Cooling tower icing risk classification driven waste heat recovery anti-freezing control method and system

Abstract

The application discloses a cooling tower icing risk grading driven waste heat recovery anti-freezing control method and system, and belongs to the technical field of industrial automation control, which comprises the following steps: constructing an icing sensitive characteristic matrix by fusing multi-dimensional parameters such as environmental wind speed, humidity, cooling water temperature difference and tower body health degree, and performing risk grading evaluation; performing collaborative optimization scheduling according to the risk grade and available waste heat resources to generate anti-freezing control instructions; collecting state parameters to calculate effect deviation after execution, and online optimizing risk mapping rules through grading correction vectors. Through the combination of multi-parameter fusion risk evaluation, waste heat scheduling based on collaborative optimization and closed-loop feedback adaptive learning, precise, energy-saving and intelligent anti-freezing control of the cooling tower can be realized.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, and in particular to a waste heat recovery and antifreeze control method and system driven by the risk classification of cooling tower icing. Background Technology

[0002] Cooling towers, as crucial heat exchange equipment, utilize the evaporation of water to remove heat through direct or indirect contact between circulating water and air. They are widely used in power, petrochemical, metallurgical, and data center industries, playing a vital role in ensuring industrial production and normal equipment operation. During cold seasons, when ambient temperatures drop below freezing, components of cooling towers such as air inlets, packing, and water collectors are highly susceptible to icing. Icing not only severely impacts the heat exchange efficiency of the cooling tower but can also damage the tower structure and even trigger systemic safety accidents. Therefore, effective antifreeze control for cooling towers is essential.

[0003] Existing cooling tower antifreeze control systems typically employ relatively simple control strategies. A common approach is based on single-point monitoring of the cooling water tank temperature or the outlet water temperature. When the temperature falls below a preset fixed threshold, antifreeze measures are triggered, such as activating an electric heater to heat the water tank, or reducing cooling intensity by adjusting fan speed or opening bypass pipes. Some systems also employ a strategy of periodically reversing the fan speed to utilize hot air inside the tower to melt the ice layer at the air inlet. The logic of these control methods is relatively straightforward, relying on simple judgments of a single physical quantity and fixed execution actions.

[0004] However, the control methods in the aforementioned existing technologies have significant drawbacks. First, their risk assessment relies on too single-factor analysis. Icing risk is the result of multiple factors, including ambient temperature, humidity, wind speed, and the cooling tower's own heat load. Relying solely on water temperature cannot accurately predict icing, often leading to delayed or misjudged antifreeze actions. Second, existing antifreeze measures are generally energy-intensive. For example, electric heating is expensive, while adjusting fans or bypassing operations sacrifices the cooling tower's normal heat exchange performance, impacting production efficiency. Furthermore, the parameters and logic of these control systems are relatively fixed once set, unable to adaptively adjust to equipment performance degradation or long-term changes in environmental conditions, resulting in low robustness and intelligence. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a waste heat recovery and antifreeze control method and system driven by the risk classification of cooling tower icing. By combining multi-parameter fusion risk assessment, waste heat scheduling based on collaborative optimization, and adaptive learning with closed-loop feedback, it can achieve precise, energy-saving, and intelligent antifreeze control of cooling towers.

[0006] The above objectives can be achieved through the following approach:

[0007] A waste heat recovery and antifreeze control method driven by cooling tower icing risk classification includes: acquiring environmental wind speed parameters, environmental humidity parameters, cooling circuit inlet water temperature parameters, cooling circuit return water temperature parameters, and the health parameters of the tower's heat exchange elements; fusing these parameters to generate an icing-sensitive feature matrix; performing risk factor identification and level mapping processing on the icing-sensitive feature matrix to generate an icing risk level; performing collaborative optimization calculations based on the icing risk level and currently available waste heat resources to generate a waste heat recovery scheduling vector; converting the waste heat recovery scheduling vector into control parameters for the cooling tower actuators to generate an antifreeze control command set; executing the antifreeze control command set and collecting system state parameters after execution to generate implementation effect parameters; calculating the deviation between the implementation effect parameters and the expected antifreeze recovery effect to generate a classification correction vector; and using the classification correction vector to optimize the level mapping process.

[0008] Optionally, generating the icing-sensitive feature matrix includes: acquiring environmental wind speed and humidity parameters of the area where the cooling tower is located, performing dimensionless processing to generate environmental state parameters; acquiring inlet water temperature parameters at the tower inlet and return water temperature parameters at the tower outlet of the cooling circuit, calculating the temperature difference value, and combining it with the environmental state parameters to perform weighted fusion to generate cooling load feature parameters; acquiring health parameters characterizing the degree of scaling, blockage, or damage of the heat exchange elements in the tower, and spatiotemporally aligning and coupling the health parameters with the cooling load feature parameters to generate the icing-sensitive feature matrix.

[0009] Optionally, the process of identifying risk factors and mapping levels in the icing-sensitive feature matrix to generate an icing risk level includes: performing principal component analysis on the icing-sensitive feature matrix to extract key feature dimensions affecting icing risk and generating a dimensionality-reduced feature vector; based on the dimensionality-reduced feature vector, using a preset fuzzy membership function to identify the contribution of each feature dimension to icing risk and generating a risk factor membership set; and matching and mapping the risk factor membership set using a preset risk level mapping rule base to output a discretized icing risk level.

[0010] Optionally, based on the icing risk level and currently available waste heat resource information, collaborative optimization calculations are performed to generate a waste heat recovery scheduling vector. This includes: obtaining the temperature parameters, flow parameters, and location distribution parameters of waste heat resources that can be used for antifreeze, conducting availability assessments, and generating a waste heat resource state vector; using the icing risk level as a constraint, and aiming to maximize waste heat utilization efficiency and minimize antifreeze energy consumption, establishing a multi-objective optimization function; inputting the waste heat resource state vector into the multi-objective optimization function for solving, calculating the waste heat allocated to different cooling tower loops or areas and their commissioning sequence, and generating a waste heat recovery scheduling vector.

[0011] Optionally, converting the waste heat recovery scheduling vector into control parameters for the cooling tower actuators and generating an antifreeze control instruction set includes: calculating the opening control parameters of the corresponding regulating valve and the speed control parameters of the circulating pump based on the waste heat distribution information in the waste heat recovery scheduling vector; generating a timing control logic sequence based on the commissioning timing information in the waste heat recovery scheduling vector and the response characteristics of each actuator in the cooling tower; and encapsulating and format-converting the opening control parameters, speed control parameters, and timing control logic sequence to generate an antifreeze control instruction set that can directly drive the actuators.

[0012] Optionally, executing the antifreeze control command set and collecting system status parameters after execution to generate implementation effect parameters includes: executing the antifreeze control command set to the regulating valve and circulating pump of the cooling tower, and collecting the actual feedback parameters of each actuator in real time; collecting the wall temperature parameters of key areas of the cooling tower, the real-time water temperature change parameters of the cooling circuit, and the environmental state change parameters after the execution of the commands; and performing fusion analysis on the actual feedback parameters, the wall temperature parameters, the real-time water temperature change parameters, and the environmental state change parameters to generate implementation effect parameters characterizing the immediate effect of the antifreeze measures.

[0013] Optionally, calculating the deviation between the implementation effect parameters and the expected antifreeze recovery effect to generate a graded correction vector includes: obtaining the expected antifreeze effect threshold and the expected waste heat recovery efficiency target value corresponding to the current icing risk level; comparing the antifreeze effect index in the implementation effect parameters with the expected antifreeze effect threshold to calculate a first deviation value; comparing the waste heat recovery efficiency index in the implementation effect parameters with the expected waste heat recovery efficiency target value to calculate a second deviation value; and generating a graded correction vector based on the magnitude and direction of the first and second deviation values, combined with the icing risk level.

[0014] Optionally, optimizing the risk level mapping process using the graded correction vector includes: inputting the graded correction vector into the adjustment interface of the risk level mapping rule base; correcting the feature threshold or membership function shape corresponding to the risk level in the risk level mapping rule base according to the deviation information and risk level association information carried in the graded correction vector; obtaining historical optimization records, and using the historical optimization records to perform stability verification on the corrected risk level mapping rule base.

[0015] Optionally, after optimizing the grade mapping process using the graded correction vector, the method further includes: acquiring the sequence of graded correction vectors generated during historical optimization, performing time-series analysis and feature extraction, and generating a correction trend feature vector; based on the correction trend feature vector, identifying the dynamic response lag characteristics between the icing risk level and the waste heat recovery scheduling vector, and generating feedforward compensation parameters; and in the next control cycle, using the feedforward compensation parameters to correct the waste heat resource state vector.

[0016] Based on the same inventive concept, this invention also provides a waste heat recovery and antifreeze control system driven by cooling tower icing risk classification. The system includes: a data acquisition and fusion module for acquiring environmental wind speed parameters, environmental humidity parameters, cooling circuit inlet water temperature parameters, cooling circuit return water temperature parameters, and the health parameters of the tower's heat exchange elements, and performing fusion processing to generate an icing-sensitive feature matrix; a risk classification assessment module for identifying risk factors and mapping levels in the icing-sensitive feature matrix to generate an icing risk level; a collaborative scheduling calculation module for performing collaborative optimization calculations based on the icing risk level and currently available waste heat resources to generate a waste heat recovery scheduling vector; an instruction generation and execution module for converting the waste heat recovery scheduling vector into control parameters for the cooling tower actuators to generate an antifreeze control instruction set; an effect monitoring and feedback module for executing the antifreeze control instruction set and collecting system status parameters after execution to generate implementation effect parameters; an online learning module for calculating the deviation between the implementation effect parameters and the expected antifreeze recovery effect to generate a classification correction vector; and a correction module for optimizing the classification mapping process using the classification correction vector.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] This invention constructs an icing-sensitive feature matrix by integrating multi-dimensional information such as ambient wind speed, humidity, cooling water temperature difference, and tower health, thereby assessing the true potential risk of icing. Based on this, a tiered processing mechanism allows antifreeze control strategies to be matched with risk levels, avoiding the problems of delayed or excessive antifreeze measures caused by inaccurate assessments in traditional control methods, thus improving the safety and reliability of system operation.

[0019] This invention utilizes available waste heat resources within the system as a heat source for freeze protection. By establishing a multi-objective optimization function, it coordinates the scheduling of waste heat utilization efficiency and minimizes transmission energy consumption while ensuring freeze protection safety. This replaces traditional freeze protection methods such as electric heating, which consume a large amount of primary energy, transforming the freeze protection process into an energy recovery and reuse process, thus possessing environmental value.

[0020] This invention constructs a complete closed loop encompassing performance monitoring, deviation calculation, and feedback correction. The system can compare the actual effect of each anti-freeze control operation with the expected target and automatically adjust and optimize its core risk level mapping rules based on deviation information. This online learning mechanism enables the control system to continuously accumulate experience and improve itself, thereby automatically adapting to dynamic conditions such as equipment aging, climate change, and operating condition fluctuations. This reduces reliance on manual intervention and debugging, achieving truly intelligent and adaptive control. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic flowchart of the waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing in an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the cooling circuit temperature parameter acquisition according to an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of the icing-sensitive feature matrix according to an embodiment of the present invention.

[0025] Figure 4 This is a schematic diagram of the feature space distribution of PCA dimensionality reduction according to an embodiment of the present invention.

[0026] Figure 5 This is a schematic diagram of the multi-source waste heat collaborative scheduling optimization results according to an embodiment of the present invention.

[0027] Figure 6 This is a schematic diagram of the waste heat recovery and antifreeze control system driven by the risk classification of cooling tower icing in an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1One embodiment of the present invention proposes a waste heat recovery antifreeze control method driven by the risk classification of cooling tower icing. By combining risk assessment with multi-parameter fusion, waste heat scheduling based on collaborative optimization, and adaptive learning with closed-loop feedback, it can achieve precise, energy-saving and intelligent antifreeze control of cooling towers.

[0030] The method described in this embodiment specifically includes:

[0031] The system acquires and merges environmental wind speed parameters, environmental humidity parameters, cooling circuit inlet water temperature parameters, cooling circuit return water temperature parameters, and the health parameters of the heat exchange elements in the tower body to generate an icing-sensitive feature matrix.

[0032] The icing-sensitive feature matrix is ​​subjected to risk factor identification and level mapping to generate an icing risk level.

[0033] Based on the icing risk level and the currently available waste heat resources, a collaborative optimization calculation is performed to generate a waste heat recovery scheduling vector.

[0034] The waste heat recovery scheduling vector is converted into control parameters for the cooling tower actuator, and an antifreeze control instruction set is generated.

[0035] Execute the antifreeze control command set, collect the system status parameters after execution, and generate implementation effect parameters;

[0036] Calculate the deviation between the implementation effect parameters and the expected antifreeze recovery effect, and generate a graded correction vector;

[0037] The grade mapping process is optimized using the graded correction vector.

[0038] Optionally, the generation of the icing-sensitive feature matrix includes:

[0039] The environmental wind speed and humidity parameters of the area where the cooling tower is located are obtained, and dimensionless processing is performed to generate environmental state parameters.

[0040] The inlet water temperature parameters at the tower body inlet and the return water temperature parameters at the tower body outlet of the cooling circuit are obtained, the temperature difference value is calculated, and combined with the environmental state parameters, a weighted fusion is performed to generate the cooling load characteristic parameters.

[0041] Obtain health parameters that characterize the degree of scaling, blockage, or damage to the heat exchange elements in the tower. Then, perform spatiotemporal alignment and correlation coupling between the health parameters and the cooling load characteristic parameters to generate an icing-sensitive feature matrix.

[0042] Specifically, ultrasonic anemometers and capacitive humidity sensors deployed at or near the cooling tower's air inlet acquire real-time environmental wind speed and humidity parameters of the area at a sampling frequency of at least 0.5 Hz. To eliminate computational obstacles caused by different physical dimensions and numerical ranges, the system performs dimensionless processing on the collected raw data. For example, a min-max normalization method is used to map real-time wind speed and humidity values ​​to a range of 0 to 1, generating normalized wind speed parameters. and standardized humidity parameters Subsequently, the system performs linear weighting based on preset weighting coefficients to generate a comprehensive environmental state parameter E:

[0043] ,

[0044] Here, E represents the final environmental state parameter, which is a dimensionless comprehensive index. For the standardized environmental wind speed, This refers to the standardized ambient humidity. and These are the weighting coefficients for wind speed and humidity, respectively, with a sum of 1. Their specific values ​​are determined based on the climatic statistical characteristics of the geographical location of the cooling tower and historical icing data, in order to reflect the degree to which different environmental factors dominate the icing process.

[0045] like Figure 2 As shown, the system utilizes Pt100 platinum resistance temperature sensors installed on the main inlet and return pipes of the cooling circuit to simultaneously acquire the inlet water temperature parameters of the cooling circuit at the tower inlet. and the return water temperature parameters at the tower outlet The system calculates the difference between the two to obtain the temperature difference value. This value directly reflects the actual heat dissipation of the cooling tower under current operating conditions. Since the risk of icing is not only related to the heat dissipation but also significantly affected by the environmental heat dissipation capacity, the system will use the temperature difference value... The environmental state parameter E calculated in the previous step is weighted and fused to generate the cooling load characteristic parameter L:

[0046] ,

[0047] Where L represents the cooling load characteristic parameter, with the same unit as temperature. ΔT is the inlet and outlet water temperature difference, in degrees Celsius. C is a proportionality coefficient related to the cooling tower design and the physical properties of the fluid medium, used to calibrate the base load level. E is the dimensionless environmental state parameter calculated above. k is the environmental impact gain coefficient, which is positive and used to adjust for the amplification effect of environmental factors on the cooling load. For example, in environments with high wind speed and low humidity, a higher E value leads to an increased L value, indicating a higher cooling intensity and a correspondingly increased risk of icing.

[0048] The system acquires a health parameter H, characterizing the heat exchange elements of the tower, primarily the packing and water collector, to assess the degree of scaling, blockage, or damage. This parameter H is a comprehensive evaluation value, typically ranging from 0 to 1, where 1 represents a brand-new state and 0 represents complete failure. It can be obtained through comparative analysis of operating and design pressure differentials, or through periodic offline monitoring and model estimation. Since the variation period of the health parameter H is much longer than that of the operating parameters, the system performs a spatiotemporal alignment operation, correlating the frequently collected cooling load characteristic parameter L with the less frequently updated health parameter H at timestamps. Next, the system couples the environmental state parameter E, the cooling load characteristic parameter L, and the health parameter H to generate the final icing sensitivity feature matrix M. The structure of this matrix within a control cycle, such as one minute, can be a feature vector. :

[0049] ,

[0050] in, This represents the icing-sensitive feature vector at time t. and These represent the environmental state parameters and cooling load characteristic parameters at that moment, respectively. H is the currently active health parameter. Over a continuous time series, these feature vectors constitute the icing sensitivity feature matrix used for risk assessment, such as... Figure 3 As shown in the figure, this matrix integrates the driving forces of the external environment, the instantaneous state of the internal heat load, and the long-term performance degradation of the heat exchange elements, providing a comprehensive and multi-dimensional data foundation for accurate icing risk classification.

[0051] Optionally, the step of performing risk factor identification and level mapping processing on the icing-sensitive feature matrix to generate an icing risk level includes:

[0052] Principal component analysis was performed on the icing-sensitive feature matrix to extract the key feature dimensions that affect the icing risk and generate a dimensionality-reduced feature vector.

[0053] Based on the reduced feature vector, the contribution of each feature dimension to the icing risk is identified using a preset fuzzy membership function, and a risk factor membership set is generated.

[0054] Based on the set of membership degrees of the risk factors, matching and mapping are performed through a preset risk level mapping rule base to output a discrete icing risk level.

[0055] Specifically, such as Figure 4As shown, principal component analysis (PCA) is performed on the obtained icing-sensitive feature matrix M. PCA is a statistical method whose engineering application lies in transforming multiple original correlated indicators into a few linearly independent comprehensive indicators, i.e., principal components. The system first standardizes the data in matrix M, including environmental state parameters, cooling load characteristic parameters, and health parameters, ensuring their mean is 0 and variance is 1. Then, it calculates the covariance matrix and solves for the eigenvalues ​​and eigenvectors. Based on a preset cumulative variance contribution rate threshold, for example, selecting the top k principal components whose total eigenvalues ​​exceed 95%, the system projects the original eigenvectors onto a new orthogonal coordinate system composed of these k principal components, thereby generating a dimensionality-reduced eigenvector Z.

[0056] ,

[0057] Where Z represents the dimensionality-reduced feature vector. This is the normalized matrix M of the original icing-sensitive feature matrix. W is the transformation matrix consisting of k principal eigenvectors calculated from the covariance matrix, and T represents the matrix transpose. Through this step, the original multidimensional features are condensed into a few key feature dimensions, reducing the complexity of subsequent calculations.

[0058] The system, based on the dimensionality-reduced feature vector Z, uses a pre-defined fuzzy membership function to identify the contribution of each feature dimension to the icing risk. A fuzzy membership function is a function that maps input variables to a range of 0 to 1, representing the degree to which the variable belongs to a certain fuzzy set. For example, for the first principal component representing the overall cooling intensity... The system pre-defines three fuzzy sets of risk contribution levels: low, medium, and high, and defines them using trapezoidal or Gaussian membership functions. The system then uses each component of the dimensionality-reduced feature vector Z... As input, through the corresponding membership function Calculate the degree to which it belongs to the j-th risk contribution fuzzy set, and finally generate the risk factor membership set U:

[0059] ,

[0060] Where U represents the set of membership degrees of risk factors. Represents the i-th principal component. For the j-th risk contribution fuzzy set, such as "high-risk contribution", the membership value is given. k is the number of feature dimensions after dimensionality reduction, and m is the preset number of risk contribution fuzzy sets.

[0061] Based on the generated risk factor membership set U, matching and mapping are performed using a pre-defined risk level mapping rule base. This rule base consists of a series of "IF-THEN" rules based on expert experience and historical data mining, such as "IF..." Belonging to high-risk contribution AND The system is classified as "belonging to the high-risk contribution level THEN, with an icing risk level of 4". It employs a fuzzy inference engine, such as a Mamdani or Sugeno model, taking the risk factor membership set U as input, activating relevant rules in the rule base, and defuzzifying the outputs of all activated rules. For example, it uses the centroid method to calculate a precise comprehensive risk assessment value. Finally, the system compares this comprehensive assessment value with a preset level classification threshold and outputs a discretized icing risk level R, for example, classifying the risk level into 1-5 levels, where level 1 represents no risk and level 5 represents extremely high risk. This icing risk level R will serve as the core driving signal for subsequent waste heat recovery and antifreeze coordinated optimization control.

[0062] Optionally, the step of generating a waste heat recovery scheduling vector by performing collaborative optimization calculations based on the icing risk level and currently available waste heat resource information includes:

[0063] The temperature, flow, and location distribution parameters of waste heat resources that can be used for frost prevention are obtained, and an availability assessment is conducted to generate a waste heat resource state vector.

[0064] Using the aforementioned icing risk level as a constraint, and aiming to maximize waste heat utilization efficiency and minimize antifreeze energy consumption, a multi-objective optimization function is established.

[0065] The waste heat resource state vector is input into the multi-objective optimization function for solution, and the waste heat and commissioning sequence allocated to different cooling tower circuits or areas are calculated to generate a waste heat recovery scheduling vector.

[0066] Specifically, such as Figure 5 As shown, various parameters of the waste heat resources that can be used for freeze protection are acquired in real time through a communication interface. Specifically, the temperature parameters are acquired using a Pt100 temperature sensor deployed on the waste heat source pipeline. The maximum usable flow rate parameter is obtained using an electromagnetic flowmeter. The system determines the location distribution parameters between the pipeline network and the antifreeze circuits of each cooling tower using the preset pipeline network topology information. The system evaluates and integrates this information to generate a multi-dimensional waste heat resource state vector. :

[0067] ,

[0068] in, This represents the state vector of waste heat resources. , , These represent the temperature, maximum available flow rate, and location code of the i-th waste heat source, respectively, where n is the total number of currently available waste heat sources. This vector serves as the resource input for optimization calculations.

[0069] The system uses the previously generated icing risk level R as the core constraint and establishes a multi-objective optimization function J with the dual objectives of maximizing waste heat utilization efficiency and minimizing antifreeze energy consumption. The icing risk level R is transformed into a specific physical constraint, namely, the predicted wall temperature of the critical area of ​​the cooling tower after the implementation of antifreeze measures. It must be higher than the safety threshold determined by the risk level R. For example, a level 4 risk might correspond to... The temperature is 2 degrees Celsius. The optimization objective is represented by a weighted function, where waste heat utilization efficiency is quantified by the recovered heat, and antifreeze energy consumption mainly refers to the pumping energy consumption for transporting the waste heat medium.

[0070] ,

[0071] ,

[0072] in, Let be the multi-objective optimization function to be maximized. x is a decision variable vector, representing the waste heat allocated to different loops and their commissioning timing, i.e., the waste heat recovery scheduling vector to be solved. It is a normalized waste heat recovery efficiency index calculated based on the decision vector x. This is the corresponding normalized antifreeze energy consumption index. and These are weighting coefficients, which sum to 1 and are dynamically adjusted according to the operating strategy, for example, increasing when energy costs are high. Constraints It ensures antifreeze safety, among which It is a wall temperature predicted based on a heat transfer model and the scheduling vector x.

[0073] The system will use the waste heat resource state vector The known parameters are substituted into the established optimization model. Then, the system calls built-in optimization algorithms, such as Particle Swarm Optimization (PSO) or Genetic Algorithm (GA), to iteratively solve the constrained nonlinear optimization problem. The algorithm searches for the optimal solution within the feasible region by continuously adjusting the values ​​of the decision variable x. After calculation, the final output solution is the waste heat recovery scheduling vector D:

[0074] ,

[0075] Where D represents the final generated waste heat recovery scheduling vector. This represents the waste heat flow allocated to the j-th antifreeze circuit. and These represent the start time and duration of the waste heat recovery system activation, respectively, and m is the total number of loops or areas requiring antifreeze intervention. This vector provides a precise quantitative basis for generating specific control commands subsequently.

[0076] Optionally, the waste heat recovery scheduling vector is converted into control parameters for the cooling tower actuator, and the antifreeze control instruction set is generated, including:

[0077] Based on the waste heat distribution information in the waste heat recovery scheduling vector, calculate the opening control parameters of the corresponding regulating valve and the speed control parameters of the circulating pump;

[0078] Based on the commissioning timing information in the waste heat recovery scheduling vector and combined with the response characteristics of each actuator of the cooling tower, a timing control logic sequence is generated.

[0079] The opening control parameters, speed control parameters, and timing control logic sequence are encapsulated and format-converted to generate an anti-freeze control instruction set that can directly drive the actuator.

[0080] Specifically, analyze the waste heat recovery scheduling vector D for information about waste heat allocation, that is, the waste heat flow allocated to the j-th antifreeze loop. For each loop, the system calls a pre-calibrated device characteristic curve model to calculate the flow rate. This is converted into the corresponding control parameters for the valve opening and the circulating pump speed. For example, for a control valve, the system queries its flow-opening characteristic curve (…). The target flow rate is calculated using an inverse function (value curve). Required target opening percentage For variable frequency circulating pumps, the system calculates the target flow rate based on the pump performance curve and pipeline resistance curve. The required operating frequency :

[0081] ,

[0082] ,

[0083] in, For the target opening degree of the j-th regulating valve, This represents the inverse function of the valve's flow-opening characteristic curve. Let j be the target operating frequency of the j-th circulating pump. This represents the inverse function of the pump's flow-frequency relationship model, which also needs to consider pipeline resistance. As a parameter.

[0084] Extract the commissioning timing information from the waste heat recovery scheduling vector D, including the start time. and duration Considering the response delay of actuators such as control valves and pumps from receiving commands to completing actions, the system combines pre-stored response characteristic parameters of each actuator, such as valve opening time (e.g., 3-5 seconds) and pump start / stop delay, to generate a timing control logic sequence with lead times. This sequence not only specifies the precise trigger time of each action but may also include coordination logic between actions, such as safety strategies like "open the valve before starting the pump" or "slowly open the valve to prevent water hammer in the pipeline." For example, the command might be set at... The valve will begin to open pre-open in the first 5 seconds. The pump starts at the target frequency at a certain time.

[0085] The opening control parameters generated in the previous two steps Speed ​​control parameters The timing control logic sequence is encapsulated. This process involves data format conversion, such as converting floating-point opening or frequency values ​​into integers or digital quantities corresponding to 4-20mA analog signals that the controller can recognize. Simultaneously, the timing logic is compiled into executable logic blocks or scripts. Finally, this information is integrated into one or more structured data frames, forming an anti-freeze control instruction set C that can be directly sent to PLCs or DCS systems via industrial buses such as Modbus or PROFIBUS. This instruction set C is the direct basis for the control system to execute anti-freeze operations, ensuring the automation and accuracy of the entire anti-freeze recovery process.

[0086] Optionally, the antifreeze control instruction set is executed, and the system status parameters after execution are collected to generate implementation effect parameters, including:

[0087] The antifreeze control command set is executed on the regulating valve and circulating pump of the cooling tower, and the actual feedback parameters of each actuator are collected in real time.

[0088] Collect and execute commands to obtain wall temperature parameters of key areas of the cooling tower, real-time water temperature change parameters of the cooling circuit, and environmental condition change parameters.

[0089] The actual feedback parameters, wall temperature parameters, real-time water temperature change parameters, and environmental state change parameters are fused and analyzed to generate implementation effect parameters that characterize the immediate effect of antifreeze measures.

[0090] Specifically, the system sends the packaged antifreeze control instruction set C to the programmable logic controller (PLC) or distributed control system (DCS) of the cooling tower control system via an industrial fieldbus. These controllers then drive the actuators in the field, such as regulating valves and circulating pumps. After the instructions are issued, the system collects real-time feedback parameters such as stroke and speed from the feedback ends of the actuators, such as the position feedback of the valve positioner and the frequency feedback of the frequency converter, at a frequency of not less than 1 Hz. These parameters reflect the actual physical state of the device, and by comparing them with the target values ​​in the instruction set, the deviation at the execution level can be quantified.

[0091] During and within a preset time window (e.g., 5 minutes) after the actuator's operation, the system simultaneously initiates high-frequency acquisition of key status parameters. This includes using surface thermocouples or infrared thermometers embedded in easily icing areas such as the packing layer, the edge of the water collection tray, and the air inlet louvers to collect wall temperature parameters of key areas of the cooling tower. Using Pt100 temperature sensors at the inlet and outlet of the cooling circuit, the real-time water temperature change parameters of the cooling circuit are continuously recorded after the command is executed to track the heating effect brought about by the waste heat injection. At the same time, the environmental conditions such as wind speed and humidity are continuously monitored to eliminate the interference of sudden changes in the external environment on the effect evaluation.

[0092] The system receives the actual feedback parameters collected above. Wall temperature parameters The analysis integrates real-time water temperature changes and environmental condition changes. This analysis is not simply a data listing, but rather generates a set of Key Performance Indicators (KPIs) through calculation. Ultimately, these indicators are encapsulated into a multi-dimensional vector, representing the implementation effect parameters. The vector is constructed as shown in the following equation:

[0093] ,

[0094] in, Parameters representing the implementation effect. Representing the lowest wall temperature collected at all measuring points during the monitoring period, it is a core indicator for assessing the freeze protection safety margin, directly derived from the wall temperature parameter. Extract from. Representing the actual waste heat recovery efficiency, it is calculated by comparing the actual recovered heat, obtained by using real-time water temperature change parameters and loop flow rate, with the theoretical maximum recoverable heat provided by the waste heat source. It quantifies the actual effectiveness of energy recovery. This represents the execution deviation, calculated by comparing the target value in the antifreeze control command set with the collected actual feedback parameters. The root mean square error between these parameters is used to determine the accuracy of the control system's execution. This parameter reflects the overall performance of the system. Vectors provide quantitative, effect-oriented inputs for subsequent deviation calculations and system optimization.

[0095] Optionally, calculating the deviation between the implementation effect parameters and the expected antifreeze recovery effect, and generating a graded correction vector, includes:

[0096] Obtain the expected antifreeze effect threshold and the expected waste heat recovery efficiency target value corresponding to the current icing risk level;

[0097] The antifreeze effect index in the implementation effect parameters is compared with the expected antifreeze effect threshold, and a first deviation value is calculated;

[0098] The waste heat recovery efficiency index in the implementation effect parameters is compared with the expected waste heat recovery efficiency target value, and a second deviation value is calculated.

[0099] Based on the magnitude and direction of the first and second deviation values, and in conjunction with the icing risk level, a graded correction vector is generated.

[0100] Specifically, based on the currently active icing risk level R, the system queries and retrieves the corresponding expected antifreeze effect threshold and expected waste heat recovery efficiency target value from a preset performance target database. For example, for an icing risk level of 3, the system might retrieve the expected antifreeze effect threshold... The temperature should not be lower than 3 degrees Celsius, while the expected waste heat recovery efficiency target value is... No less than 70%. These target values ​​are based on historical operational data statistics and engineering safety specifications, ensuring reasonable and differentiated performance expectations under different risk levels.

[0101] The system will generate the implementation effect parameters from the previous step. The antifreeze performance index, namely the minimum wall temperature. , and the threshold of the expected antifreeze effect obtained Compare and calculate the first deviation value. This deviation value directly reflects the achievement of the antifreeze safety target. Simultaneously, the system will implement effect parameters. The waste heat recovery efficiency index, i.e., the actual waste heat recovery efficiency. =With the expected waste heat recovery efficiency target value Compare and calculate the second deviation value. This deviation value quantifies the degree to which energy utilization efficiency has been achieved.

[0102] ,

[0103] ,

[0104] in, The value represents the deviation in antifreeze effect, expressed in degrees Celsius. A negative value indicates that the actual antifreeze effect did not meet expectations, posing a potential risk. This represents the deviation in waste heat recovery efficiency, and is dimensionless. A negative value indicates that the energy recovery efficiency is lower than the target level.

[0105] The system calculates the first deviation value. Second deviation value The magnitude and direction of the vector, combined with the current freezing-ice risk level R, generate a graded correction vector. This vector not only contains the specific numerical value of the deviation, but its structure or encoding method also implicitly contains the adjustment priority and strategy. For example, when A large negative value indicates a serious deficiency in freeze protection safety; in this case, a graded correction vector is needed. The priority will be given to adjusting towards "improving antifreeze strength," even if this might mean sacrificing some recovery efficiency. Conversely, if... For positive and If the value is negative, the vector will point in the direction of "optimizing recycling efficiency while ensuring safety".

[0106] ,

[0107] in, This represents the final generated hierarchical correction vector. It combines the two deviation values. , This vector is encapsulated with the icing risk level R that triggered the control. This vector will serve as input, driving subsequent learning modules to adaptively optimize the risk level mapping rules, forming a complete feedback correction closed loop. For example, if frequent occurrences are found under level 3 risk... In such cases, the system may subsequently adjust the feature thresholds for identifying level 3 risks to be more conservative.

[0108] Optionally, optimizing the grade mapping process using the graded correction vector includes:

[0109] The graded correction vector is input into the adjustment interface of the risk level mapping rule base;

[0110] Based on the deviation information and risk level association information carried in the graded correction vector, the feature threshold or membership function shape corresponding to the risk level in the risk level mapping rule base is corrected.

[0111] Obtain historical optimization records and use these records to verify the stability of the revised risk level mapping rule base.

[0112] Specifically, the hierarchical correction vector generated in the previous step... This signal is transmitted as an input to the preset adjustment interface of the risk level mapping rule base. This interface is a software module specifically designed for online updates and optimization of the rule base, ensuring that modifications to the core logic are made within a controlled and traceable framework.

[0113] Analyzing hierarchical correction vectors Deviation information carried in , And the associated risk level R. Based on this information, the system triggers a rule adjustment algorithm. For example, if the vector shows a deviation in antifreeze effectiveness at risk level R=3. If the value remains negative, indicating insufficient freeze protection, the algorithm will locate all rules in the risk level mapping rule base that output "Risk Level 3". Then, it will modify the preconditions of these rules to make them more sensitive. Specifically, this could involve lowering the threshold of the key components in the dimensionality-reduced feature vector Z required to trigger this level, or adjusting the shape of the corresponding fuzzy membership function so that its center point moves towards lower feature values, allowing the system to classify similar working conditions as Level 3 risk at an earlier stage. This modification process can be illustrated by the following formula:

[0114] ,

[0115] in, This represents the modified membership function. It is a primitive function. It is based on the hierarchical correction vector The calculated adjustment amount. When When it is negative, A positive value shifts the membership function to the left, allowing smaller feature values ​​z to acquire higher membership degrees, thus making it easier to trigger the corresponding risk level. The adjustment magnitude is proportional to the size of the bias and can be controlled by a learning rate parameter to prevent the system from over-adjusting.

[0116] After revising the risk level mapping rule base, the system does not immediately use it as the final version. Instead, it first retrieves historical optimization records stored in the database, which contain all the level correction vectors from the past period. The system uses this historical data to perform an offline stability check on the revised rule base. The check can involve using the historical icing sensitivity feature matrix as input, simulating under the new rule base, and observing whether the output icing risk level sequence is smoother than before, or whether the predicted control effect deviation shows a convergence trend. For example, it can check whether, after adjustment, historically severe icing insufficiency conditions can be correctly identified as higher risk levels. Only when the check results meet preset stability indicators, such as a decrease in the moving average of the adjusted deviation, will the newly revised risk level mapping rule base be officially activated.

[0117] Optionally, after optimizing the grade mapping process using the graded correction vector, the process further includes:

[0118] The hierarchical correction vector sequence generated during the historical optimization process is obtained, and time series analysis and feature extraction are performed to generate a correction trend feature vector.

[0119] Based on the correction trend feature vector, the dynamic response hysteresis characteristics between the icing risk level and the waste heat recovery scheduling vector are identified, and feedforward compensation parameters are generated.

[0120] In the next control cycle, the waste heat resource state vector is corrected using the feedforward compensation parameters.

[0121] Specifically, the system retrieves hierarchical correction vectors generated in the historical database over a relatively long period, such as the last 72 hours. The system analyzes and extracts features from the time series data, such as calculating the antifreeze effect deviation in the correction vector. The moving average, rate of change (first derivative), and variance of the components. These statistics constitute a correction trend feature vector that describes the overall trend of the recent system correction behavior. :

[0122] ,

[0123] in, This represents the characteristic vector of the corrected trend. It is the moving average of the historical antifreeze effect deviation sequence, reflecting the average correction direction and magnitude. The gradient or rate of change of the sequence reflects how fast the deviation changes. The variance of the sequence reflects the effectiveness of the control stability.

[0124] The system is based on the generated correction trend feature vector By combining the change sequence of the icing risk level R with the execution sequence of the waste heat recovery scheduling vector D within the same period, a correlation analysis is performed to identify the dynamic response lag characteristics between the two. For example, by analyzing the time required from the issuance of the scheduling command to the actual wall temperature reaching the new safety threshold when the risk level jumps from level 2 to level 4, and the deviation trend generated during this process, a dynamic model describing the system's response delay and insufficiency is established. This model ultimately outputs a set of feedforward compensation parameters. :

[0125] ,

[0126] in, This represents the feedforward compensation parameter, which may be a vector containing the overshoot thermal gain coefficient and the advance action time constant. It is a mapping model trained offline from historical data, such as a small neural network or radial basis function network, which establishes a nonlinear relationship between the correction trend and the required compensation amount.

[0127] At the start of the next control cycle, before the collaborative scheduling calculation module performs optimization calculations, the system first utilizes the feedforward compensation parameters calculated in the previous cycle. For the current real-time acquired waste heat resource state vector Make corrections to generate a temporary state vector for optimization. This correction does not change the physical properties of the resource, but rather adjusts its equivalent value or availability in the optimization model.

[0128] ,

[0129] in, It is the corrected waste heat resource state vector. It is a correction function. The specific correction operation can be based on... The heat gain coefficient included in the calculation is used to appropriately amplify the heat requirement of the antifreeze target during optimization; or, based on the advance action time constant, a penalty term related to the response time is added to the multi-objective optimization function to guide the optimization algorithm to select waste heat resources with faster response or more direct transportation paths.

[0130] Based on the same inventive concept, such as Figure 6 As shown, the present invention also provides a waste heat recovery and antifreeze control system driven by the risk classification of cooling tower icing, the system comprising:

[0131] The data acquisition and fusion module is used to acquire environmental wind speed parameters, environmental humidity parameters, cooling circuit inlet water temperature parameters, cooling circuit return water temperature parameters, and health parameters of the heat exchange elements in the tower, and to perform fusion processing to generate an icing-sensitive feature matrix.

[0132] The risk grading and assessment module is used to identify risk factors and map levels to the icing-sensitive feature matrix to generate icing risk levels.

[0133] The collaborative scheduling calculation module is used to perform collaborative optimization calculations based on the icing risk level and the currently available waste heat resources to generate a waste heat recovery scheduling vector.

[0134] The instruction generation and execution module is used to convert the waste heat recovery scheduling vector into control parameters for the cooling tower actuator and generate an antifreeze control instruction set.

[0135] The effect monitoring and feedback module is used to execute the antifreeze control command set, collect the system status parameters after execution, and generate implementation effect parameters;

[0136] An online learning module is used to calculate the deviation between the implementation effect parameters and the expected antifreeze recovery effect, and generate a graded correction vector;

[0137] The correction module is used to optimize the grade mapping process using the grade correction vector.

[0138] To verify the feasibility of this invention in practice, it was applied to the cooling system of a data center. This data center operates continuously year-round and faces a severe challenge in preventing its cooling towers from freezing during winter. Traditional anti-freezing methods mainly rely on energy-intensive electric heating elements or shutting down some cooling tower units during the colder nighttime hours, resulting in high operating costs and limited cooling capacity. This embodiment aims to utilize the waste heat generated by the data center server units to achieve intelligent and energy-efficient anti-freezing control of the cooling towers by applying this invention.

[0139] The test object was a counter-flow cooling tower in the data center. The test took place at 23:00 on a winter night, when the load was stable and the environmental conditions were harsh. The control system performed the control cycle according to the method of this invention.

[0140] The system first collects real-time operating parameters using weather stations and pipeline sensors deployed around the cooling tower. The current ambient wind speed is 6.0 m / s, and the ambient humidity is 25%. The system's wind speed range is set to 0-15 m / s, and the humidity range is assumed to be 20%-100%. Minimum-maximum normalization is then performed: normalized wind speed. Standardized humidity Based on the region's climate characteristics, wind speed weighting Set to 0.7, humidity weight. Set to 0.3. Calculate the environmental state parameter E according to the formula: .

[0141] The inlet water temperature of the cooling circuit was collected. Return water temperature Calculate the temperature difference: Based on the cooling tower design parameters and fluid characteristics, the proportionality coefficient C = 1.1 and the environmental impact gain coefficient k = 0.5 are set. The cooling load characteristic parameter L is calculated using the formula: .

[0142] By long-term monitoring and model estimation of the pressure difference in the tower packing, the current health parameter H of the heat exchange element is obtained as 0.88, indicating slight scaling or blockage. The above parameters are spatiotemporally aligned at the current time t to generate the icing sensitivity feature vector at that moment. As a row in the icing-sensitive feature matrix: .

[0143] The system performs principal component analysis on the continuously acquired icing-sensitive feature matrix, and then... After standardization, the coordinates are projected onto a pre-trained principal component coordinate system. Assume the analysis results show that the first two principal components... Mainly reflects the overall cooling intensity and The cumulative variance contribution rate, primarily reflecting the interaction between equipment status and the environment, reaches 96%, meeting the dimensionality reduction requirements. The dimensionality-reduced feature vector obtained in this calculation is: .

[0144] Will Input the preset fuzzy membership function. Calculate the result: The membership degree for "high-risk contribution" is 0.85; The membership degree for "medium-risk contribution" is 0.15; The membership degree for "low-risk contribution" is 0.60.

[0145] The system calls the risk level mapping rule base, one of the rules being: "IF Belonging to 'high-risk contribution' AND "Not belonging to 'high-risk contribution' THEN icing risk level is 4". Based on the above membership set, the fuzzy inference engine activates this and other related rules, and after defuzzification calculations such as the centroid method, finally outputs the discretized icing risk level R=4, i.e., high risk.

[0146] The system identifies the available waste heat resource as the liquid cooling circulation loop in the server room, and its state vector... for That is, the temperature is 42℃ and the maximum available flow rate is 8m³. 3 / h.

[0147] The system uses R=4 as a constraint condition, requiring the prediction of wall temperature in critical areas of the cooling tower after freeze protection. This refers to the safety threshold corresponding to level 4 risk. Weights are assigned based on the objectives of maximizing waste heat utilization efficiency and minimizing pump energy consumption. , The waste heat resource state vector Substitute the model into the optimization model and call the particle swarm optimization algorithm to solve it.

[0148] The result of the optimization calculation is the waste heat recovery scheduling vector D. This vector indicates the flow rate that should be drawn from the liquid cooling circuit. Hot water, immediately ( Injected into the cooling tower's water collection pan antifreeze circuit, duration minute. .

[0149] Based on the scheduling vector D, the system calculates the target speed of the antifreeze loop circulating pump to be 38Hz, corresponding to a target opening degree of 82% for the regulating valve. These parameters are encapsulated into an antifreeze control instruction set and sent to the PLC for execution. Five minutes after instruction execution, the effect monitoring and feedback module begins collecting data. The lowest wall temperature among all measuring points is collected. The actual waste heat recovery efficiency is calculated by measuring the temperature difference and flow rate of the injected hot water. The above data constitutes the implementation effect parameters. .

[0150] Obtain the expected target corresponding to Level 4 risk: Expected antifreeze effect threshold. Target value for expected waste heat recovery efficiency Calculate the first deviation value Calculate the second deviation value. .

[0151] The system generates a graded correction vector based on the deviation value and the current risk level. The vector is fed into the online learning module. Because... A positive value indicates that the antifreeze effect exceeds expectations, while A negative value indicates that energy efficiency has not met the standards, and the system judges that the current response to Level 4 risk is too conservative. Based on this, the correction module fine-tunes the risk level mapping rule base, slightly increasing the feature threshold that triggers Level 4 risk. This means that under similar operating conditions in the future, the system may output Level 3 risk or slightly reduce the amount of waste heat that needs to be dispatched, in order to seek a better balance between safety and energy efficiency.

[0152] In summary, the method of this invention transforms the parameters of multi-source heterogeneity into a risk level, such as level 4. Based on this level, the system automatically performs optimization calculations to generate a waste heat scheduling scheme, allocating 4.5 m³ of waste heat. 3 The system successfully implemented a flow rate of / h. The execution results show that the anti-freeze safety target was achieved, with an actual minimum wall temperature of 2.8℃, exceeding the 2.0℃ threshold, thus protecting the cooling tower. Through a closed-loop feedback correction mechanism, the system identified optimization space in the control strategy and generated a correction vector. By fine-tuning the risk assessment model, this invention demonstrates its ability to continuously optimize energy efficiency while ensuring safety. Compared to traditional fixed threshold control or manual intervention, it improves the intelligence and economy of cooling tower operation in winter.

[0153] It should be noted that the above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the invention. All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of the invention upon considering the disclosure of the specification and practical truths. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing, characterized in that, include: The system acquires and merges environmental wind speed parameters, environmental humidity parameters, cooling circuit inlet water temperature parameters, cooling circuit return water temperature parameters, and the health parameters of the heat exchange elements in the tower body to generate an icing-sensitive feature matrix. The icing-sensitive feature matrix is ​​subjected to risk factor identification and level mapping to generate an icing risk level. Based on the icing risk level and the current waste heat resource information, a collaborative optimization calculation is performed to generate a waste heat recovery scheduling vector; The waste heat recovery scheduling vector is converted into control parameters for the cooling tower actuator, and an antifreeze control instruction set is generated. Execute the antifreeze control command set, collect the system status parameters after execution, and generate implementation effect parameters; Calculate the deviation between the implementation effect parameters and the expected antifreeze recovery effect, and generate a graded correction vector; The grade mapping process is optimized using the aforementioned graded correction vector; The generated icing-sensitive feature matrix includes: The environmental wind speed and humidity parameters of the area where the cooling tower is located are obtained, and dimensionless processing is performed to generate environmental state parameters. The inlet water temperature parameters at the tower body inlet and the return water temperature parameters at the tower body outlet of the cooling circuit are obtained, the temperature difference value is calculated, and combined with the environmental state parameters, a weighted fusion is performed to generate the cooling load characteristic parameters. Obtain health parameters characterizing the degree of scaling, blockage, or damage to the heat exchange elements of the tower, and perform spatiotemporal alignment and correlation coupling between the health parameters and the cooling load characteristic parameters to generate an icing-sensitive feature matrix; The icing-sensitive feature matrix is ​​subjected to risk factor identification and level mapping processing to generate icing risk levels, including: Principal component analysis was performed on the icing-sensitive feature matrix to extract the key feature dimensions that affect the icing risk and generate a dimensionality-reduced feature vector. Based on the reduced feature vector, the contribution of each feature dimension to the icing risk is identified by a preset fuzzy membership function, and a risk factor membership set is generated. Based on the set of membership degrees of the risk factors, matching and mapping are performed through a preset risk level mapping rule base to output a discrete icing risk level.

2. The waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing as described in claim 1, characterized in that, Based on the icing risk level and current waste heat resource information, collaborative optimization calculations are performed to generate a waste heat recovery scheduling vector, including: The temperature, flow, and location distribution parameters of the waste heat resources used for frost prevention are obtained, and an availability assessment is conducted to generate a waste heat resource state vector. Using the aforementioned icing risk level as a constraint, and aiming to maximize waste heat utilization efficiency and minimize antifreeze energy consumption, a multi-objective optimization function is established. The waste heat resource state vector is input into the multi-objective optimization function for solution, and the waste heat and commissioning sequence allocated to different cooling tower circuits or areas are calculated to generate a waste heat recovery scheduling vector.

3. The waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing as described in claim 2, characterized in that, The waste heat recovery scheduling vector is converted into control parameters for the cooling tower actuator, and the antifreeze control instruction set is generated, including: Based on the waste heat distribution information in the waste heat recovery scheduling vector, calculate the opening control parameters of the corresponding regulating valve and the speed control parameters of the circulating pump; Based on the commissioning timing information in the waste heat recovery scheduling vector and combined with the response characteristics of each actuator of the cooling tower, a timing control logic sequence is generated. The opening control parameters, speed control parameters, and timing control logic sequence are encapsulated and format-converted to generate an anti-freeze control instruction set for the direct-drive actuator.

4. The waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing as described in claim 3, characterized in that, The antifreeze control command set is executed, and the system status parameters after execution are collected to generate implementation effect parameters, including: The antifreeze control command set is executed on the regulating valve and circulating pump of the cooling tower, and the actual feedback parameters of each actuator are collected in real time. Collect and execute commands to obtain wall temperature parameters of key areas of the cooling tower, real-time water temperature change parameters of the cooling circuit, and environmental condition change parameters. The actual feedback parameters, wall temperature parameters, real-time water temperature change parameters, and environmental state change parameters are fused and analyzed to generate implementation effect parameters that characterize the immediate effect of antifreeze measures.

5. The waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing as described in claim 4, characterized in that, Calculate the deviation between the implementation effect parameters and the expected antifreeze recovery effect, and generate a graded correction vector including: Obtain the expected antifreeze effect threshold and the expected waste heat recovery efficiency target value corresponding to the current icing risk level; The antifreeze effect index in the implementation effect parameters is compared with the expected antifreeze effect threshold, and a first deviation value is calculated; The waste heat recovery efficiency index in the implementation effect parameters is compared with the expected waste heat recovery efficiency target value, and a second deviation value is calculated. Based on the magnitude and direction of the first and second deviation values, and in conjunction with the icing risk level, a graded correction vector is generated.

6. The waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing as described in claim 2, characterized in that, Optimizing the grade mapping process using the graded correction vector includes: The graded correction vector is input into the adjustment interface of the risk level mapping rule base; Based on the deviation information and risk level association information carried in the graded correction vector, the feature threshold or membership function shape corresponding to the risk level in the risk level mapping rule base is corrected. Obtain historical optimization records and use these records to verify the stability of the revised risk level mapping rule base.

7. The waste heat recovery and antifreeze control method driven by the risk classification of cooling tower icing as described in claim 6, characterized in that, After optimizing the grade mapping process using the aforementioned graded correction vector, the process further includes: The hierarchical correction vector sequence generated during the historical optimization process is obtained, and time series analysis and feature extraction are performed to generate a correction trend feature vector. Based on the correction trend feature vector, the dynamic response hysteresis characteristics between the icing risk level and the waste heat recovery scheduling vector are identified, and feedforward compensation parameters are generated. In the next control cycle, the waste heat resource state vector is corrected using the feedforward compensation parameters.

8. A waste heat recovery and antifreeze control system driven by the risk classification of cooling tower icing, characterized in that, include: The data acquisition and fusion module is used to acquire environmental wind speed parameters, environmental humidity parameters, cooling circuit inlet water temperature parameters, cooling circuit return water temperature parameters, and health parameters of the heat exchange elements in the tower, and to perform fusion processing to generate an icing-sensitive feature matrix. The risk grading and assessment module is used to identify risk factors and map levels to the icing-sensitive feature matrix to generate icing risk levels. The collaborative scheduling calculation module is used to perform collaborative optimization calculations based on the icing risk level and the current waste heat resource information to generate a waste heat recovery scheduling vector. The instruction generation and execution module is used to convert the waste heat recovery scheduling vector into control parameters for the cooling tower actuator and generate an antifreeze control instruction set. The effect monitoring and feedback module is used to execute the antifreeze control command set, collect the system status parameters after execution, and generate implementation effect parameters; An online learning module is used to calculate the deviation between the implementation effect parameters and the expected antifreeze recovery effect, and generate a graded correction vector; The correction module is used to optimize the grade mapping process using the graded correction vector; the generation of the icing-sensitive feature matrix includes: The environmental wind speed and humidity parameters of the area where the cooling tower is located are obtained, and dimensionless processing is performed to generate environmental state parameters. The inlet water temperature parameters at the tower body inlet and the return water temperature parameters at the tower body outlet of the cooling circuit are obtained, the temperature difference value is calculated, and combined with the environmental state parameters, a weighted fusion is performed to generate the cooling load characteristic parameters. Obtain health parameters characterizing the degree of scaling, blockage, or damage to the heat exchange elements of the tower, and perform spatiotemporal alignment and correlation coupling between the health parameters and the cooling load characteristic parameters to generate an icing-sensitive feature matrix; The icing-sensitive feature matrix is ​​subjected to risk factor identification and level mapping processing to generate icing risk levels, including: Principal component analysis was performed on the icing-sensitive feature matrix to extract the key feature dimensions that affect the icing risk and generate a dimensionality-reduced feature vector. Based on the reduced feature vector, the contribution of each feature dimension to the icing risk is identified by a preset fuzzy membership function, and a risk factor membership set is generated. Based on the set of membership degrees of the risk factors, matching and mapping are performed through a preset risk level mapping rule base to output a discrete icing risk level.

Images