Cooling tower self-adaptive regulation and control method and system based on artificial intelligence

Abstract

The invention discloses a cooling tower self-adaptive regulation and control method and system based on artificial intelligence, and relates to the technical field of cooling tower regulation and control. Performing clustering analysis based on the distance of the time sequence fragments; analyzing a clustering category of the to-be-predicted feature group, and analyzing a neighbor set of the to-be-predicted feature group; analyzing the participation weight of each historical feature group in the neighbor set, and combining the participation weight to construct a regression model; the cooling capacity deviation between the predicted cooling capacity and the cooling capacity before the target cooling capacity is set is analyzed, and the new control capacity is analyzed by combining the cooling capacity deviation in the current control period and controller parameters; and collecting actual cooling capacity in real time, synchronously recording a normalized current feature group, analyzing a model prediction residual error, establishing a data buffer area, and updating a regression coefficient set and controller parameters based on a dynamic data set. The prediction precision is improved, and long-term efficient operation of the system is maintained.

Description

Technical Field

[0001] The present invention relates to the technical field of cooling tower control, and in particular to an artificial intelligence-based adaptive control method and system for a cooling tower. Background Art

[0002] The application of cooling towers originated in the early stages of industrialization, especially the development of the power industry. With the development of the times, the cooling tower industry began to develop digitally and intelligently, introducing technologies such as remote monitoring and big data analysis, and using advanced monitoring technologies to improve equipment operating efficiency.

[0003] Current cooling tower control systems generally rely on fixed thresholds or simple rules to categorize operating conditions. These fail to effectively capture the dynamic coupling between various parameters and ignore deformation characteristics over time, resulting in operating condition classifications that deviate from the actual operating state. Traditional controller parameters are fixed over time and cannot adapt to environmental changes and equipment aging. When actual cooling demand fluctuates, fixed parameters can easily cause control oscillation or response lag, often leading to inefficient coordination between fans and pumps. Frequent triggering of ineffective adjustments increases equipment energy consumption and mechanical wear.

[0004] Therefore, the present invention discloses an artificial intelligence-based cooling tower adaptive control method and system to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide an artificial intelligence-based cooling tower adaptive control method and system to solve the problems raised in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solution: an artificial intelligence-based cooling tower adaptive control method, the method comprising the following steps: S1: Synchronously collect historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical pump flow rate, normalize the historical data, and perform cluster analysis based on the distance between time series segments; S2: Analyze the clustering category of the feature group to be predicted, analyze the neighbor set of the feature group to be predicted; analyze the participation weight of each historical feature group in the neighbor set, and build a regression model based on the participation weight; S3: Analyze the cooling amount deviation between the predicted cooling amount and the set target cooling amount, analyze the new control amount based on the cooling amount deviation and controller parameters in the current control cycle, and send the new control instruction to the actuator; S4: Collect the actual cooling amount in real time and synchronously record the normalized current feature group, analyze the model prediction residual, establish a data buffer, and update the regression coefficient set and controller parameters based on the dynamic data set.

[0007] According to the above scheme, S1 includes the following contents: S101: Obtain historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical water pump flow rate to generate a historical data set; perform normalization processing on each historical data according to the corresponding minimum and maximum values ​​to obtain a normalized historical data set; S102: Based on the normalized historical data set, sequentially intercepting continuous time series segments along the time axis using a sliding window of a preset length and a step size, and calculating the dynamic time warping distance between any two time series segments; Randomly select a preset number of sample points from the sample set consisting of all time series samples as initial cluster centers. For each sample point, extract the dynamic time warping distance value between the sample point and all initial cluster centers. Assign the sample point to the category corresponding to the initial cluster center with the smallest dynamic time warping distance value, and record the cluster label of the current iteration for each sample point. For each cluster category, the sum of the dynamic time warping distances of each sample point to all other sample points in the same cluster category is analyzed, and the sample point with the smallest sum of the dynamic time warping distances is used as the cluster center point after iteration; the sample point allocation and cluster center point iteration operations are repeated until the positions of all cluster center points remain unchanged for two consecutive iterations or reach the preset maximum iteration round threshold; the clustering result data is generated and stored.

[0008] The present invention uses dynamic time warping distance to measure the similarity of time series segments, effectively solves the problems of scaling and offset of time series on the time axis, and improves the accuracy of working condition division. Iterative clustering based on dynamic time warping distance can adaptively identify typical operating modes in historical data, laying the foundation for subsequent working condition control.

[0009] According to the above scheme, S2 includes the following contents: S201: Obtain the normalized feature set to be predicted X = (T env , T wb ,△T,V f , V p ), where T env Represents the normalized value of the ambient temperature, T wb represents the normalized value of wet bulb temperature, △T represents the normalized value of the inlet and outlet water temperature difference, V f Indicates the normalized value of the fan frequency, V p Represents the normalized value of the water pump flow rate; analyzes the corresponding cluster category; analyzes the dynamic time-warped distance between the feature group to be predicted and the historical feature group at each historical moment in the corresponding cluster category; sorts in ascending order of distance, and selects the I historical feature groups with the smallest distance to form the nearest neighbor set N I ={X i|l∈[1,I]}, I is a preset constant, i is a positive integer; S202: Based on the dynamic time warping distance between the feature group to be predicted and the historical feature groups in the neighboring set, the participation weight of each historical feature group is analyzed, and the participation weight of the i-th historical feature group is recorded as W i , W i =exp(-D i 2 / h 2 ); where D i represents the dynamic time warping distance between the i-th historical feature group and the feature group to be predicted, and h represents the preset participation coefficient; S203: Obtain the historical cooling amount corresponding to each historical feature group in the neighbor set, and record the historical cooling amount of the i-th historical feature group as Q i , based on the historical feature group, historical cooling amount and corresponding participation weight, the regression coefficient set {β0, β1, β2} is solved by minimizing the weighted residual sum of squares; the specific formula is as follows: ; Construct a regression model based on a set of regression coefficients: ; Substitute the feature group to be predicted X into the regression model to output the predicted cooling amount Q.

[0010] This method first locates the cluster category to which the predicted data belongs, then searches for nearest neighbors within that category, significantly narrowing the search scope, improving real-time performance, and ensuring data homogeneity. Similar samples are given higher weights through exponential weighting, enhancing local fitting capabilities. The regression coefficient is solved by minimizing the weighted sum of squared residuals, significantly improving the accuracy of cooling capacity prediction. Only the clustering results and nearest neighbor samples need to be stored, eliminating the need for complex neural networks, reducing computing resource requirements, and making it suitable for embedded deployment.

[0011] According to the above solution, S3 includes the following: S301: extracting the set target cooling capacity and calculating the cooling capacity deviation in real time based on the predicted cooling capacity; calling pre-stored controller parameters according to the cluster category corresponding to the feature group to be predicted; the controller parameters include fan control parameters and water pump control parameters; the fan control parameters and water pump control parameters each include a proportional coefficient and an integral coefficient; S302: Analyze the real-time fan frequency increment △f and the real-time pump flow increment △q based on the cooling amount deviation and the corresponding controller parameters. ; ; Among them, Ka f Indicates the proportional coefficient in the fan control parameters, Kb fIndicates the integral coefficient in the fan control parameters, Ka q Indicates the proportional coefficient in the pump control parameters, Kb q Represents the integral coefficient in the water pump control parameters; e represents the cooling capacity deviation, ∑e represents the accumulated value of the historical cooling capacity deviation in the current control cycle; △t is the preset control cycle time step; The fan frequency and pump flow execution quantities of the previous control cycle are read, the new control quantity is analyzed in combination with the real-time fan frequency increment and the real-time pump flow increment, and the new control instruction is sent to the actuator.

[0012] This invention pre-stores independent controller parameters for different cluster categories, achieving precise "one-category-one-policy" control. The control variable is generated based on the cooling capacity deviation and its integral term, dynamically eliminating steady-state errors, with fast response speed and strong anti-disturbance performance.

[0013] According to the above scheme, S4 includes the following contents: S401: Collect the actual cooling amount in real time and synchronously record the normalized current feature group. Subtract the predicted cooling amount from the actual cooling amount and record it as the model prediction residual. Establish a first-in-first-out data buffer of length N. Store the current feature group and the corresponding model prediction residual as a data pair in the buffer. When the number of data pairs in the buffer exceeds N, remove the earliest historical data pair and obtain the most recent N data pairs to form a dynamic data set. Based on the dynamic data set, recursively update the regression coefficient set and controller parameters using the least squares method. S402: Feeding back the updated regression coefficient set to the regression model to replace the original regression coefficients, and feeding back the updated controller parameters to the controller to replace the original controller parameters.

[0014] This invention utilizes recursive least squares to update regression coefficients and controller parameters in real time, addressing model drift caused by equipment aging, scaling, and other factors. A first-in, first-out buffer retains only the most recent set of data, ensuring the model continuously tracks the system's latest status and preventing the impact of outdated historical data. The residual difference between actual cooling capacity and predicted values ​​is fed back to the model and controller, forming a closed "perception-prediction-control-calibration" loop to maintain efficient system operation over the long term.

[0015] Another aspect of the present application is an artificial intelligence-based cooling tower adaptive control system, which is applied to the above-mentioned artificial intelligence-based cooling tower adaptive control method, and the system includes a data acquisition and clustering module, a cooling capacity prediction module, a control instruction analysis module, and a parameter update module; The data collection and clustering module is used to synchronously collect historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency and historical water pump flow rate, and perform cluster analysis based on the distance of time series segments after normalizing the historical data; The cooling amount prediction module is used to analyze the clustering category of the feature group to be predicted, analyze the neighboring set of the feature group to be predicted; analyze the participation weight of each historical feature group in the neighboring set, and construct a regression model based on the participation weight; The control instruction analysis module is used to analyze the cooling amount deviation between the predicted cooling amount and the set target cooling amount, analyze the new control amount in combination with the cooling amount deviation and controller parameters in the current control cycle, and send the new control instruction to the actuator; The parameter update module is used to collect the actual cooling amount in real time and synchronously record the normalized current feature group, analyze the model prediction residual, establish a data buffer, and update the regression coefficient set and controller parameters based on the dynamic data set.

[0016] According to the above solution, the data acquisition and clustering module includes a cooling tower data acquisition unit and a data clustering unit; The cooling tower data acquisition unit is used to obtain historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical water pump flow rate to generate a historical data set; each historical data is normalized according to the corresponding minimum and maximum values ​​to obtain a normalized historical data set; The data clustering unit is used to intercept continuous time series segments along the time axis based on the normalized historical data set, using a sliding window of preset length and step size, and calculate the dynamic time regularization distance between any two time series segments; repeat the sample point allocation and cluster center point iterative operations until the positions of all cluster center points remain unchanged for two consecutive iterations or reach a preset maximum iteration round threshold; generate and store clustering result data.

[0017] According to the above solution, the cooling amount prediction module includes a participation weight analysis unit and a prediction model construction unit; The participation weight analysis unit is used to obtain the normalized feature group to be predicted and analyze the corresponding cluster category; analyze the dynamic time warping distance between the feature group to be predicted and the historical feature group at each historical moment in the corresponding cluster category; sort the distances in ascending order to construct a neighbor set, and analyze the participation weight of each historical feature group based on the dynamic time warping distance between the feature group to be predicted and the historical feature groups in the neighbor set; The prediction model construction unit is used to obtain the historical cooling amount corresponding to each historical feature group in the neighbor set, solve the regression coefficient set by minimizing the weighted residual square sum based on the historical feature group, the historical cooling amount and the corresponding participation weight, and construct a regression model based on the regression coefficient set.

[0018] According to the above solution, the control instruction analysis module includes a controller parameter calling unit and a control instruction analysis unit; The controller parameter calling unit is used to extract the set target cooling amount and calculate the cooling amount deviation in real time based on the predicted cooling amount; call the pre-stored controller parameters according to the cluster category corresponding to the feature group to be predicted; the controller parameters include fan control parameters and water pump control parameters; the fan control parameters and water pump control parameters each include a proportional coefficient and an integral coefficient; The control instruction analysis unit is used to analyze the real-time fan frequency increment and the real-time water pump flow increment based on the cooling amount deviation and the corresponding controller parameters; read the fan frequency and water pump flow execution amount of the previous control cycle, analyze the new control amount in combination with the real-time fan frequency increment and the real-time water pump flow increment, and send the new control instruction to the actuator.

[0019] According to the above solution, the parameter updating module includes a parameter analyzing unit and a parameter transmitting unit; The parameter analysis unit is used to collect the actual cooling amount in real time and simultaneously record the normalized current feature group, subtract the predicted cooling amount from the actual cooling amount as the model prediction residual, establish a first-in-first-out data buffer, store the current feature group and the corresponding model prediction residual as a data pair in the buffer, and obtain a dynamic data set; based on the dynamic data set, recursively update the regression coefficient set and controller parameters using the least squares method; The parameter transmission unit is used to feed back the updated regression coefficient set to the regression model to replace the original regression coefficient, and feed back the updated controller parameters to the controller to replace the original controller parameters.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention uses dynamic time warping distance to measure the similarity of time series segments, effectively solves the problems of scaling and offset of time series on the time axis, improves the accuracy of working condition division, and iterative clustering based on dynamic time warping distance can adaptively identify typical operating modes in historical data, laying the foundation for subsequent working condition control. The present invention first locates the cluster category to which the data to be predicted belongs, and then searches for neighbors within the category, greatly narrowing the search scope, improving real-time performance and ensuring data homogeneity. Similar samples are given higher weights through exponential weights to enhance local fitting capabilities. The regression coefficient is solved by minimizing the weighted residual sum of squares, significantly improving the accuracy of predicted cooling capacity. Only clustering results and neighbor samples need to be stored, without the need for complex neural networks, reducing computing resource requirements, and being suitable for embedded deployment. The present invention pre-stores independent controller parameters for different cluster categories to achieve precise control of "one category, one policy". The control quantity is generated based on the cooling quantity deviation and its integral term, dynamically eliminating steady-state errors, with fast response speed and strong anti-disturbance performance. This invention utilizes recursive least squares to update regression coefficients and controller parameters in real time, addressing model drift caused by equipment aging, scaling, and other factors. A first-in, first-out buffer retains only the most recent set of data, ensuring the model continuously tracks the system's latest status and preventing the impact of outdated historical data. The residual difference between actual cooling capacity and predicted values ​​is fed back to the model and controller, forming a closed "perception-prediction-control-calibration" loop to maintain efficient system operation over the long term. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 Schematic diagram of the process of the cooling tower adaptive control method based on artificial intelligence of the present invention; Figure 2 The figure is a schematic structural diagram of the cooling tower adaptive control system based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] Example 1: Please refer to Figure 1 The present invention provides a technical solution: an artificial intelligence-based cooling tower adaptive control method, the method comprising the following steps: S1: Synchronously collect historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical pump flow rate, normalize the historical data, and perform cluster analysis based on the distance between time series segments; In S1, the following are included: S101: Obtain historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical water pump flow rate to generate a historical data set; perform normalization processing on each historical data according to the corresponding minimum and maximum values ​​to obtain a normalized historical data set; S102: Based on the normalized historical data set, sequentially intercepting continuous time series segments along the time axis using a sliding window of a preset length and a step size, and calculating the dynamic time warping distance between any two time series segments; Randomly select a preset number of sample points from the sample set consisting of all time series samples as initial cluster centers. For each sample point, extract the dynamic time warping distance value between the sample point and all initial cluster centers. Assign the sample point to the category corresponding to the initial cluster center with the smallest dynamic time warping distance value, and record the cluster label of the current iteration for each sample point. For each cluster category, the sum of the dynamic time warping distances of each sample point to all other sample points in the same cluster category is analyzed, and the sample point with the smallest sum of the dynamic time warping distances is used as the cluster center point after iteration; the sample point allocation and cluster center point iteration operations are repeated until the positions of all cluster center points remain unchanged for two consecutive iterations or the preset maximum iteration round threshold is reached; the clustering result data is generated and stored.

[0024] S2: Analyze the clustering category of the feature group to be predicted, analyze the neighbor set of the feature group to be predicted; analyze the participation weight of each historical feature group in the neighbor set, and build a regression model based on the participation weight; In S2, the following are included: S201: Obtain the normalized feature set to be predicted X = (T env , T wb ,△T,V f , V p ), where T env Represents the normalized value of the ambient temperature, T wb represents the normalized value of wet bulb temperature, △T represents the normalized value of the inlet and outlet water temperature difference, V f Indicates the normalized value of the fan frequency, V p Represents the normalized value of the water pump flow rate; analyzes the corresponding cluster category; analyzes the dynamic time-warped distance between the feature group to be predicted and the historical feature group at each historical moment in the corresponding cluster category; sorts in ascending order of distance, and selects the I historical feature groups with the smallest distance to form the nearest neighbor set NI ={X i |l∈[1,I]}, I is a preset constant, i is a positive integer; S202: Based on the dynamic time warping distance between the feature group to be predicted and the historical feature groups in the neighboring set, the participation weight of each historical feature group is analyzed, and the participation weight of the i-th historical feature group is recorded as W i , W i =exp(-D i 2 / h 2 ); where D i represents the dynamic time warping distance between the i-th historical feature group and the feature group to be predicted, and h represents the preset participation coefficient; S203: Obtain the historical cooling amount corresponding to each historical feature group in the neighbor set, and record the historical cooling amount of the i-th historical feature group as Q i , based on the historical feature group, historical cooling amount and corresponding participation weight, the regression coefficient set {β0, β1, β2} is solved by minimizing the weighted residual sum of squares; the specific formula is as follows: ; Construct a regression model based on a set of regression coefficients: ; Substitute the feature group to be predicted X into the regression model to output the predicted cooling amount Q.

[0025] S3: Analyze the cooling amount deviation between the predicted cooling amount and the set target cooling amount, analyze the new control amount based on the cooling amount deviation and controller parameters in the current control cycle, and send the new control instruction to the actuator; In S3, include the following: S301: Extracting the set target cooling capacity and calculating the cooling capacity deviation in real time based on the predicted cooling capacity; calling pre-stored controller parameters according to the cluster category corresponding to the feature group to be predicted; the controller parameters include fan control parameters and water pump control parameters; the fan control parameters and water pump control parameters both include proportional coefficients and integral coefficients; S302: Analyze the real-time fan frequency increment △f and the real-time pump flow increment △q based on the cooling amount deviation and the corresponding controller parameters. ; ; Among them, Ka f Indicates the proportional coefficient in the fan control parameters, Kb f Indicates the integral coefficient in the fan control parameters, Ka q Indicates the proportional coefficient in the pump control parameters, Kb qRepresents the integral coefficient in the water pump control parameters; e represents the cooling capacity deviation, ∑e represents the accumulated value of the historical cooling capacity deviation in the current control cycle; △t is the preset control cycle time step; The fan frequency and pump flow execution quantities of the previous control cycle are read, the new control quantity is analyzed in combination with the real-time fan frequency increment and the real-time pump flow increment, and the new control instruction is sent to the actuator.

[0026] S4: Collect the actual cooling amount in real time and synchronously record the normalized current feature group, analyze the model prediction residual, establish a data buffer, and update the regression coefficient set and controller parameters based on the dynamic data set.

[0027] In S4, include the following: S401: Collect the actual cooling amount in real time and synchronously record the normalized current feature group. Subtract the predicted cooling amount from the actual cooling amount and record it as the model prediction residual. Establish a first-in-first-out data buffer of length N. Store the current feature group and the corresponding model prediction residual as a data pair in the buffer. When the number of data pairs in the buffer exceeds N, remove the earliest historical data pair and obtain the most recent N data pairs to form a dynamic data set. Based on the dynamic data set, recursively update the regression coefficient set and controller parameters using the least squares method. S402: Feeding back the updated regression coefficient set to the regression model to replace the original regression coefficients, and feeding back the updated controller parameters to the controller to replace the original controller parameters.

[0028] Example 2: Please refer to Figure 2 , the present invention provides a technical solution: another aspect of the present application is an artificial intelligence-based cooling tower adaptive control system, the system is applied to the above-mentioned artificial intelligence-based cooling tower adaptive control method, the system includes a data acquisition clustering module, a cooling capacity prediction module, a control instruction analysis module and a parameter update module; The data collection and clustering module is used to synchronously collect historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical pump flow rate. After normalizing the historical data, cluster analysis is performed based on the distance of time series segments. The cooling capacity prediction module is used to analyze the clustering categories of the feature group to be predicted and the neighboring set of the feature group to be predicted; analyze the participation weights of each historical feature group in the neighboring set and construct a regression model based on the participation weights; The control instruction analysis module is used to analyze the cooling amount deviation between the predicted cooling amount and the set target cooling amount, analyze the new control amount based on the cooling amount deviation and controller parameters in the current control cycle, and send the new control instruction to the actuator; The parameter update module is used to collect the actual cooling amount in real time and synchronously record the normalized current feature group, analyze the model prediction residual, establish a data buffer, and update the regression coefficient set and controller parameters based on the dynamic data set.

[0029] The data acquisition and clustering module includes a cooling tower data acquisition unit and a data clustering unit; The cooling tower data acquisition unit is used to obtain historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical water pump flow rate to generate a historical data set; each historical data is normalized according to the corresponding minimum and maximum values ​​to obtain a normalized historical data set; The data clustering unit is used to intercept continuous time series segments along the time axis based on the normalized historical data set, using a sliding window of preset length and step size, and calculate the dynamic time regularization distance between any two time series segments; repeat the sample point allocation and cluster center point iterative operations until the positions of all cluster center points remain unchanged for two consecutive iterations or reach the preset maximum iteration round threshold; generate and store clustering result data.

[0030] The cooling amount prediction module includes a participation weight analysis unit and a prediction model construction unit; The participation weight analysis unit is used to obtain the normalized feature group to be predicted and analyze the corresponding cluster category; analyze the dynamic time warping distance between the feature group to be predicted and the historical feature group at each historical moment in the corresponding cluster category; sort them in ascending order of distance, construct a neighbor set, and analyze the participation weight of each historical feature group based on the dynamic time warping distance between the feature group to be predicted and the historical feature groups in the neighbor set; The prediction model construction unit is used to obtain the historical cooling amount corresponding to each historical feature group in the neighbor set, solve the regression coefficient set by minimizing the weighted residual square sum based on the historical feature group, historical cooling amount and corresponding participation weight, and build a regression model based on the regression coefficient set.

[0031] The control instruction analysis module includes a controller parameter calling unit and a control instruction analysis unit; The controller parameter calling unit is used to extract the set target cooling capacity and calculate the cooling capacity deviation in real time based on the predicted cooling capacity; the pre-stored controller parameters are called according to the cluster category corresponding to the feature group to be predicted; the controller parameters include fan control parameters and water pump control parameters; both fan control parameters and water pump control parameters include proportional coefficients and integral coefficients; The control instruction analysis unit is used to analyze the real-time fan frequency increment and the real-time water pump flow increment based on the cooling amount deviation and the corresponding controller parameters; read the fan frequency and water pump flow execution quantities of the previous control cycle, analyze the new control quantity in combination with the real-time fan frequency increment and the real-time water pump flow increment, and send the new control instruction to the actuator.

[0032] The parameter updating module includes a parameter analyzing unit and a parameter transmitting unit; The parameter analysis unit is used to collect the actual cooling amount in real time and simultaneously record the normalized current feature group. The actual cooling amount minus the predicted cooling amount is recorded as the model prediction residual. A first-in-first-out data buffer is established. The current feature group and the corresponding model prediction residual are stored in the buffer as a data pair to obtain a dynamic data set. Based on the dynamic data set, the regression coefficient set and controller parameters are recursively updated using the least squares method. The parameter transmission unit is used to feed back the updated regression coefficient set to the regression model to replace the original regression coefficient, and feed back the updated controller parameters to the controller to replace the original controller parameters.

[0033] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0034] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. The cooling tower adaptive control method based on artificial intelligence is characterized by: The method comprises the following steps: S1: Synchronously collect historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical pump flow rate, normalize the historical data, and perform cluster analysis based on the distance between time series segments; S2: Analyze the clustering category of the feature group to be predicted, analyze the neighbor set of the feature group to be predicted; analyze the participation weight of each historical feature group in the neighbor set, and build a regression model based on the participation weight; S3: Analyze the cooling amount deviation between the predicted cooling amount and the set target cooling amount, analyze the new control amount based on the cooling amount deviation and controller parameters in the current control cycle, and send the new control instruction to the actuator; S4: Collect the actual cooling amount in real time and synchronously record the normalized current feature group, analyze the model prediction residual, establish a data buffer, and update the regression coefficient set and controller parameters based on the dynamic data set.

2. The cooling tower adaptive control method based on artificial intelligence according to claim 1 is characterized in that: In S1, the following are included: S101: Obtain historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency, and historical water pump flow rate to generate a historical data set; perform normalization processing on each historical data according to the corresponding minimum and maximum values; Get the normalized historical data set; S102: Based on the normalized historical data set, sequentially intercepting continuous time series segments along the time axis using a sliding window of a preset length and a step size, and calculating the dynamic time warping distance between any two time series segments; Randomly select a preset number of sample points from the sample set consisting of all time series samples as initial cluster centers. For each sample point, extract the dynamic time warping distance value between the sample point and all initial cluster centers. Assign the sample point to the category corresponding to the initial cluster center with the smallest dynamic time warping distance value, and record the cluster label of the current iteration for each sample point. For each cluster category, the sum of the dynamic time warping distances of each sample point to all other sample points in the same cluster category is analyzed, and the sample point with the smallest sum of the dynamic time warping distances is used as the cluster center point after iteration; the sample point allocation and cluster center point iteration operations are repeated until the positions of all cluster center points remain unchanged for two consecutive iterations or reach the preset maximum iteration round threshold; the clustering result data is generated and stored.

3. The cooling tower adaptive control method based on artificial intelligence according to claim 2 is characterized in that: In S2, the following are included: S2 01: Get the normalized feature set to be predicted X = (T env , T wb ,△T,V f , V p ), where T env Represents the normalized value of the ambient temperature, T wb represents the normalized value of wet bulb temperature, △T represents the normalized value of the inlet and outlet water temperature difference, V f Represents the normalized value of the fan frequency, V p Represents the normalized value of the water pump flow rate; analyzes the corresponding cluster category; analyzes the dynamic time-warped distance between the feature group to be predicted and the historical feature group at each historical moment in the corresponding cluster category; sorts in ascending order of distance, and selects the I historical feature groups with the smallest distance to form the nearest neighbor set N I ={X i |l∈[1,I]}, I is a preset constant, i is a positive integer; S202: Based on the dynamic time warping distance between the feature group to be predicted and the historical feature groups in the neighboring set, the participation weight of each historical feature group is analyzed, and the participation weight of the i-th historical feature group is recorded as W i , W i =exp(-D i 2 / h 2 ); where D i represents the dynamic time warping distance between the i-th historical feature group and the feature group to be predicted, and h represents the preset participation coefficient; S203: Obtain the historical cooling amount corresponding to each historical feature group in the neighbor set, and record the historical cooling amount of the i-th historical feature group as Q i , based on the historical feature group, historical cooling amount and corresponding participation weight, the regression coefficient set {β0, β1, β2} is solved by minimizing the weighted residual sum of squares; A regression model is constructed based on the regression coefficient set; the feature group to be predicted X is substituted into the regression model to output the predicted cooling amount Q.

4. The cooling tower adaptive control method based on artificial intelligence according to claim 3 is characterized in that: In S3, include the following: S301: extracting the set target cooling capacity and calculating the cooling capacity deviation in real time based on the predicted cooling capacity; calling pre-stored controller parameters according to the cluster category corresponding to the feature group to be predicted; the controller parameters include fan control parameters and water pump control parameters; the fan control parameters and water pump control parameters each include a proportional coefficient and an integral coefficient; S302: Analyzing the real-time fan frequency increment Δf and the real-time water pump flow increment Δq based on the cooling capacity deviation and the corresponding controller parameters; The fan frequency and pump flow execution quantities of the previous control cycle are read, the new control quantity is analyzed in combination with the real-time fan frequency increment and the real-time pump flow increment, and the new control instruction is sent to the actuator.

5. The cooling tower adaptive control method based on artificial intelligence according to claim 4 is characterized in that: In S4, include the following: S401: Collect the actual cooling amount in real time and synchronously record the normalized current feature group. Subtract the predicted cooling amount from the actual cooling amount and record it as the model prediction residual. Establish a first-in-first-out data buffer of length N. Store the current feature group and the corresponding model prediction residual as a data pair in the buffer. When the number of data pairs in the buffer exceeds N, remove the earliest historical data pair and obtain the most recent N data pairs to form a dynamic data set. Based on the dynamic data set, recursively update the regression coefficient set and controller parameters using the least squares method. S402: Feeding back the updated regression coefficient set to the regression model to replace the original regression coefficients, and feeding back the updated controller parameters to the controller to replace the original controller parameters.

6. An artificial intelligence-based cooling tower adaptive control system, wherein the system is applied to the artificial intelligence-based cooling tower adaptive control method according to any one of claims 1 to 5, characterized in that: The system includes a data acquisition clustering module, a cooling capacity prediction module, a control instruction analysis module and a parameter update module; The data collection and clustering module is used to synchronously collect historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency and historical water pump flow rate, and perform cluster analysis based on the distance of time series segments after normalizing the historical data; The cooling amount prediction module is used to analyze the clustering category of the feature group to be predicted, analyze the neighboring set of the feature group to be predicted; analyze the participation weight of each historical feature group in the neighboring set, and construct a regression model based on the participation weight; The control instruction analysis module is used to analyze the cooling amount deviation between the predicted cooling amount and the set target cooling amount, analyze the new control amount in combination with the cooling amount deviation and controller parameters in the current control cycle, and send the new control instruction to the actuator; The parameter update module is used to collect the actual cooling amount in real time and synchronously record the normalized current feature group, analyze the model prediction residual, establish a data buffer, and update the regression coefficient set and controller parameters based on the dynamic data set.

7. The artificial intelligence-based cooling tower adaptive control system according to claim 6, characterized in that: The data acquisition and clustering module includes a cooling tower data acquisition unit and a data clustering unit; The cooling tower data acquisition unit is used to obtain historical ambient temperature, historical wet-bulb temperature, historical inlet and outlet water temperature difference, historical fan frequency and historical water pump flow rate to generate a historical data set; each historical data is normalized according to the corresponding minimum and maximum values; Get the normalized historical data set; The data clustering unit is used to intercept continuous time series segments along the time axis based on the normalized historical data set, using a sliding window of preset length and step size, and calculate the dynamic time regularization distance between any two time series segments; repeat the sample point allocation and cluster center point iterative operations until the positions of all cluster center points remain unchanged for two consecutive iterations or reach a preset maximum iteration round threshold; generate and store clustering result data.

8. The artificial intelligence-based cooling tower adaptive control system according to claim 6, characterized in that: The cooling amount prediction module includes a participation weight analysis unit and a prediction model construction unit; The participation weight analysis unit is used to obtain the normalized feature group to be predicted and analyze the corresponding cluster category; analyze the dynamic time warping distance between the feature group to be predicted and the historical feature group at each historical moment in the corresponding cluster category; sort the distances in ascending order to construct a neighbor set, and analyze the participation weight of each historical feature group based on the dynamic time warping distance between the feature group to be predicted and the historical feature groups in the neighbor set; The prediction model construction unit is used to obtain the historical cooling amount corresponding to each historical feature group in the neighbor set, solve the regression coefficient set by minimizing the weighted residual square sum based on the historical feature group, the historical cooling amount and the corresponding participation weight, and construct a regression model based on the regression coefficient set.

9. The artificial intelligence-based cooling tower adaptive control system according to claim 6, characterized in that: The control instruction analysis module includes a controller parameter calling unit and a control instruction analysis unit; The controller parameter calling unit is used to extract the set target cooling amount and calculate the cooling amount deviation in real time based on the predicted cooling amount; call the pre-stored controller parameters according to the cluster category corresponding to the feature group to be predicted; the controller parameters include fan control parameters and water pump control parameters; the fan control parameters and water pump control parameters each include a proportional coefficient and an integral coefficient; The control instruction analysis unit is used to analyze the real-time fan frequency increment and the real-time water pump flow increment based on the cooling amount deviation and the corresponding controller parameters; read the fan frequency and water pump flow execution amount of the previous control cycle, analyze the new control amount in combination with the real-time fan frequency increment and the real-time water pump flow increment, and send the new control instruction to the actuator.

10. The artificial intelligence-based cooling tower adaptive control system according to claim 6, characterized in that: The parameter updating module includes a parameter analyzing unit and a parameter transmitting unit; The parameter analysis unit is used to collect the actual cooling amount in real time and synchronously record the normalized current feature group, subtract the predicted cooling amount from the actual cooling amount as the model prediction residual, establish a first-in-first-out data buffer, store the current feature group and the corresponding model prediction residual as a data pair in the buffer, and obtain a dynamic data set; Recursively updating the regression coefficient set and controller parameters using the least squares method based on the dynamic data set; The parameter transmission unit is used to feed back the updated regression coefficient set to the regression model to replace the original regression coefficient, and feed back the updated controller parameters to the controller to replace the original controller parameters.

Images