group; the baseline PID control parameters and the K candidate PID control parameters are retrieved from the scenario parameter mapping table using the scenario IDs of the real-time operating scenario and the K candidate operating scenarios as indices; the K scenario membership probabilities of the K candidate operating scenarios are used as contribution coefficients to perform intersection consistency correction on the K candidate PID control parameters and the baseline PID control parameters to obtain corrected PID control parameters; the corrected PID control parameters are used for disturbance-free smooth coarse adjustment of the cooling tower constant pressure control.

[0067] Furthermore, the smoothing coarse adjustment module 40 is used to perform the following operation steps: Calculate the K parameter deviations of the K candidate PID control parameters relative to the benchmark PID control parameters; perform consistency fusion on the K scenario membership probabilities and the K parameter deviations to obtain K correction contribution weights; use the benchmark PID control parameters as the benchmark anchor point, and perform weighted correction on the K candidate PID control parameters according to the K correction contribution weights to obtain the corrected PID control parameters.

[0068] Furthermore, the operating condition characteristic index group consists of ambient wet-bulb temperature, cooling water supply main flow rate, first-order rate of change of flow rate, supply and return water temperature difference, terminal total load feedback signal, and instantaneous water replenishment flow rate; the deviation feedback index group consists of real-time pressure value of the controlled point, pressure deviation of the controlled point, and first-order rate of change of deviation.

[0069] Furthermore, the PID fine-tuning module 50 is used to perform the following operation steps: Based on the real-time pressure value of the controlled point in the deviation feedback index group, a pressure condition label is determined; using the pressure deviation and first-order change rate of the controlled point in the deviation feedback index group as input, the PID parameter increment is matched in the preset fine-tuning rule library; after biasing and superimposing the PID parameter increment according to the pressure condition label, amplitude limiting and smoothing processing is performed to obtain an effective fine-tuning increment; the effective fine-tuning increment is superimposed on the base PID control parameter to obtain the fine-tuned PID control parameter for constant pressure control of the cooling tower.

[0070] Furthermore, if the matching result of the operating condition vector in the scenario mode library is an empty set, then the operating condition vector is used for fault mode matching to trigger a cooling tower fault alarm.

[0071] Through the foregoing detailed description of the cooling tower constant pressure PID adaptive control method based on process monitoring, those skilled in the art can clearly understand the cooling tower constant pressure PID adaptive control system based on process monitoring in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to in the method section.

[0072] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A constant-pressure PID adaptive control method for cooling towers based on process monitoring, characterized in that, The method includes: Decouple the operating condition characteristic index group and the deviation feedback index group from the process monitoring data pool; Based on the aforementioned set of operating condition characteristic indicators, an operating condition vector is constructed, and an initial operating scenario group is matched in the scenario pattern library. Using the collection timestamp of the working condition characteristic index group as the starting point for backtracking, the historical state transition chain of the preset time window is retrieved, and the initial running scenario group is corrected based on the scenario state transition prediction to output the real-time running scenario. Based on the real-time operating scenario, the baseline PID control parameters are matched to perform a smooth, non-disruptive coarse adjustment of the cooling tower constant pressure control. Using the aforementioned baseline PID control parameters as the starting point for online control, the constant pressure fine-grained PID fine-tuning of the cooling tower driven by the aforementioned deviation feedback index group is executed. The scenario switching analysis under incremental monitoring is performed on the aforementioned working condition characteristic index group until the operating scenario changes, triggering a fine-grained PID fine-tuning loop under non-disruptive smooth coarse adjustment.

2. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 1, characterized in that, The method involves constructing an operating condition vector based on the aforementioned operating condition feature index group, and matching an initial operating scenario group with the scenario pattern library. Predefine multiple combinations of operational characteristic conditions for various typical operational scenarios; Based on feature generality ranking, the multiple combinations of operational feature conditions are hierarchically divided to construct a scene hierarchy model tree, and the multiple typical operational scenarios are associated with multiple leaf nodes of the scene hierarchy model tree. The operating condition vector is mapped to the scene hierarchical model tree, and joint pruning under hierarchical progressive search is performed to obtain the matching node tree; Extract typical operating scenarios associated with multiple matching nodes from the matching nodes to form the initial operating scenario group.

3. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 2, characterized in that, The method involves mapping the operating condition vector to the scene hierarchical model tree, performing joint pruning under hierarchical progressive search, and obtaining a matching node tree. The operating condition vector is mapped to the scene hierarchy model tree, and coarse-grained node filtering under breadth-first search is performed to prune the scene hierarchy model tree as candidate subtrees. Fine-grained node verification is performed on the candidate subtrees using a depth-first search of the operating condition vector, and the candidate subtrees are pruned to form the matching node tree.

4. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 2, characterized in that, Using the collection timestamp of the aforementioned working condition characteristic index group as the starting point for backtracking, the historical state transition chain of a preset time window is retrieved, and the initial running scenario group is corrected based on the scenario state transition prediction to determine the scenario membership probability, thereby outputting the real-time running scenario. The method includes: The preset time window is dynamically adjusted based on the frequency of scene switching; Using the collection timestamp as the starting point for backtracking, and combining it with the preset time window to construct a backtracking interval, the historical state transition chain is retrieved from the circular log buffer; By combining and enumerating the various typical operating scenarios, multiple scenario jump state pairs are constructed. The frequency of the historical state transition chain is traversed to obtain multiple empirical transition probabilities and a dynamic empirical transition matrix is ​​constructed. The initial running scenario group is mapped to the dynamic experience transition matrix to obtain the scenario membership probability group; The initial running scenario group is sorted in descending order according to the probability group to which the scenario belongs, and the real-time running scenario with high confidence is located.

5. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 4, characterized in that, Based on the real-time operating scenario, a baseline PID control parameter is matched to perform a non-disruptive, smooth coarse adjustment of the cooling tower constant pressure control. The method includes: The scene membership probability group is filtered based on a preset membership probability threshold, so as to retrieve K candidate running scenarios from the initial running scene group; Using the scenario IDs of the real-time running scenario and the K candidate running scenarios as indexes, the baseline PID control parameters and the K candidate PID control parameters are retrieved from the scenario parameter mapping table respectively. The K scenario membership probabilities of the K candidate operating scenarios are used as contribution coefficients to perform intersection consistency correction on the K candidate PID control parameters and the benchmark PID control parameters to obtain the corrected PID control parameters. The modified PID control parameters are used for perturbation-free smooth coarse adjustment of constant pressure control in the cooling tower.

6. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 5, characterized in that, Using the K scenario membership probabilities of the K candidate operating scenarios as contribution coefficients, the intersection consistency of the K candidate PID control parameters and the baseline PID control parameters is corrected to obtain the corrected PID control parameters. The method includes: Calculate the K parameter deviations of the K candidate PID control parameters relative to the baseline PID control parameters; The K scene membership probabilities and K parameter deviations are fused to obtain K correction contribution weights. Using the baseline PID control parameters as the baseline anchor point, the K candidate PID control parameters are weighted and corrected according to the K correction contribution weights to obtain the corrected PID control parameters.

7. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 1, characterized in that, The operating condition characteristic index group consists of ambient wet-bulb temperature, cooling water supply main flow rate, first-order rate of change of flow rate, supply and return water temperature difference, terminal total load feedback signal, and instantaneous water replenishment flow rate. The deviation feedback index group consists of real-time pressure value of the controlled point, pressure deviation of the controlled point, and first-order rate of change of deviation.

8. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 7, characterized in that, Using the aforementioned baseline PID control parameters as the starting point for online control, the method includes executing fine-grained PID tuning of the cooling tower under constant pressure driven by the aforementioned deviation feedback index group, comprising: Based on the real-time pressure value of the controlled point in the deviation feedback index group, the pressure condition label is determined; Using the controlled point pressure deviation and the first-order rate of change of the deviation in the deviation feedback index group as inputs, the PID parameter increment is matched in the preset fine-tuning rule base; After biasing and superimposing the PID parameter increments according to the pressure condition label, a limiting and smoothing process is performed to obtain an effective fine-tuning increment; The effective fine-tuning increment is superimposed on the base PID control parameters to obtain the fine-tuned PID control parameters, which are then used for constant pressure control of the cooling tower.

9. The cooling tower constant pressure PID adaptive control method based on process monitoring as described in claim 1, characterized in that, If the matching result of the operating condition vector in the scenario mode library is an empty set, then the operating condition vector is used for fault mode matching to trigger a cooling tower fault alarm.

10. A constant-pressure PID adaptive control system for cooling towers based on process monitoring, characterized in that, For implementing the cooling tower constant pressure PID adaptive control method based on process monitoring as described in any one of claims 1-9, the system comprises: The indicator decoupling module is used to decouple the operating condition characteristic indicator group and the deviation feedback indicator group from the process monitoring data pool; The scenario matching module is used to construct a running condition vector based on the set of operating condition feature indicators and match the initial running scenario group in the scenario pattern library. The probability correction module is used to retrieve the historical state transition chain of the preset time window, take the collection timestamp of the working condition characteristic index group as the backtracking starting point, perform scene membership probability correction on the initial running scene group based on scene state transition prediction, and output the real-time running scene. The smooth coarse adjustment module is used to match the benchmark PID control parameters according to the real-time operating scenario and perform non-disruptive smooth coarse adjustment of the cooling tower constant pressure control. The PID fine-tuning module is used to perform fine-grained PID fine-tuning of the cooling tower under constant pressure driven by the deviation feedback index group, with the reference PID control parameters as the starting point for online control. The fine-tuning loop module is used to perform scene switching analysis under incremental monitoring of the working condition characteristic index group until the operating scene changes, triggering a fine-grained PID fine-tuning loop under non-disturbance smooth coarse adjustment.

Images